Developing Metalworking Fluids for Titanium Cutting

Research Collection

Doctoral Thesis

Author(s): Meier, Linus

Publication Date: 2020

Permanent Link: https://doi.org/10.3929/ethz-b-000413413

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library DISS. ETH NO. 26546

Developing Metalworking Fluids for Titanium Cutting

A thesis submitted to attain the degree of DOCTOR OF SCIENCES of ETH ZURICH (Dr. sc. ETH Zurich)

presented by LINUS MEIER

MSc ETH Mechanical Engineering

born on 17.04.1992 citizen of Rafz (ZH)

accepted on the recommendation of Prof. Dr.-Ing. Dr. h.c. K. Wegener, examiner Prof. Dr.-Ing. habil. V. Schulze, co-examiner

2020

Acknowledgements

This thesis was written during my occupation as a research assistant at the Institute of Machine Tools and Manufacturing (IWF) at ETH Zurich. Most of the research was conducted as part of a research project between the company Blaser Swisslube and IWF, which was supported by innosuisse. Foremost, I would like to express my deepest gratitude to my supervisor, Prof. Wegener. He was able to quickly grasp the challenges of my research and gave great advice, while at the same time leaving enough freedom for me to pursue my own ideas. A special thank-you goes to my co-examiner Prof. Schulze. He gave me valuable advice concerning this work and the corresponding publications. I am also very grateful for the collaboration with Blaser Swisslube. I gained deep insights in the industrial development of metalworking fluids and I was supported by the staff while doing basic research. I would like to specially mention Simon Balz, who performed many of the cutting tests and gave advice on test design; Manfred Schneeberger, who led the project; Dr. Michael Eglin, who shared his knowledge about neat oil development; Dr. Olivia Bossard, who shared her knowledge about emulsion development; Marco Tognali, who planned and prepared many of the required test fluids and Dr. Niklaus Ruttimann,¨ who coordinated the industrial and scientific research. I thank my colleagues at ETH Zurich: my longtime office colleagues Martina Spahni and Sebastian Böhl for the fruitful discussions, Jona Engel for the advice in topics related to material science, Knut Krieger for the introduction in metallurgical lab techniques, Philipp Stoll for the advice in designing for additive manufacturing and the manufacturing of a part in SLM, the workshop team Sandro Wigger and Albert Weber for their excellent manufacturing skills, Jens Boos for the advice on process control, Dr. Oscar Laurent for the introduction into advanced surface analytics, PD Dr. Jean-Paul Kunsch for the advice in fluid dynamics and Prof. Bernhard Elsener for the advice in corrosion phenomena. I thank Michael Bettati for his excellent proofreading, including both linguistic and technical advice.

III Finally, I thank my partner, family and friends for the support during the preparation of this work.

Linus Meier Zurich, October 2019

IV Contents

1 Introduction 1 1.1 Introduction to Metalworking Fluids ...... 1 1.2 TasksofMetalworkingFluids ...... 2 1.3 CategoriesofMetalworkingFluids...... 3 1.4 Composition Metalworking Fluids ...... 4 1.5 IntroductiontoTitanium...... 4

2 State of the Art 7 2.1 Development of Metalworking Fluids ...... 7 2.1.1 MetalworkingFluidEﬀect ...... 7 2.1.2 Additives ...... 9 2.1.3 BaseFluids ...... 11 2.1.4 TestingMethods ...... 12 2.1.5 MWFSelectionStrategies ...... 13 2.2 Tribometers ...... 14 2.3 Titanium Cutting Process, Tools, and Coatings ...... 16 2.4 CuttingForces ...... 19 2.5 ToolWear...... 23

3 Task and Aim 29

4 Turning Tests 33 4.1 TurningTestBench...... 33

V 4.1.1 Machine ...... 33 4.1.2 Sensors...... 34 4.1.3 DataAcquisition ...... 36 4.2 ReferenceTurningProcess ...... 38 4.3 OrthogonalTurningProcess ...... 41 4.4 CuttingForcesinTurning ...... 42 4.5 ChipFlowModel ...... 43 4.6 Reducing Variance in Turning Tool Life ...... 49 4.6.1 Illuminationstage...... 50 4.6.2 InterpolationoftheWearTrend ...... 51 4.6.3 AutomaticWearMeasurement...... 52 4.6.4 Assessment of the Initial Cutting Edge Micro-Geometry with Opti- calMeasurement ...... 53 4.6.5 Assessment of the Initial Cutting Edge Micro-Geometry with Pro- cessForces...... 58 4.7 3DWear...... 59 4.8 NewToolLifeCriteria ...... 61 4.8.1 ForceCriterion ...... 61 4.8.2 3DWearCriterion ...... 62 4.9 ExperimentalResults...... 64 4.10Conclusion...... 64

5 Milling Tests 67 5.1 MillingTestBench ...... 67 5.1.1 Machine ...... 67 5.1.2 Tools...... 67 5.1.3 SensorsandDataAcquisition ...... 68 5.2 ReferenceMillingProcess ...... 69 5.3 DataProcessing...... 69 5.4 WearCriteria ...... 72

VI 5.4.1 Automatic Wear Land Measurement ...... 72 5.4.2 ForceCriteria ...... 74 5.5 Reducing Variance in Milling Tool Life ...... 76 5.5.1 Application of Force-Based Criteria ...... 76 5.5.2 Single-Flute Finish Milling Tests ...... 77 5.5.3 Single-Flute Rough Milling Tests ...... 80 5.5.4 RobustDataEvaluation ...... 81 5.6 Conclusion...... 85

6 In-process Tribometer 87 6.1 History...... 87 6.2 Design...... 90 6.2.1 Requirements ...... 90 6.2.2 Solutions...... 92 6.3 Specifications ...... 94 6.4 HeatFlowModel ...... 96 6.4.1 ModelSetup...... 96 6.4.2 ModelResults...... 97 6.5 WorkpieceSurfaceProperties ...... 98 6.6 OperatingModes ...... 100 6.7 ExperimentalResults...... 103 6.7.1 ComparisonofDifferentSetups ...... 103 6.7.2 InfluenceoftheLoad ...... 105 6.7.3 TransientBehavior ...... 106 6.7.4 InfluenceoftheMWFComposition ...... 106 6.7.5 InfluenceofthePin...... 110 6.7.6 Temperature in the Friction Zone ...... 116 6.8 FrictionModels ...... 119 6.8.1 AnalyticalDryFrictionModel ...... 119 6.8.2 LubricatedFrictionModel ...... 123

VII 6.9 Conclusion...... 128

7 Wear Mechanisms 129 7.1 AnalysisofCuttingTools ...... 129 7.1.1 ExperimentalSetup...... 129 7.1.2 Results...... 131 7.2 HighTemperatureOxidationTest ...... 133 7.2.1 ExperimentalSetup...... 133 7.2.2 Results...... 134 7.3 PinWearOil ...... 139 7.4 PinWearEmulsion ...... 142 7.5 DissolutionoftheCobaltBinder...... 143 7.6 OxidationofTungstenCarbide ...... 145 7.7 Decarburizationoftungstencarbide ...... 145 7.8 CrackFormation ...... 147 7.9 CombinedWearModel ...... 147 7.10Conclusion...... 149

8 Summary and Outlook 151 8.1 WearMechanisms...... 151 8.2 Metal Working Fluids in Continuous Cutting ...... 152 8.3 Metal Working Fluids in Interrupted Cutting ...... 153 8.4 WearAnalogyTests ...... 154 8.5 ToolLifeTesting ...... 156 8.6 Developing Metalworking Fluids ...... 157 8.7 Other Inﬂuences on the Cutting Productivity ...... 158

VIII Symbols and Abbreviations

Notations

Generally, vectors are represented by underlined small letters, e.g. g. Matrices are represented by double-underlined capital letters, e.g. H.

Latin Symbols

A Undeformed chip cross-section area

Apolar Polar plot area ac Undeformed chip thickness ae Radial depth of cut ap Axial depth of cut aw Chip width C Constant

C1 Constant

C2 Constant

Cp Process capability D Diﬀusion constant D Tool diameter

EA Activation Energy E∗ Equivalent elastic modulus F Load f scalar ﬁeld

Fc Cutting force

Fchip Force exerted by the chip normally to the rake face

Ff Feed force

Fmax Peak force during a cut

IX Fmax,0 Initial peak force

FN Normal force

Fstiction Stiction force

FT Tangential force fz Feed per tooth H Hardness H Hessian matrix h Convective heat flow coefficient iu.r. Constant relating force to the deviation in chip flow speed k Rate of force increase kA Constant relating chip load to chip area kc Contribution of the cutting edge to the cutting force kf Contribution of the cutting edge to the feed force kf Transversal stiffness ku.r. Constant relating force to the deviation in chip flow angle L Sliding length m Equivalent mass mfall Falling slope of the cutting force mrise Rising slope of the cutting force n Sample size n Spindle revolution speed ′ Pcut,wp Specific heat conducted in the workpiece in cutting ′′ Ppin Heat density generated by friction p Pressure, equivalent to the normal stress at a surface pcut Force integral over time, equal to the impulse pmax Maximum pressure Q Material volume loss due to adhesive wear R Gas constant R Spherical radius R Cutting edge radius, constrained

Rn Cutting edge radius, unconstrained r Pearson’s correlation coeﬃcient S Skewness

Sα Distance from the ideal cutting edge to the point where the real cutting edge meets the ideal ﬂank face

Sγ Distance from the ideal cutting edge to the point where the real cutting edge meets the ideal rake face T Tool life

X teng Engagement time t0 Time of force increase

Vtot Total material volume removed v Speed vc Cutting speed vf Feed rate vrange Width of the transition range vtrans Transition speed VB Flank wear land width

VBmax Maximum ﬂank wear land width W Wear depth

Wadh Adhesive wear depth z Number of cutting edges of a milling tool

Greek Symbols

α Clearance angle of a cutting tool

αeng Engagement angle of a milling tool β Wedge angle of a cutting tool γ Rake angle of a cutting tool

∆µFM Reduction in the coeﬃcient of friction through additives ∆r Distance of the ideal to the real cutting edge η Chip ejection angle

ηwp Fraction of the cutting energy conducted into the workpiece κ Tool entering angle θ Absolute temperature θ Cutting edge lead angle, equivalent to 90 degree − κ

θenv Absolute temperature of the environment λ Eigenvalue

λs Inclination angle of the cutting edge µ Coeﬃcient of Friction

µcomb Combined coeﬃcient of Friction

µdry Quasi-dry coeﬃcient of Friction

µlub Lubricated coeﬃcient of Friction

τmax Maximum shear stress

XI Abbreviations

ASTM American Society for Testing and Materials AW Anti Wear BIB Broad Ion Beam BUE Built-Up Edge CAD Computer-Aided Design CAM ComputerAidedManufacturing CBN Cubic Boron Nitride CFRP Carbon Fiber Reinforced Plastic CNC Computer Numerical Control DAQ Data Acquisition DIN Deutsches Institut fur¨ Normung (German Institute for Standardization) EDM Electrical Discharge Machining EDTA Ethylenediaminetetraacetic Acid EDX Energy Dispersive X-Ray Spectroscopy EP Extreme Pressure ETH Eidgen¨ossische Technische Hochschule (Swiss Federal Institute of Tech- nology) FM Friction Modiﬁer FSPT Full Scale Production Test GUI Graphical User Interface HSC High-Speed Cutting HSK Hohlschaft-Kegel (Hollow Shank Taper) IWF Institut fur¨ Werkzeugmschinen und Fertigung (Institute of Machine Tools and Manufacturing) MQL Minimum Quantity Lubrication MRR Material Removal Rate MSAC M-Estimator Sample Consensus MWF Metalworking Fluid PI Proportional-Integral PLC Programmable Logic Controller PVD Physical Vapor Deposition RANSAC Random Sample Consensus Algorithm SEM Scanning Electron Microscope SLM Selective Laser Melting TCO Total Cost of Ownership

XII XRD X-Ray Diﬀraction XRF X-Ray Fluorescence

XIII XIV Abstract

Economical titanium cutting remains a challenge in manufacturing, especially with the titanium use rising. The productivity is low and tool wear is high. Metalworking fluids have a major influence on the tool wear. The working mechanism of metalworking fluids is poorly understood. New products are therefore developed in expensive trial-and-error cycles. The development process of new metalworking fluids for titanium should be made easier, more target-oriented, and less expensive. In this thesis, the whole fluid development chain from concept to machining tests with the final product is analyzed and optimized. The causes of tool wear in titanium cutting are investigated. Therefrom, the required properties of the metalworking fluid to mitigate wear are derived. Tribological laboratory tests are developed. As they are not able to replace the testing of a newly developed metalworking fluid in a real cutting situation, cutting tests are optimized as well. The test repeatability has one of the biggest leverages on the expenses of a full-scale cutting test. Disturbances are therefore identified and eliminated. It is found that notch wear can be attributed to tribo-chemical binder degradation. Crater wear in contrast can be attributed to the dissolution of carbide grains. Strategies to mitigate both types of wear are discussed. The tendency for notch wear can be assessed with an in-process tribometer. In full scale turning tests, the tool life standard deviation could be reduced by a factor of three by compensating the influence of the initial cutting edge radius. In full scale milling tests, the resolution of the tool life measurement could be increased at a decreased workpiece material consumption by using a force-based tool life criterion and single-fluted milling tools. The exact conditions a metalworking fluid meets close to the cutting edge remain unclear. Therefore, full-scale cutting tests are still needed before launching a new metalworking fluid. The methods to evaluate metalworking fluids are developed in an industrial environment and are directly adaptable by metalworking fluid producers. The methods are easy to use and can be automated.

XV Zusammenfassung

Obwohl Titanlegierungen immer häufiger eingesetzt werden, bleibt ihre wirtschaftliche Be- arbeitung wegen des Werkzeugverschleisses schwierig. Kuhlschmierstoffe¨ können den Ver- schleiss stark verringern. Der Wirkmechanismus ist dabei ungenugend¨ verstanden, weshalb neue Kuhlschmierstoffe¨ teuer empirisch entwickelt werden. Die Entwicklung von neuen Kuhlschmierstoffen¨ fur¨ Titan soll daher einfacher, zielgerich- teter und gunstiger¨ werden. In der vorliegenden Arbeit wird die Entwicklungskette vom Konzept bis zum Zerspa- nungstest mit dem fertigen Produkt umfassend analysiert und optimiert. Die Verschleiss- mechanismen werden genauer untersucht und daraus die benötigten Kuhlschmierstoffei-¨ genschaften abgeleitet. Einfache Tribologie-Labortests werden entwickelt, welche aber die Zerspanungsversuche nicht vollständig ersetzen können. Deshalb werden die Zerspanungs- versuche ebenfalls optimiert. Die Wiederholbarkeit der Zerspanungsversuche, als grösster Einfluss auf die Kosten, wird durch die Eliminierung von Störgrössen verbessert. Als Ergebnis kann Kerbverschleiss auf tribo-chemischen Angriff des Binders zuruckge-¨ fuhrt¨ werden. Der Kolkverschleiss hingegen wird durch die Auflösung der Karbidkörner bestimmt. Strategien, um beide Verschleissmechanismen zu unterdrucken,¨ werden vorge- stellt. Mit einem In-Prozess Tribometer können Ole¨ auf ihre Kerbverschleissneigung uber-¨ pruft¨ werden. In Drehversuchen kann die Streuung um den Faktor drei reduziert werden, wenn man den Einfluss der Schneidkantenradien kompensiert. Bei Fräsversuchen kann bei gesenktem Materialeinsatz die Auflösung der Standzeitmessung verbessert werden, wenn ein kraftbasiertes Standzeitkriterium verwendet wird. Die genauen Bedingungen, die nahe der Schneidkante herrschen, bleiben unklar. Deshalb werden in der Kuhlschmierstoffentwicklung¨ immer noch abschliessende Zerspanungsversu- che notwendig sein. Die neu entwickelten Methoden zur Kuhlschmierstoffbewertung¨ entstammen einem indus- triellen Umfeld, sind deshalb leicht anwendbar und können einfach automatisiert werden.

XVI 1

Chapter 1

Introduction

1.1 Introduction to Metalworking Fluids

Metalworking fluids (MWFs) were used since the early days of metalworking to improve machining processes. A first reference was given by Vitruvius in 31 B.C., where the use of olive oil is mentioned for turning pistons and cylinders [78]. With the development of grinding machines in the 16th century, water became more important as a metalworking fluid. Johannes Stradanus depicted water chutes leading to grinding wheels in a copper engraving from 1575. With the industrial revolution and the discovery of petroleum in the 19th century, MWFs gained influence and at the same time, the first scientific investigations showed increased productivity when using MWFs. In 1915 the first water-miscible MWFs were made commercially available [30]. With the development of harder and more heat-resistant tool materials, tool loads were continuously increased. To still be effective under these conditions, MWFs have been continuously enhanced to the present day. Some effective additives from the early days are outdated, some are even forbidden due to toxicity and environmental impact. On the other hand, new promising additives are developed on a daily basis. Some processes require the use of MWFs. For instance, deep-hole drilling requires a well- lubricating oil to prevent adhesion to the guide pads in certain materials. Additionally, according to Byers [30], insurances may require the use of MWFs in the case of titanium cutting to mitigate fire hazards. In most cases however, MWFs have to compete against dry cutting, minimum quantity lubrication (MQL), and cryogenic cooling. More than ever, the total costs of ownership (TCO) are of importance to users. These costs include the costs of acquisition and maintenance of MWF conditioning equipment, supervision, additional additives, energy, pumps for delivery, hazardous incident prevention, and disposal costs. Bierma et al. [22] report a ratio of hidden costs ranging from 1.5:1 up to 5.5:1, meaning 2 1. Introduction

that companies spend up to 5.5 times the purchase costs of a metalworking fluid for using the fluid. According to the same study, an improvement larger than a factor of 5 in the benefit-to-purchase cost ratio results from an increase of the fluid performance by 20%. The benefits mostly result from higher productivity and lower tool costs. Accounting for this fact, MWF-producers invested a lot in the development of high performance, high price MWFs. According to Astakhov [10], in the 1980’s, MWFs accounted for less than 3% of the machining costs, whereas in the 2000’s, car manufacturers report more than 30%. On the other hand, according to [23], the cost share of MWFs in costs per part is only 0.5%, if human resources are included. In 2015, the global MWF market had a value of $ 9.91 billion, of which approximately half can be accounted for metal cutting fluids [151]. This corresponds to a total of 1.8 megatons of metal cutting fluid production per year. The market is expected to grow by 0.6% per year [151]. For comparison, the global market for cutting tools is three times the size of that for metal cutting fluids.

1.2 Tasks of Metalworking Fluids

The main tasks of MWFs are to lubricate the contact between tool and workpiece, to cool the tool and workpiece, and to flush away chips. A well-lubricated contact leads to reduced tool wear and can lead to lower friction forces, which in turn reduces the amount of heat that has to be removed with the fluid. Moreover, the surface finish and surface integrity are strongly dependent on the MWF’s lubricational properties. The cooling properties mainly depend on the bulk physical properties of the cutting fluid, namely the specific heat capacity, heat conductivity, viscosity, and density. In some cases, molecular effects such as surface energies or boiling characteristics may play a role. Flushing only depends on the viscosity and density and is important when cutting in confined spaces, e.g. in drilling, tapping, or pocket milling. Even in unconfined conditions, high pressure flushing can help to get shorter chips when turning or milling long-chipping materials. Besides the main tasks, the MWF has to fulfill a number of secondary tasks. It has to prevent natural corrosion of the workpieces and machines. It has to cool the surface of the machine bed, the spindle, and sometimes it is even used to internally cool the machine structure. It has to clean the machine, if possible during machining. Sometimes the MWF is used as a hydraulic fluid as well or to check for tool breakage. The MWF has to meet additional requirements, which are not directly related to its tasks. It must be nontoxic and non-irritating for the skin. It must be non-corrosive for any part of the machine or the workpiece. It has to be compatible with rubber seals and the machine 1.3 Categories of Metalworking Fluids 3

paint. It must not produce foam or mist nor separate itself when used with high pressure. It must be stable over a long period, thus it should inhibit excessive growth of bacteria and fungi, should not use up its surface-active substances too quickly, and should not be carried out with the chips. It should quickly drain from the machine windows without leaving a residue to allow for a quick inspection of the machining result. It should be transparent to allow for inspection during machining and the odor should be pleasant. Finally, it should be easy and harmless to be disposed of.

1.3 Categories of Metalworking Fluids

Four classes of MWFs have been established over the years: neat oils, oil based emulsion, semi-synthetic emulsions, and synthetic fluids. Neat oils are used without any water. They provide the best possible lubrication, although combined with a comparably weak cooling effect. Equipping a machine tool with neat oil leads to high initial costs, as the whole fluid volume has to be bought from the MWF producer. Neat oils lead to low adhesion between work material and tool, low cutting forces, and good surface finish. They have a long sump life due to the lack of growth of bacteria and fungi. Depending on the additives and the base fluid, neat oils can be black to fully transparent. They are mainly used in grinding, deep-hole drilling, Swiss-type turning, and gear cutting. Oil based emulsions are delivered as concentrate and are mixed with water by the customer. Oil percentages from 1% to 20% can be used, however, 4% to 12% are usually recommended. Due to the low oil phase content, the cooling properties are similar to those of pure water. The droplet size is usually in the range of 3 µm, leading to the typical white, milky appearance. Semi-synthetic emulsions are similar to conventional emulsions, but contain a larger fraction of synthetic oil. This allows for the formation of droplets with sub-micrometer size, making the MWF translucent to transparent. Synthetic fluids do not contain an oil phase; they are fully soluble in water. Synthetic fluids do not achieve as good lubricating properties as emulsions. Therefore, they are used in applications where cooling has the highest priority, such as grinding or titanium machining. Synthetic concentrates and MWFs made thereof are usually transparent, having an appearance like water. A more detailed overview is presented by Byers [30] and by Evans [50]. 4 1. Introduction

1.4 Composition Metalworking Fluids

Neat oils are based on mineral oil, synthetic apolar fluids, plant oils, synthetic esters or a mixture thereof. The base fluid, to a large part, determines the viscosity, the solubility of the additives, and the tribological properties at low loads. Friction modifiers (FM) are added to lower friction and wear under boundary friction conditions. They reversibly adsorb to the surface and therefore the molecules have to consist of a polar portion paired with a non-polar tail. Typical examples are esters. Anti-wear (AW) additives are added to lower wear at medium friction conditions. They may form a protective film when loaded. Extreme-pressure (EP) additives are added to lower wear at severe conditions. At higher temperatures or loads, they react with the workpiece and tool material to form a stable separation layer. Not directly connected to the main task are antioxidants, which prevent the oil from getting rancid; corrosion inhibitors; metal deactivators to prevent staining of non-ferrous materials; defoamers; anti-mist; dyes; and fragrances. Concentrates for oil-based emulsions are based either on mineral oil or on ester oil and have similar types of additives as neat oils. However, no anti-mist additives are required. In order to form stable emulsions with water, carefully selected emulsifiers are added. In order to prevent corrosion and inhibit bacterial and fungal growth, an alkaline pH of around 9 is usually targeted in emulsions. The pH is raised by adding alkaline agents. Stable MWFs are also achievable by using neutral corrosion inhibitors and biocides. Concentrates for semi-synthetic emulsions additionally contain water-soluble FM additives. These additives require a water phase in the concentrate, making the development of such formulations more complex. Fully synthetic concentrates only consist of water-soluble FM, AW, and EP additives. A large amount of corrosion inhibitors is needed. A more detailed overview is presented by Byers [30].

1.5 Introduction to Titanium

Titanium is an extraordinary metal: with a density of only 4.81 g/cm3, its alloys can reach tensile strengths of 1250 MPa, making it the metal with the highest speciﬁc strength and therefore an ideal material for lightweight construction. The fatigue strengths are higher than in steel alloys, making it even more interesting for structural aerospace components. A low Young’s modulus combined with the high strength leads to a high elastic energy absorption capacity, making it an excellent material for springs. In addition to the mechanical properties, titanium is exceptionally corrosion resistant through the self-passivation of 1.5 Introduction to Titanium 5

the surface with an oxide layer. Therefore, it is optimally suited for marine applications and in chemical plants. Because of the good biocompatibility, medical implants and bone screws are often made from titanium. Titanium alloys can have a high strength and low creep rates up to temperatures of 600 ◦C, making them suitable for compressor blades of modern gas turbines.

100% 9 7 3 13 12 5 7 6 6 5 5 80% 5 14 4 7 15 18 22

60%

53 40% 75 73 68 67 61 20% 20 0% A350 A300 A310-200 A320-200 A340 A380 A350 1970 1980 1985 1990 2005 2013 Miscellaneous GLARE Steel Titanium Composites Aluminum

Figure 1.1: Material mix in commercial planes developed by Airbus. The plane with the highest titanium share, the A350-XWB, is shown to the right. Image courtesy of Airbus (S.Ramadier).

The interest in titanium and its alloys is continuously rising. As an example, Figure 1.1 shows the massive increase in titanium alloy use which is accompanying the increase in CFRP use in newly developed aircraft. Already today, the aerospace industry accounts for over 60% of the total worldwide titanium alloy use with a further increase estimated [153]. This in turn causes an increasing need for titanium cutting and therefore an increasing need for MWFs suitable for titanium cutting. Despite the advances in cutting tool substrate and coating technology, titanium is still considered as hard-to-cut. The extraordinary strength at high temperatures leads to high cutting forces. The low thermal conductivity causes heat accumulation near the cutting edge. Although titanium is inert in its oxidized state, its nascent surfaces generated in machining react with almost everything that touches them, including the tool. Altogether, this causes high wear rates and limits productivity due to slow machining and frequent tool changes. An overview on the material titanium is given by Lutjering¨ [88]. A book focused on titanium machining is presented by Davim [112]. 6 1. Introduction 7

Chapter 2

State of the Art

2.1 Development of Metalworking Fluids

The development of MWFs is based mainly on experience and is strongly connected to industrial companies. Therefore, relatively few scientific publications are available on the topic. Publications from industrial companies concerning the composition of cutting fluids are seldom fully replicable, as the components are often made anonymous. MWFs are developed in an iterative procedure, combining experience from existing products and test results to achieve an improvement. Many combinations of base fluids and additives are prepared and tested in the laboratory. Some promising candidates are further tested in a full-scale cutting test to determine the composition of a new product. An overview is given by Canter [32] and an exemplary procedure is presented by Abdalla et. al [1]. If the best fluid does not meet all the primary and secondary requirements, the process is repeated with modified compositions. Scientific research provides the companies with new ingredients and concepts for MWFs and the scientific community uses MWFs to investigate machining processes, compare dry, minimum quantity lubrication, cryo, flood cooling and high-pressure delivery strategies. A review on MWF acting mechanisms and new concepts is presented by Brinksmeier et al. [27]. A review focused on new MWF constituents is presented by Osama et al. [111].

2.1.1 Metalworking Fluid Eﬀect

MWFs act by cooling and lubricating the process. While the cooling eﬀect is easily graspable on a macroscopic scale, the lubricating eﬀect is much harder to understand, as the pressure on the tool surface is much higher than in normal lubricated contacts. In the early research on MWFs, two main hypotheses existed. In 1947, Rehbinder [125] in Russia 8 2. State of the art

showed that the adsorption of carbon tetrachloride to the surface of tensile test specimens leads to lower tensile strength. Applying this to chip formation would explain the lower cutting forces when cutting with carbon tetrachloride. In the USA, Merchant and Shaw [97] assumed that the lower cutting forces stem from the formation of a chemical reaction layer with low shear strength between the chip surface and the rake face. In 1958, Shaw [131] further investigated in turning with carbon tetrachloride. He used oleic acid and dry machining as comparison. He measured high coefficients of friction with carbon tetrachloride but still observed low cutting forces. The effect was explained by the smaller chip-curling radius, which leads to a lower cutting force. In 1966, Barlow [16] extended the Rehbinder effect to any surface modification and showed that electro-polished surfaces lead to lower cutting forces. At the same time, he dismissed the Rehbinder effect as the main effect of MWFs, as it only occurred when the liquid was dried up before the deformation occurs. Cassin et al. [33] proposed diffusion through the chip to the cutting edge be the main acting mechanism of carbon tetrachloride. In a cutting chamber with a controlled atmosphere, Wakabayashi et al. [159] proved that the limiting factor of carbon tetrachloride lubrication is not the transport to the surface but the interaction with the surface. The penetration of cutting fluid to the cutting edge in continuous cutting can easily be explained in the case of carbon tetrachloride, as it is a very small and mobile molecule. Postnikov [117] looked for mechanisms with which bigger molecules, such as oil molecules could reach the friction zone. The workpiece and the tool were electrically isolated from ground and the electrochemical potential between tool and workpiece is measured. For this method, the MWF has to be sufficiently conductive, therefore salt solutions with isopropyl alcohol as a surface-active substance were chosen. The main penetration mechanisms were proposed to be through breaking-off of the built-up edge or through vibrations of the cutting tool. Williams et al. [178] measured cutting forces under different controlled environments. They assumed capillaries between the chip and workpiece and showed that smaller lubricant molecules, such as carbon tetrachloride, were able to penetrate further than bigger molecules, such as hexachlorobuta-1,3-diene. Furthermore, they calculated lower penetration depths for vapors than for liquids at cutting speeds lower than 30 m/min. At higher speeds, vapor lubrication becomes more important. Naerheim et al. [102] measured the electrochemical impedance of workpiece material immersed in MWF to quantify the extent of additive adsorption. They found a correlation between the cutting force and the adsorption and concluded that the adsorption of additives is important to lower the cutting forces. Smith et al. [136] presented a theoretical model for cutting fluid penetration depth. Barker et al. [15] investigated emulsions in an EHD test rig. At low rolling speeds, a reservoir consisting of only the oil-phase can form in the rolling contact. At higher speeds, starvation of the oil depot leads to the water-phase 2.1 Development of Metalworking Fluids 9

entering the frictional contact. Madhavan et al. [90] used transparent glass and sapphire tools to cut metal and observe the interface between chip and tool. Contrary to earlier assumptions, they observed sliding even in the intimate contact zone. Ackroyd et al. [3] confirmed the results for more realistic cutting speeds. Huang et. al [63] used the same test setup: They cut lead with a transparent sapphire tool using a luminescent MWF. By measuring the light intensity, they found a linearly increasing film thickness from 0 µm at the cutting edge to 15 µm at the edge of the intimate contact zone on the rake face. Wang et al. [160] used an analytical model for the laminar flow into the cutting zone. They showed a limited penetration depth for conventional jet cooling due to the chip motion away from the cutting edge. Emulsion spray and cryogenic gases were able to penetrate much further. Inversely fitted friction models for a FE-simulation of the cutting process by Banerjee et al. [14] showed a lower coefficient of friction for MQL than for flood cooling, thereby supporting the theory of better penetration.

2.1.2 Additives

Over the years, many chemical compounds were screened for their use as additives in MWFs. Watanabe et al. [104, 148, 164–176] tested a whole range of promising new additives for their surface tension, coefficient of friction and corrosion inhibition, however, no tool wear tests are performed. Vasyshak et al. [157] added different polymers to a base oil. The drill time in a force-controlled operation was reduced at the cost of a higher torsional moment. Soshko et al. [139] used polyvinyl chloride as an additive with the result of longer tool life and an improved surface finish. Jain et al. [66] tested different inorganic salts for their activity as EP additives. Bataller et al. [19] gave the complete composition of an emulsion without anionic surfactants and tested its stability and droplet size. A similar study was conducted by John et al. [70], investigating different oil and surfactant combinations for their stability. Cambiella et al. [31] gave a complete description of different bench test emulsions to study the interaction between emulsifiers and EP-additives. Schulz et al. [127] and later Huesmann-Cordes et al. [64] developed a theory considering the different oxide to hydroxide ratios on the surface of different alloys. They found that the ratio between additives acting via ionic and van der Waals bonds plays a bigger role than the shear amount of each additive. They proved their theory in a Brugger setup. Minfray et al. [99] studied the influence of organic pentasulfide, an additive that has already been in widespread use at the time. Yan et al. [184] investigated novel compounds containing phosphorous and boron and found a reduced coefficient of friction and lower wear in a four ball tester. Zhang et al. [187] tested the feasibility of lignin as an additive and found lower cutting forces and better surface quality than with conventional emulsion. The results were later confirmed by Mu et al. [101]. Ionic liquids, 10 2. State of the art

either as base fluid or as additives, have gained interest in recent years. In 2013, Libardi et al. [85] considered using ionic liquids as MWFs and tested the lubricating properties. Pham et al. [115] used ionic liquids for micro milling with good results. They highlighted the low evaporation rate, low flammability and possible biodegradability. Later studies used ionic liquids as additives to a base oil and report major reductions in friction force even at low concentrations [5, 6, 55, 56]. A review on ionic liquids as MWFs was given in [7]. Polymers were discussed as a single additive in water by Winter et al. [179, 180]. A completely new approach was followed in the group of Brinksmeier [98, 123, 124]: Microorganisms, which are normally unwanted in an MWF, were selectively cultivated to act as lubrication enhancement. The organisms produced surface-active substances, such as fatty acids and polymeric compounds. Under light friction conditions, the cells as a whole could prevent the frictional partner from touching each other, whereas at higher loads, the microorganisms were crushed, further releasing surface-active substances. Solid lubricants are successfully used in metal forming to deal with the extreme pressures. Their use in cutting is discussed as well. Bagchi et al. [12, 13] reported better surface finish with solid lubricants dispersed in water or oil. Suresh Kumar Reddy et al. [143] supplied dry graphite or molybdenum disulfide powder to a milling process and achieved lower cutting forces than in wet cutting. Nageswaro Rao et al. [103] investigated the influence of particle size in solid dry lubrication of a turning process. Micrometer-sized particles require additional technical measures to be kept in suspension and cannot penetrate far into the wedge-shaped gap between chip and tool or between tool and workpiece. Furthermore, the suspensions are opaque, which limits the visibility for the machine tool operator. This leads to an intensified investigation of nanometer-sized particles as additives for MWFs in the recent years. Due to the large surface area, nanoparticles can

behave diﬀerently than the bulk material. For example, even though Al2O3 is considered an abrasive as bulk material, Vasu et al. [156] reported less tool wear when turning with

Al2O3-nanoparticles. Four main working mechanisms were discussed in [82]: a rolling or ball bearing effect where the nanoparticles keep the surfaces apart and lead to lower rolling friction, shown in Figure 2.1a; a continuously replenished protective film on the tool, shown in Figure 2.1b; a mending effect where surface defects in the friction partners are filled with coagulated nanoparticles and thereby increase the load bearing area, shown in Figure 2.1c; or a polishing effect where the nanoparticles continuously remove asperities on the friction partners, shown in Figure 2.1d. Srikant et al. [140] concentrated on the thermal properties of nanofluids. They calculated the heat transfer coefficient for different concentrations of copper oxide nanoparticles in water and found a maximum increase of 18% at a concentration of 6%, leading to a temperature decrease of 30 K in their case. Sharma et al. [129] presented a review on the effects of nanoparticles in machining pro- 2.1 Development of Metalworking Fluids 11

cesses. A comprehensive review on all aspects of metal cutting with nanoﬂuids was given by Singh et al. [135]. a b

c d

Figure 2.1: Proposed working principles of nanoparticles [82].

2.1.3 Base Fluids

Although the first MWFs were water, vegetable oil, or animal fat, they were quickly replaced by mineral oils in the petroleum era and later complemented by mineral oil based emulsions. In the recent years, more emphasis is put on ecologically sustainable metal cutting. According to Lawal et al. [80], the environmental impact of vegetable oils and animal fats is lower than the one of mineral oils, but their price is significantly higher. With the current trend to minimize metalworking usage with MQL, the price of the MWF plays a lower role than in conventional flood cooling. Other than the good biodegradability, vegetable oils have a good compatibility with additives, low toxicity and a high flash point [130]. The polar ester bond leads to an anti-wear behavior even without additives and makes vegetable oils ideal for MQL fluids. The properties can be further tailored by transesterification with a different alcohol than glycerol, leading to synthetic vegetable oils. Suda et al. [142] demonstrated the good lubricating properties, high biodegradability, and oxidation stability of polyol esters. Xavior et al. [183] showed benefits of pure coconut oil in tool wear compared to a commercial neat oil. Ozcelik et al. tested different fluids based on canola and sunflower oils and found much lower wear than with two unspecified commercial oils. Lawal et al. [81] reported good results with an emulsion based on vegetable oil. Burton et al. [29] prepared a canola oil in water emulsion in-situ without any surfactant and supplied it as MQL to the cutting zone, leading to a lower cutting force in micro milling than with a commercial emulsion. Bork et al. [25] compared MWFs based on jatropha oil, synthetic jatropha oil, canola oil, and mineral oil 12 2. State of the art

to conclude that jatropha oil leads to the longest tool life and the lowest surface roughness. Sashidhara et al. [130] reviewed the development and testing of vegetable-based MWFs. Lawal et al.[79, 80] prepared two reviews on the application of vegetable based MWFs. Debnath et al. [43] reviewed diﬀerent types of environmentally friendly MWFs and cooling techniques.

2.1.4 Testing Methods

Corresponding to the wide range of functions of MWFs described in Section 1.2, many different methods for testing MWFs exist. For each different function, e.g. reduction of tool wear, different tests with various costs exist, e.g. a tribometer test or a full-scale production test. Various authors [38, 41, 87, 154] pointed out that cheap laboratory tests have only a limited relevance, which is defined as the correlation with the full-scale production test results. De Chiffre et al. compared a wide range of MWF tests. In an early work [38], de Chiffre et al. assumed the same ranking of different MWFs in every possible test, i.e. for example the MWF leading to the lowest cutting force also leads to the lowest surface roughness and to the longest tool life. This allows for an easy comparison of different tests after assessing their reproducibility, variability range and cost. Cutting force tests proved to have the best ratio between resolution and cost. In a later study [39], the same authors tried to verify the assumption of invariant ranking for emulsions but observed different rankings in different machining tests for neat oils. However, according to another study [20] of the same authors, at least the thrust force of unworn drills correlates well with the tool life when drilling austenitic steel with different neat oils. In a further step, the constant ranking assumption was replaced by the correlation with the real performance value. Belluco arranged the methods in a 2D-diagram shown in Figure 2.2, featuring both the cost and the correlation between the test results and the real performance. Popov et al. [116] developed a special set of turning parameters to specifically target the anti-adhesion properties of an MWF. They proposed to use this quick test between Reichert tests and full-scale cutting tests. Axinte et al. [11] presented an error budget for tool life tests, depending on the parameters used. Using this data, they could reduce the number of necessary tests to determine the parameters of Taylor’s law reliably to a minimum. The tapping torque test is a popular test for MWFs, as it bridges the gap between simple tribometers and cutting tests. It is typically performed on tapping machines as the one shown in Figure 2.3. Even though a cutting process is involved, tapping torque tests can still be performed in a conventional tribology laboratory. The good correlation with the real performance is confirmed by many researchers in the past decades 2.1 Development of Metalworking Fluids 13

Figure 2.2: 2D-diagram of diﬀerent MWF tests by Belluco [21]. Tribological analyses (Trib.An.) are cheap but have a low correlation with the performance. Full-scale production tests (FSPT) are expensive but do correlate well with performance.

[21, 32, 40, 41, 87, 188]. The ASTM D 5619 standard, treating tapping torque tests, has been withdrawn due to lack of interest. It stated: “[the tapping test] method can be used to more accurately predict the lubricating properties of a metal removal fluid than previously available laboratory scale tests”. Up to now, no standard has established on how to perform such tests, therefore, according to Zimmerman et al. [188], the results of different test series are hard to compare. The tapping torque test is not applicable for titanium cutting, as entangled chips and cold welding lead to large fluctuations in torque, sometimes even causing tool breakage.

2.1.5 MWF Selection Strategies

When selecting a new MWF, the tools, cutting parameters, and workpiece materials are usually defined first. Then, a number of different MWFs are tested for the fulfillment of the different tasks. New methods to reduce the required amount of tests are proposed by the scientific community. Different multicriteria decision making (MCDM) methods are applied to simplify the selection process. For example, Rao et al. [121] used a digraph 14 2. State of the art

Figure 2.3: Tapping torque test setup [152]. and matrix method to select the appropriate cutting fluid. Jagadish et al. [65] used the multi-objective optimization on the basis of simple ratio analysis (MOOSRA). Prasad et al. [118] applied the quality function deployment (QFD) method to generate a task-specific suitability ranking for a database of MWFs. It generated reliable results independent of user intuition. These methods may help an MWF customer to decide whether an MWF is suited for a specific purpose. However, the methods do not provide any guidance for developing a new MWF.

2.2 Tribometers

Traditionally, tribometers were developed to emulate friction and wear conditions of machine elements such as bearings, pistons and gears. These elements are usually designed to achieve a lifetime of several hundred hours and therefore are designed to feature light friction conditions. For this reason, most of the tribometers are only designed for light friction conditions. Of the conventional bench tribometers, only a few are used in MWF testing [32, 71]: The Brugger Test, according to DIN 51347, is used to measure the load bearing capacity of neat oils. It uses two steel rolls with a limited signiﬁcance for the tool-workpiece contact. The similar, but non-standardized, Reichert Test [72] is used for water-miscible ﬂuids. The Shell Four Ball Test, according to ASTM D 2783 and DIN 51350, provides similar information as the Brugger Test and is more common in the Anglo-Saxon 2.2 Tribometers 15

region. Usually, only the amount of wear after the tests described above is evaluated. The measurement of the coefficient of friction requires more expensive test devices. In order to better understand tribological effects in metal cutting, special cutting tribometers were developed to emulate the tribological interaction between workpiece, chip, and tool. Cutting tribometers are usually designed as open tribometers, i.e. the friction partner representing the tool is always producing a new track, never reusing a previous track. This prevents the formation of a protective layer over multiple passes. Cutting tribometers should allow for severe friction conditions, i.e. the achievable contact stresses should be in the same range as the yield stress. Hedenqvist et al. [61] used a cylinder of workpiece material on a lathe. The lateral surface of a cylinder made of tool material was pressed radially against the surface of the workpiece as shown in Figure 2.4b. This setup is simple in design and allows for a great range of different parameters to be tested. It is however not suitable to measure the coefficient of friction on a nascent surface. While the workpiece was turning, the tool cylinder was moved axially to create spiral track. Rech et al. [122] used a similar setup, but the cylindrical surface of the tool was replaced by a spherical shape as shown in Figure 2.4e. It has similar advantages and limitations as the design by Hedenqvist. Another design, even closer to cutting, was presented by Puls et al. [119, 120]. The flank face of a cutting insert was pressed against the workpiece with an extremely negative rake angle (−80°) as shown in Figure 2.4c. This suppresses chip formation and allows for an easy separation of frictional forces from metal forming forces. The tribometer can easily be equipped with temperature sensors and a high-speed camera. The setup is not sensitive to the oxide layer on titanium, as a semi-solid titanium paste is formed at the interface. This tribometer’s is however restricted to severe friction conditions. Especially when cutting quickly oxidizing metals such as aluminum, stainless steel or titanium, open tribometers do not emulate the situation in cutting well. In open tribometers, the surface is prepared minutes before the actual friction test is started. This allows for the formation of an oxide layer, which is not comparable to machining, where no oxide layer is present between the workpiece, chip and tool. Therefore, Olsson et al. [107] developed the in-process tribometer shown in Figure 2.4a. A longitudinal face turning process using a tool with entering angle of 90° generated an even, freshly cut surface. Immediately behind the cutting tool, the friction probe was pressed on the freshly generated surface, thereby limiting the effect of oxidation. Olsson only presents data for AISI 4340 steel and it remains unclear if the setup was suitable for titanium cutting. In-process tribometers are not in widespread use, as they have a very specific use case and are tedious to use. Zemzemi et al. [186] as well as Smolenicki et al. [137] presented in-process tribometers based on the principle of Olsson. Both required tubular workpiece material or preparation 16 2. State of the art

Figure 2.4: Different cutting tribometers. a) In-process tribometer by Olsson et al. [107] (figure from [186]). b) Open tribometer by Hedenqvist et al. [61] (figure from [186]). c) Open tribometer by Puls et al. [120]. d) In-process tribometer by Zemzemi et al. [186]. e) Open tribometer by Rech et al. [122]. with a face grooving operation because of stability problems in the otherwise required big depth of cut.

2.3 Titanium Cutting Process, Tools, and Coatings

When developing MWFs, especially very specialized ones, advances in the tool material, coating and geometry, as well as developments in the machining strategies always have to be considered. In this way, a good performance at state-of-the-art conditions is ensured. In 1990, Machado et. al [89] presented a review on titanium cutting. Although the MWFs, tool substrates and coatings are continuously optimized, the basic issues stayed constant. A book by Paulo Davim [112] gave a more recent overview on the state of the art in titanium machining. Titanium is considered to be diﬃcult to machine. Its high strength, ductility, high reac- tivity, and poor heat conduction lead to high tool wear and limit the achievable cutting speeds. According to a review of Abele et al. [2], cutting speeds around 50 m/min are common for cutting titanium with cemented carbide tools. Titanium chips are serrated in 2.3 Titanium Cutting Process, Tools, and Coatings 17

all industrially used cutting conditions. According to the model of Komanduri et al. [76], the serrations are induced by a combination of adiabatic shear bands and cracks forming in the shearing zone. The low Young’s modulus of titanium leads to a significant spring-back effect behind the cutting edge, as shown by Schaal et al. [126]. This causes a wide contact zone at the flank face, higher feed forces, higher temperatures, and increased tool wear compared to a situation without spring-back. Although the temperatures at the cutting edge initially increase when machining steel or aluminum at higher speeds, once the cutting speed reaches five to ten times the conventional cutting speed, the temperature of the cutting edge drops again and allows economic machining in the high-speed cutting (HSC) regime, where most of the heat is transported away with the chips. Li et al. [84] milled titanium with cutting speeds of up to 700 m/min and measured the temperature at the cutting edge but could not observe any temperature drop. However, lower cutting forces are measured at high cutting speeds, which is attributed to thermal softening. Eckstein et al. [48] observed a drop in cutting force between 200 m/min and 300 m/min in milling of titanium, allowing for slightly higher tool lives when cutting with speeds in this region compared to neighboring regions. However, they used a small feed per tooth of 0.045 mm which only renders this approach feasible for finishing. Sutter et al. [144] used a ballistic setup to obtain titanium cutting speeds of up to 4800 m/min and observed a dropping temperature up to speeds of 1800 m/min. At higher cutting speeds, thermal softening is compensated by dynamic strain hardening and strain rate hardening. One possibility to deal with the rising temperatures is to use a cutting tool with a round, rotating insert. This way, each part of the cutting tool only spends a short time in the hottest part of the cutting zone and is cooled afterward. In this manner, Lei et al. [83] increased the tool life in Ti6Al4V turning by a factor of 64 compared to a stationary tool. If manual indexing with up to 37 positions is considered, an increase in tool life by a factor of at least 1.7 remains, allowing for higher cutting speeds with the same tool life. Ezugwu et al. [52] reported worse tool life when turning Ti6Al4V with CBN than with uncoated cemented carbide due to notch wear. Under very light cutting conditions in finishing, according to Kuljanic et al. [77] polycrystalline diamond can be used to cut titanium up to cutting speeds of 110 m/min with tool lives of over 6 h. A passivation effect through the formation of a dense TiC-layer was suggested as an explanation. Another promising material for high-speed cutting of titanium is binder-less CBN, which has better mechanical properties at elevated temperatures than conventional CBN. Wang et al. [161, 162] showed longer tool life than with conventional CBN at speeds around 350 m/min. Currently, titanium is always cut wet when seeking highest productivity. Sørby et al. [138] achieved 300% longer tool life when using high pressure coolant at up to 30 MPa in 18 2. State of the art

a grooving operation. A further increase of 100% was achieved with additional flank face cooling. Attempts to replace the flood cooled process with MQL or even a dry process are motivated by environmental concerns and cost reduction but do not reach the same tool life at conventional cutting speeds [44, 114]. According to a review of Ezugwu et al. [51] in 2003 cryogenic cooling has the potential to reduce tool wear below the level of flood cooling, which was later confirmed by Deiab et al. [44]. Tool coatings do not improve the tool life when machining titanium cutting as significantly as when machining steel. This may be attributed to the high affinity of titanium to the common titanium-based coatings. Therefore, some large tool manufacturers still recom- mend the use of uncoated tools. However, massive improvements in tool life for drilling titanium with AlTiN-coated carbide tools compared to uncoated tools are reported by Sharif et al. [128]. Klocke et al. [74] inverted the usual process development and selected the MWF first. They choose a biodegradable polyol ester base fluid and optimized the tool coating and process afterwards, achieving roughly the same productivity as in the original case with emulsion. The cutting edge micro-geometry has an influence on chip formation and cutting forces, tool wear and workpiece surface integrity [45]. Depending on the weight of those factors, the desired process parameters and the tool material properties, an optimal tool micro- geometry exist. The sharper the tool is, the higher the risk of edge breaking is. In addition, the surface roughness may be impaired by remaining grinding marks and the compressive residual stress drops in general. Blunter tools lead to an increase in cutting force as shown by Wyen et al. [181], a shift of the maximum temperature from the rake face closer to the cutting edge as shown by Bassett et al. [18], and more burr formation as shown by Denkena et al. [47]. The micro-geometry of the cutting edge can be described using several different parameters. The micro-geometry of rounded cutting edges can be described with the cutting edge radius. A real cutting edge profile will never exactly consist of two straight flanks and a perfect circle, tangent to both flanks. Thus, robust algorithms such as the one developed by Wyen et al. [182] are necessary. Wyen’s algorithm determines the part of the cutting edge that belongs to the micro-geometry first and then fits two circles to the profile of the micro-geometry. The circle with radius R is constrained to be tangential to the two flanks,

whereas the circle with radius Rn has no constraints. An example is shown in Figure 2.5. When describing asymmetrical cutting edges, the form-factor method, developed by Denkena et al. [46] can be used, which is shown in Figure 2.5 as well. First, a profile of an ideally sharp cutting edge is fitted to the measured profile. Then, the points where the measured profile meets the ideal profile are evaluated on both the rake face and the flank 2.4 Cutting Forces 19

face. The distance from those points to the ideal cutting edge are designated Sγ for the

rake face and Sα for the ﬂank face. The ratio Sγ/Sα is a measure for the cutting edge asymmetry. Karpat et al. [73] report an industrially achievable hone radius repeatability of 5 µm for CBN tools but expects worse repeatability for the softer carbide tools. Bassett et al. [17] used a tolerance band from 25 µm to 35 µm for both Sα and Sγ and determined the process capability Cp to be 0.41 and 0.46 respectively for the brushing hone method. Evaluating their tool life map in the small variation range of cutting edge radii from 25 µm to 35 µm, tool life is expected to vary by more than 10% because of variations in the micro-geometry.

2.4 Cutting Forces

As one of the first researchers investigating cutting processes with defined cutting edges, Merchant [93, 94] developed an analytical cutting force model for orthogonal cutting. He assumed a constant shearing stress in the infinitely thin, straight shear plane and applied the coulomb friction model. Furthermore, the tool was assumed infinitely sharp with no friction on the flank face. The cutting force and thus the cutting power reaches a minimum value at a certain angle of the shear plane. Merchant assumed this value to be equal to

S S r α γ Δ

R lank face n �

rake face

Figure 2.5: Overview of the parameters to describe the cutting edge micro-geometry. The measured proﬁle is shown in black with the part identiﬁed to belong to the micro-geometry highlighted in blue. 20 2. State of the art

the actual shear plane angle observed in cutting. Despite its very limiting assumptions, the model is still used today, for example to estimate the coefficient of friction in new tool types [57]. Merchant’s model leads to expected cutting forces, which are proportional to the undeformed chip thickness. In reality, the forces show a less than proportional increase with chip thickness. Kienzle [Kienzle 1952] used an empirical approach and suggested a power law with an exponent between 0 and 1 for the uncut chip thickness. Kienzle’s model is frequently used today, as it is available for cutting force predictions in many CAM tools and myriads of measuring data are available and tabulated. An even simpler approach is presented by Altinta¸set al. [4], commonly referred to as the Altinta¸smodel. This model separates the cutting force into a constant portion caused by the cutting edge, which is not affected by the depth of cut and a variable portion, which is proportional to the local uncut chip thickness. Wyen et al. [181] investigated the cutting and feed forces and found the relations shown in Figure 2.6. They concluded that the Altinta¸smodel is valid for uncut chip thicknesses bigger than the cutting edge radius. The influence of the cutting edge radius on the cutting force is small compared to the influence in the feed force. When a straight, oblique cutting edge with an inclination angle is used, the chip will not flow orthogonally to the cutting edge. Stabler [141] suggests that the chip flow deviates

400 400

350 350 r n== 50 µm g r n== 40 µm 300 300 r n== 30 µm r n== 20 µm [N] c 250 [N] 250

f r n== 10 µm

200 200

150 r n== 50 µm 150 feed force force feed F

cutting force cutting force F r n== 40 µm 100 r n== 30 µm 100 F Pl,c r n== 20 µm 50 50 r n== 10 µm FPl,f

0 0 0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25 uncut chip thickness h [mm] uncut chip thickness h [mm]

Figure 2.6: Experimentally determined cutting forces Fc and feed forces Ff for turning Ti6Al4V with diﬀerent cutting edge radii rn at diﬀerent uncut chip thicknesses h, the data is standardised to a cutting width of b = 1 mm (vc = 70 m/min), by Wyen et al. [181]. 2.4 Cutting Forces 21

from the direction orthogonal to the cutting edge by the same chip ejection angle ηc as the inclination angle of the cutting edge λS, being referred to as Stabler’s law. Later studies confirm this tendency but obtain values up to 20% smaller than the inclination angle due to friction effects on the rake face [110]. The deviations of the chip flow direction from Stabler’s law at higher inclination angles are explained by Moufki et al. [100] with a temperature dependent coefficient of friction and a viscoplastic material model. The rake angle is defined as the angle between a line orthogonal to the cutting speed vector and the intersection line of the measurement plane containing both the cutting speed vector and the orthogonal line with the rake face. Considering an oblique cutting tool, the value depends on the orientation of the orthogonal line. Brown et al. [28] define the three most meaningful rake angles:

• the normal rake angle, with the measurement plane being equal to the wedge measurement plane, i.e. orthogonal to the cutting edge

• the velocity rake angle, with the measurement plane being equal to the working plane, i.e. parallel to the feed direction

• the eﬀective rake angle, with the measurement plane parallel to the chip velocity

Brown et al. [28] performed cutting tests with varying obliqueness and found that the normal rake angle is the most significant rake angle in terms of chip formation and cutting force. Shi et al. [133, 134] studied the influence of confined chip movement. They used sym- metrical turning tools with both positive and negative lead angles as shown in Figure 2.7. In the case of θ < 0, if two independent chips would form on every side of the cutting tool, they would cross each other in the middle. However, as a single chip is formed, compressive stresses in the chip force both sides of the chip to move parallel to the feed direction. This restriction in movement causes higher cutting forces. The same effect can be observed in the case of θ > 0. Two independent chips would flow apart from each other. However, the chip is held together by tensile stresses, forcing the chip to move straight in the feed direction, again causing an increase in cutting force. For θ > 0 the effect is limited to values below 20°, presumably because of the chip breaking apart in the middle, leaving two non-interfering chips at higher angles θ. The feed has been increased and the workpiece width has been decreased in the tests with higher absolute values of θ in order to keep the uncut chip thickness ac and width aw constant. A simple empirical model of the increase in cutting force is presented:

2 Fc = Fc0 · (1 + µ1θ ) (2.1) 22 2. State of the art

rotation 1200

1000 aw workpiece ac aw 800 ac θ [N] force cutting θ tool

600 feed

<0 >0 -40 -20 0 20 40 cutting edge lead angle [°] θ θ Figure 2.7: Left: Experimental setup to measure the force increase through restθ ricted chip motion. Right: Experimental data. Modiﬁed after Shi [134].

Equation (2.1) features the cutting force Fc with the effect of chip restriction, the cutting force Fc0 without chip restriction, a parameter µ1, and the deviation of the actual chip flow direction from the unrestricted direction θ. An increased force is also observed if the chip is not able to locally move with the optimal speed. But, the influence of the restricted chip speed is much more difficult to test. Shi [134] used a tool stack with different rake angles engaging all tools at the same time to form a single chip. The higher rake angles lead to less chip compression and therefore to a higher chip speed. By stacking a tool with high rake angle between two tools with low rake angles, the chip of the middle tool was forced by shear stresses in the chip to move at a lower than optimal speed, whereas the chip parts of the outer segments were forced to move at a higher than optimal speed. By using different combinations of rake angles, the effect of the speed restriction can be isolated. It turns out that a similar relation as in Equation (2.1) is valid:

r − r 2 F = F · 1 + µ 0 (2.2) c c0 2 r 0 !

Equation (2.2) features the cutting force Fc with the eﬀect of chip restriction, the cutting force Fc0 without chip restriction, a parameter µ2, the actual cutting ratio r, and the unrestricted cutting ratio r0. 2.5 Tool Wear 23

2.5 Tool Wear

The accurate measurement of tool wear is difficult, as small changes in the tool micro- geometry should be measured accurately, non-destructively, and quickly. A well-established measure is the tool flank wear land width (VB), shown in Figure 2.9. It is easily measurable with a microscope and correlates well with the expected tool failure and workpiece quality. This is also reflected in the respective standards. Both ISO 3685 for tool life testing with single-point turning tools and ISO 8688 for tool life testing in milling recom- mend a given critical VB as basis of a tool life criterion. The flank wear land can have a non-uniform shape. Therefore, the standards include recommendations on how to evaluate the width. Tool life tests can be accelerated by either accelerating the wear itself or by changing the tool life criterion. Merchant et al. [96] opposed to accelerating the wear rate, as this could completely change the wear behavior. Simply changing the critical VB to smaller values leads to shorter tests, but in turn leads to higher measurement uncertainties. Therefore, they introduced a new wear measure: the radioactive tracer method. The tool was irradiated with particles from a nuclear reactor or a particle accelerator, forming a thin radioactive layer around the cutting edge. By measuring the difference in radioactivity, they could quickly and accurately measure the worn volume even in early wear stages. The same authors [95] later extended the measurements and found that over 90% of the worn tool material was adhering to the chips. The setup was later refined by Opitz et al. [108]: Two opposing tools with the same geometry, one radioactive and one not, were used to longitudinally turn a workpiece. The radioactive tool itself provided information about the total wear volume. The activity of the chips from this tool provided information about the worn volume that was transported over the rake face. The other tool cleaned the workpiece’s surface from radioactive traces of the other tool’s flank face and therefore provided information about the worn volume that was transported over the flank face. Cook et al. [37] criticized the use of the radioactive tracer method. They showed that the results from the radioactive method correlate well with the conventional method of measuring VB, but only for a given process. As soon as cutting parameters, workpiece material or tool material were changed, the altered wear behavior lead to a different relation between the radioactivity and VB. Today, the radioactive method is rarely used anymore. The method is potentially hazardous and similar information is obtainable today with 3D-microscopy. Other approaches to quantify small amounts of wear were suggested by Uehara et al. [150]. C45 Chips with adhering tungsten carbide wear are pickled in sulfuric acid to release the tungsten carbide from the chips. The particles are then separated from the chips and the amount of tungsten is measured colorimetrically. Another approach is 24 2. State of the art

to place a film resistor on the flank face and measure the resistance between the far end of the resistor and the workpiece. In this manner, a continuous in-process measurement of wear is possible. The measurements are closely related to the mean flank wear land width in the resistor area. Both methods are laborious and are therefore rarely used today. Several approaches have been made to accelerate wear, usually by increasing the cutting speed. The principle is based on Taylor’s equation of tool life:

n vcT = C (2.3)

In this empirical relation, the cutting speed vc and the tool life T depend on each other via an exponent n and a constant C. By performing experiments with higher than economic cutting speeds, tool life is kept short, leading to less expensive experiments. In theory, the results can later be extrapolated to economic cutting speeds. Lorenz [86] investigated possible rapid machinability tests. He doubled the cutting speed and found the same constants of Taylor’s law to be valid for three of five of the investigated tool life criteria. Malakooti et al. later [91] successfully applied this method and showed the same ranking of machinability under different cutting conditions. However, according to Jhita et al. [69], Taylor’s law is only valid as long as the dominant wear mechanism does not change. In Figure 2.8, Klocke [75] shows the contribution of different wear mechanisms at different temperatures for the example of steel cutting. Just below 700 ◦C, the wear rate even decreases with increasing temperature or cutting speed. This clearly shows a deviation from Taylor’s law. Different activation energies at different temperatures, as described by Cook [36]; formation of a built-up edge (BUE) at lower cutting speeds; or changes in the dominant wear mechanism from diffusion to adhesion; further limit Taylor’s law to piecewise application. Tool wear can be traced down to several basic mechanisms. However, sometimes the effects are not easily distinguishable and therefore the type and number of categories may vary [9, 30, 75]. Adhesion wear, abrasion wear, diffusion wear, chemical wear, and mechanical overload are chosen in the present work. Adhesion wear is caused by welding between the tool and workpiece. In severe cases, this leads to the formation of a BUE, which periodically breaks off, removing parts of the tool as well. On a local scale, adhesive wear can be modeled using several semi-empirical models. In the model of Archard et al. [8], the total adhesive worn depth Wadh is assumed to be proportional to the normal load p, the sliding distance L, and the constant K, while being inversely proportional to the hardness H of the softer partner:

dW Kp adh = (2.4) dL H 2.5 Tool Wear 25

700°C

Diffusion

Abrasion Total Wear Rate Wear Total Adhesion Oxidation

Cutting Temperature (vc , fZ , ...)

Figure 2.8: Contribution of different wear mechanisms at different temperatures for steel cutting. Modified after [75].

In the deduction, Archard only assumes wear of the softer friction partner, therefore wear of a cutting tool cannot be described by this equation, as the cutting tool always has to be harder than the workpiece. Abrasion is caused by the rubbing of hard particles and mainly leads to flank wear. Hard particles may stem from the casting process or from oxide and carbide precipitates from within the workpiece material. The strain-hardened BUE can also act abrasively when it breaks off and passes by the flank or rake face. Diffusion wear is caused by mobile atoms, which move from the tool into the workpiece and chip, thereby leading to the dissolution of the tool. Diffusion in the reverse direction can indirectly lead to increased wear as well. By changing the composition of the tool material, its properties change, possibly promoting the other wear mechanisms. In its simplest form, Fick’s law can be used to predict the dissolution of the tool material in the workpiece material, leading to a constant wear rate after the establishment of a steady- state gradient, as it has been done by Takeyama and Murata [146]:

dW − EA = De Rθ (2.5) dt

Equation (2.5) features the worn depth W , a constant related to diﬀusion D, the activation energy EA, the Boltzmann-constant R, and the absolute temperature θ. The Archard model for adhesive wear is later extended by Usui et al. [155] for the application on cutting tool systems. They use an Arrhenius-term to account for the combined 26 2. State of the art

Width of lank wear land VB

� N α VB KB: Crater width SV

KM: Distance betw. crater VB max.

centers B

KT: Crater depth VB SV : Displacement of cutting edge toward lank face α Wear notches on SV : Displacement of cutting major cutting edge toward rake� face γ edge

sectional view A-A KT A

KB crater

SV A

α γ

Figure 2.9: Nomenclature and measurements for tool wear phenomena [75].

eﬀect of softening of the tool and hardening of the workpiece by diﬀusion as well as thermal softening of the tool:

dW − C2 = C e( θ ) (2.6) pdL 1

Equation (2.6) features the worn depth W , the normal stress p, the sliding length L, the

absolute temperature θ and the constants C1 and C2. Usui et al. [155] estimated the stress, velocity and temperature distribution on a cutting tool and showed a good prediction of the tool wear shape. Their model is valid at diﬀerent tool wear regimes and can be used for abrasive wear as well, although the constants must be adjusted accordingly. The constants can only be derived experimentally, with the inﬂuence of tool material, workpiece material, 2.5 Tool Wear 27

and lubrication being unclear. Usui’s model is very similar to Equation 2.5 with the only difference being, that the wear rate is proportional to the contact pressure. Chemical wear is caused by a chemical reaction between tool and workpiece material or the MWF. With conventional workpiece materials, chemical wear is not a big concern. Ox- idation of the tool material can be prevented with coatings or doping of the tool material, as Klocke [75] shows. Mechanical overload can be divided into direct mechanical stress caused by the cutting forces and thermo-mechanical loads due to temperature gradients in the tool. Furthermore, it can be distinguished between brittle fracture, plastic deformation and fatigue cracking. Fatigue is only an issue in interrupted cuts, when mechanical loads and thermo-mechanical loads are periodically changing. Typically, the cemented carbide cutting tools used for cutting titanium show very pronounced wear craters with their maximum depth near or even at their cutting edges. Abrasion is not a major issue because of the lack of hard precipitates in titanium. How- ever, adhesion plays a major role, because of the high affinity of unoxidized titanium to almost any surface. Despite the strength of cemented carbide initially being much higher than that of titanium; mechanically, thermally, or chemically weakened portions of the cutting tool can be removed by adhesion. Various researchers presented models for the crater formation. Hartung [60] and Hartung et al. [59] showed that a simple dissolution model as in steel cutting does not provide accurate predictions of tool wear for different coatings. Instead, a reaction layer consisting mainly of TiC forms between the tool and chip. It acts as a diffusion barrier for carbon and therefore the reaction layer thickness and stability govern the wear rate. Dearnley et al. [42] found signs of attrition wear caused by adhesion in addition to diffusion wear. Jawaid et al. [67] additionally found cracks and plastic deformation of the tool. Jayal et al. [68] investigated the wear after a short cutting time at high speed. They found ridges in the crater in direction of the chip flow and higher wear than they expect. This lead them to the conclusion that self-abrasion, i.e. tungsten carbide grains scratching over the tool surface, was the major wear mechanism. Ghani et al.[54] used a different cutting parameter set and found microcracks to be the main source of wear. They were believed to stem from mechanical overload due to adhesion of workpiece material to the chip. Bordin et al. [24] described adhesion as the main wear mechanism and showed a reduction in adhesive properties when reducing the temperature with liquid nitrogen. Odelros et al. [106] explored the diffusion wear in more detail. They predicted the formation of a carbon-depleted zone on the surface of the tools and confirmed it with SEM-images of a tool cross-section and XRD-scans, where a small portion of metallic tungsten was visible. Furthermore, they predicted the formation of a 28 2. State of the art

◦ T i2Co-phase at the interface with a melting point below 1000 C, explaining the fast wear at elevated temperatures. However, no experimental evidence of this phase was found. Cemented carbide titanium cutting tools often exhibit notch wear. Although it is not as pronounced as with ceramic inserts, it may still be the critical factor determining the end of tool life. Chandrasekaran et al. [34] investigated notch wear in stainless steel cutting and concluded that adhesion and abrasion of the saw-like chip edge leads to brittle material removal. Adhesion is promoted by lateral deformation of the chip in the notch zone, exposing nascent workpiece material to the tool. Klocke [75] added abrupt transitions in mechanical and thermal stresses at the borders of the contact zone and oxidation eﬀects as possible reasons. Because the cutting temperatures are generally high in the case of titanium cutting, thermal fatigue wear can be an issue when machining with interrupted cuts. 29

Chapter 3

Task and Aim

The aim of the present work stems from a need in the MWF industry to develop new MWFs more efficiently. In this long-term goal, the iterative approach described in Section 2.1 should be replaced by a more target oriented approach. This requires a comprehensive model to predict any of the MWF properties from its formulation, which is not possible in the medium term due to the high number of relevant, complex interdependences. There- fore, in the medium term, the iterative process will persist, but the number of iterations can be reduced and be made less expensive. A very wide-ranging approach is chosen for this purpose. Every step in the development process of a new MWF is analyzed and possibly improved. The current methods are complemented with new methods where necessary. The approach can be visualized in the 2D graph shown in Figure 3.1 which is based on Figure 2.2 by Belluco [21]. It shows the cost and relevance of the tests in the current situation and the aim to reduce costs and improve relevance. The dashed line through the currently applied methods in white boxes represents the state of the art. It divides the plane in two regions. Test methods above and to the left of the line are feasible but have a high cost and a low relevance, which renders them uneconomical. Methods below and to the right of the curve would be economically favorable. However, they are not available due to the seemingly arbitrary differences between test results and the actual quality of an MWF which limit the relevance. The arbitrary differences are caused by systematic errors in the test result which result from a lack of understanding of the underlying influence factors. With research, more influence factors can be understood and the dashed line can be shifted down and right, resulting in the target dot-dashed line with more economical and relevant tests. Therefore, the area between the state of the art line and the target line can be interpreted as the research gap. Currently, cutting tests results have a high variation and are very expensive. Turning tests shall be introduced as a reliable and less expensive cutting test to complement the milling tests. Both types of tests shall be made less expensive. The required number of repetitions, 30 3. Task and Aim

Full Scale Production Test feasible, but Milling uneconomical Test

p Turning a g Tapping Test

Torque ch cost Test ar se Brugger In-process re Test Tribometer Test Pin-on-disc unfeasible, but Test Wear desirable Prediction

correlation with real performance

Figure 3.1: The research gap visualized according to Belluco [21]. New blocks are shown in green. Improvements and substitutions of existing blocks are indicated with arrows. The research gap is indicated in light green. the amount of material and tools needed can be reduced by using wear progression models, more data sources and an improved data evaluation. Furthermore, the operator’s effort to set up the test and evaluate the data shall be reduced. Full-scale production tests are not in the scope of this work, as they are rarely used when developing new universal MWFs but instead are performed by the customer when optimizing his production process. Conventional tribological analysis is mostly focused on steel friction partners. However, in titanium cutting, the tool is in contact with unoxidized titanium. There is no established method to assess friction on an unoxidized titanium surface in an industrial setup. Pin-on- disc or similar methods to determine the coefficient of friction under different conditions shall therefore be replaced by a method that works with unoxidized surfaces. The in- process tribometer setup allows to reduce the influence of the oxide layer significantly. The tapping torque test is not available for titanium, therefore, the in-process friction tests shall substitute it as well. The friction measurements should be complemented with a friction model to reduce the number of necessary tests. Likewise, the Brugger test shall be replaced with wear assessment after in-process measurements of friction. This test is 31

more expensive, but a much better correlation with the real performance of the MWF is expected as well due to the correct material combination. Currently known mechanisms of titanium cutting tool wear were either described for dry cutting or without taking the chemical effects of the MWF into account. Therefore, a wear prediction model considering the influence of the MWF shall be introduced. The relevant mechanisms should be identified and existing wear models should be modified, revealing the relevant parameters. Once the wear prediction model is developed, it will be cheap to use since it will not involve material or machine costs. On the other hand, it will have a low relevance due to the small number of effects that can be captured. In each of the following sections, one of the methods shown in Figure 3.1 is treated. The methods are roughly sorted by their cost and relevance in descending order, with the cutting tests in turning and milling followed by the tribological in-process tribometer test. Lastly, wear know-how and prediction is treated as a very important and cheap tool to design MWFs, but with a limited relevance. 32 3. Task and Aim 33

Chapter 4

Turning Tests

4.1 Turning Test Bench

4.1.1 Machine

Figure 4.1: Lathe (left) with external MWF tank and high-pressure delivery system (right).

A NLX 4000/750 CNC lathe by DMG Mori as shown in the left part of Figure 4.1 is used in all turning tests. According to the manufacturer, the machine uses the biggest hydrodynamic bearings of its class, providing excellent stiﬀness and damping. The machine does not feature the optional y-axis, which would have reduced the stiﬀness. The maximum turning diameter is 600 mm. The original MWF tank and delivery system of the machine are not used. Instead, one of two specialized systems is used: an external high-pressure unit located next to the machine as seen in the right part of Figure 4.1 or an internal low-pressure unit. 34 4. Turning Tests

The pumps of the external MWF unit are able to achieve 200 bar at 30 l/min. Two frequency-controlled motors are used to drive piston pumps. At the highest pressure, the MWF has to be supplied directly to the tool. For lower pressures up to 70 bar, the MWF can be conveniently supplied through the turret. The unit incorporates a ﬁlter to remove suspended particles. The tank holds 800 liters and is actively cooled. This volume is needed when working with high pressure in order to ensure that no entrapped air is supplied to the tool and to provide enough thermal latency for the cooling system to react. Changing the MWF takes between one and two workdays depending on the types of MWF used before and afterward. The small internal MWF cycle allows to capture the used MWF before it reaches the machine tool fairing and feeds it back to a small tank holding 30 liters immediately. This unit uses an external gear pump with a frequency-controlled motor. The pressure is limited to 16 bar and no cooling is available. However, the MWF can be changed in 30 min, making it the ideal choice for screening tasks.

4.1.2 Sensors

The machine operation and the cutting process are monitored using a variety of different sensors. Whereas the cutting force and tool temperature are directly evaluated for the assessment of an MWF, the other data is stored and used to detect deterioration of the MWF unit, leaks, and blockage over time. The temperature of the MWF is measured in the external tank and used as a controller input for the cooling jacket. The pressure is measured with two piezoresistive sensors: one at the exit of the MWF unit and one directly before the MWF enters the tool. A gear flow meter is used to measure the total MWF flow rate at the exit of the MWF unit. Calculating the pressure drop inside the machine tool and knowing the flow rate, the viscosity of the MWF can be monitored. The absolute pressure before entering the tool is used as input for controlling the pump speed. A piezo electric dynamometer of type 9129AA with a measurement range of 10 kN is mounted on a modified tool holder, which is in turn mounted onto the turret. The tool is mounted on the dynamometer with specially designed brackets. The signal of the dynamometer is amplified with a charge amplifier to obtain three analogue voltage signals, one for each direction. The same sensors are used with the internal MWF unit. However, only one pressure sensor near the tool is used and the flow sensor is omitted. 4.1 Turning Test Bench 35

Of the several methods to measure the temperature near the cutting edge described in literature, most are not suitable for the use with MWFs or for longer measurements. Fiber-pyrometric measurement in a hole drilled to the underside of the rake face, as shown in Figure 4.2, does fulﬁll all requirements. It produces reliable data at temperatures above 300 ◦C. As the temperature distribution has a steep gradient near the cutting edge with an unknown shape due to the distributed heat sources and sinks, the temperature is measured as close as possible to the cutting edge in order to avoid the need for extrapolation. Blind holes are drilled by EDM drilling, leaving a characteristic hole shape due to electrode wear. The remaining wall thickness is measured through tactile methods. Holes with a deviation of more than 20 µm from the desired 200 µm are discarded.

Ø 0.5 mm Ø 0.6 mm

0.25 mm

0.2 mm 0.2

lank face lank � rake face

Figure 4.2: Sketch of the situation in pyrometric temperature measurements of the cutting edge. A cross-section through the middle of the ﬁber (yellow) is shown.

Figure 4.2 shows how the fiber is placed as far down the drilled hole as possible. The temperature is measured with the two-color pyrometer FIRE-3 by En2Aix. Although the temperature calculated from the ratio of emission at two different wavelengths produces a temperature measurement, which is somewhat averaged over the acceptance cone of the fiber, the value is strongly shifted towards the maximum temperature, since the radiated power is proportional to the temperature to the power of four. The fiber has to be secured tightly in the hole of the insert to prevent it from slipping out when vibrations occur or the tool moves. The fiber has a minimum allowable bending radius of 50 mm. This poses a challenge in internal turning. By guiding the fiber in a metal tube as shown in Figure 4.3, it is protected from the chips and the bending radius is prevented from dropping below its minimal value. In this way, internal face turning down

to a diameter of 46 mm is still possible with a maximum depth of cut ap of 1.5 mm. 36 4. Turning Tests

Figure 4.3: Modified turning tools to receive the fiber of a pyrometer. The external shaft tool is shown to the left. The internal turning head is shown to the right. The fiber is not shown to make the holes in the inserts visible. They are indicated with red arrows.

A flexible mantle thermocouple with the same outer diameter of 0.5 mm as the fiber is available as well. It is easier to use, less expensive, and has a better-suited measuring range. However, it has a higher latency due to the heat transfer between the drilled hole wall and the thermocouple. Furthermore, the measured temperatures are always smaller than those measured pyrometrically because of the distance of the thermocouple from the rake face and a stronger, linear averaging effect from all the surrounding areas through heat conduction.

4.1.3 Data Acquisition

The data acquisition (DAQ) is designed to be as automated as possible in order to not present an additional workload for the machine operator. The system is based on a data acquisition device of type USB-6361 by National Instruments. It features 8 differential analog input channels, 2 analog output channels and 24 Digital I/O pins. The acquisition software as well as the GUI are programmed in LabVIEW. Figure 4.4 shows the screen, which is divided into a small configuration area, indicators, and two large plot areas. The upper area shows the unfiltered signal in close to real time from the beginning of the current acquisition period. The lower area shows a compressed view of all data since the last manual reset. Both diagrams are important to understand the process and to take action if necessary. The upper, fast diagram is used to detect subtle changes in the 4.1 Turning Test Bench 37

tool geometry, which may lead to a change in signal variance. The lower, slower diagram is used to visualize long-term wear behavior of the tool. A tool life test can take more than half an hour, with up to 4 minutes of continuous cut. In simpler programs, where only the last few seconds of data are shown, the visualization is possible in time with the data packages arriving. In this case, however, the visualization takes more time than the time available between two data packages. Therefore, the program is split in two cycles operating at independent cycle times. The faster cycle pulls data from the onboard memory of the DAQ device in packages. The data is scaled from voltage back to the physically meaningful measurement unit depending on the configuration data given by the user. The data is then put in a queue in the RAM of the PC. The slower cycle pulls data from this queue, filters it for the lower diagram and the numerical indicators, updates the diagrams and saves the data to the hard disc. Further features of the program include configuration files, which store all setup parameters, and a special setup-mode without data saving, which is used to calibrate sensors and adjust the pump speeds.

Figure 4.4: GUI of the acquisition software.

The acquisition start is triggered by an impulse that can be sent by an M-code from the PLC of the lathe. The DAQ device then sends a digital signal to the charge ampliﬁer to bring it to the measuring state. Data is then acquired and processed as described above. With the second pulse, the acquisition stops, and the charge ampliﬁer is brought back to the reset state. The data saving continues until the queue is empty and the system is ready for another acquisition. 38 4. Turning Tests

4.2 Reference Turning Process

A reference parameter set is needed to compare different MWFs. The reference parameter set must be chosen wisely, as the performance of the MWF at the reference parameter set is considered the real performance of the MWF. In other words, the quality of a quick test is later determined by comparing its results to those of the reference parameter set. If a universal MWF is to be developed, the real performance can be defined as an average or a minimum over multiple processes, process parameters, and materials. In this case, a single roughing turning process in titanium is selected. The reference parameter set has to satisfy a number of conflicting criteria. It should be industrially relevant, meaning that the obtained ranking of MWF performance holds true for many of the later applications. The measurement value should have a small variation in order to reduce the number of required repetitions to obtain a reliable mean value. It should further lead to a good separation of the test candidates in order to reduce the required repetitions to obtain a reliable ranking. The process should be stable in the long term, meaning that it produces the same results if repeated after a year. Furthermore, the process should be economical in terms of time used for the machine tool and operator as well as material and tooling costs. The titanium alloy Ti6Al4V (Titanium Grade 5) is chosen since it is the most prevalent in titanium cutting industry, is challenging to cut, and is cheaper compared to more advanced titanium alloys. It is purchased as rolled bar material peeled to a diameter of 305 mm. Every test is made with material of the same lot in order to keep the influence of the workpiece material as low as possible. The bar is sawn into pieces of 250 mm to make it suitable for single sided clamping on the lathe. The insert of type CNMG 12 04 12-SM 1125 by Sandvik Coromant is selected. It is made of a tough cemented carbide substrate, which leads to an even and uniform wear. The PVD coating is based on (T i, Al)N and (Al, Cr)2O3 with a golden T iN top-layer. According to the manufacturer, the coating does not promote tool life in this case. However, it helps to unambiguously detect the flank wear land. The big corner radius of 1.2 mm is chosen to increase the corner stability. The negative insert features a constant wedge angle of 75° along the perimeter. In its installed position, this leads to a positive rake angle with values between 5° and 9°, depending on the direction and tool used. As usual for a roughing insert, it is not ground after sintering, resulting in bigger variations of the micro-geometry and insert size. A face turning operation is preferred over a longitudinal turning operation in tool life testing of large workpieces, despite the longitudinal process being more prevalent in industrial applications. In longitudinal turning, the tool life in the outer diameters would be in the 4.2 Reference Turning Process 39

order of one to three passes, meaning that one insert is only tested at a specific diameter. The material at the core of bar stock is usually softer than at the outside either due to less work hardening or a slower cooling rate in the manufacturing process. In addition, the differences in workpiece radius lead to differences in the contact zone at the secondary flank face. At smaller diameters, the tool reaches the shoulder more often in the same time and the cut has to be interrupted more frequently. Combining these effects, it is impossible to compare results from a new piece of raw material with a large diameter (respectively a small diameter in internal turning) to those obtained near the end of life of a workpiece at a smaller diameter (respectively a larger diameter in internal turning). In face turning, each insert undergoes this change in radius several times, blurring the radius influence over the whole tool life. Of course, any gradient of material properties in the longitudinal direction has a much greater influence in this case. Through the manufacturing process, this gradient is, however, much less distinct. At regular intervals, the process is stopped and the insert is observed with a microscope. To ensure that every insert experiences the same load case at the same time of the test, the face turning tests are always started at the same diameter. Any remaining material from a previous test, where the tool reached its end of life in the middle of a pass is removed with a secondary insert. The cutting process parameters have a strong influence on the expected tool life and the wear behavior. As a starting point, the recommended parameters of the insert manu-

facturer, shown in Table 4.1, are used. ap has a minor inﬂuence on tool life, as the cutting action can be described as a two-dimensional process which does not vary along the cut-

ting edge. In industrial applications, ap is chosen as high as possible while still preventing vibrations to reach the highest possible material removal rate (MRR). In tool life testing, ap is chosen as small as reasonably possible to limit material cost. In this case, it is reduced to 1.5 mm to still have the straight part of the main cutting edge engaged. The cutting speed is raised to obtain tool lives in the order of 20 min. Longer tool lives are uneconomical, shorter tool lives limit the industrial relevance and lead to higher uncertainties. The cutting is performed using a high performance universal emulsion at a concentration of 8% at a pressure of 70 bar. The workpiece is turned from a diameter of 302 mm inwards to a diameter of 40 mm in four steps of equal duration. After each step, the wear is measured. The ISO 3685 standard for turning tool life tests is only partially applicable to deﬁne the tool life criterion, as the tool corner region occupies most of the depth of cut. A smooth ﬂank wear land shape with no break-outs is usually achieved. The wear shape for the parameter sets A and D are shown in Figure 4.5. The wear land has no constant region,

therefore the easily measurable and in this case robust VBmax is used as criterion. The

critical value of VBmax = 0.6 mm proposed by the standard is too large, as the risk of

tool breakouts increases signiﬁcantly at values of VBmax between 0.2 mm and 0.6 mm.

Therefore, VBmax = 0.2 mm is chosen as the tool life criterion. 40 4. Turning Tests

Parameter Set A Parameter Set D

Figure 4.5: Wear land shapes of cutting tools at the end of life in external face turning.

The cutting speed has an immense influence on tool life, as a comparison between parameter sets A and C from Table 4.1 shows. An increase in cutting speed of 10 m/min leads to a decrease in tool life by a factor of three. This effect is much less pronounced at lower feeds as can be seen when comparing parameter sets B and D. Parameter sets A and D lead to an optimal tool life close to 20 min. Both sets show a smooth wear pattern, shown in Figure 4.5. Despite the material consumption with parameter set D being significantly lower than with parameter set A, A is chosen to be the reference parameter set for turning, as it is closer to the recommended process parameters.

Parameter Symbol Unit Startvalue SetA SetB SetC SetD

Cutting vc [m/min]26 50 50 60 60 speed Feed f [mm] 0.28 0.28 0.2 0.28 0.2

Depth of cut ap [mm] 2 1.5 1.5 1.5 1.5 Rep. n = 1 n = 1 n = 3 n = 3 Results Tool life T [min] 21 30 7 19 3 Total mate- Vtot [cm ] 446 782 170 342 rial removal

Table 4.1: Diﬀerent parameter sets tested in external face turning with the corresponding tool life and removed material volume. 4.3 Orthogonal Turning Process 41

If there were a strong dependence of the wear progression on the current radius in face turning, this would limit the effective resolution because of an uneven wear progression. The extent of this effect is tested by using one tool for only the outer half of the face while using a second tool for the inner half. The results are summarized in Table 4.2. Although a small effect of the radius is likely present and justifies the use of the face turning process, a significant drop in resolution through non-uniformly spaced tool lives is not expected.

The reference process is later adapted for internal turning. The same cutting speed vc, feed and depth of cut ap as in external turning are used. The workpiece is predrilled to a diameter of 43 mm. It is machined outwards to a diameter of 275 mm in four steps of equal duration. After each step, the wear is measured. The resulting tool lives are higher than in external turning, as is shown in Table 4.3, which can be traced back to the higher clearance angle in cutting direction. At the same worn volume, a higher clearance angle leads to a smaller ﬂank wear land, which in turn results in longer tool life. The low pressure is used by the internal MWF unit. The lower pressure slightly decreases the tool life and slightly increases the variance of the result, which is accepted when facing the massively decreased cost for a test.

4.3 Orthogonal Turning Process

The reference process described above produces helical chips with a non-uniform cross- section. This poses problems when trying to prepare chip cross-sections and reliably determine geometrical parameters, such as the mean chip thickness. An orthogonal cutting test, which could be performed with the internal MWF unit, is therefore devised. Similarly to external orthogonal cutting, grooves with a width of 1.5 mm are prepared, leaving ﬁns in between with a width of 1.8 mm. The internal grooving turning head SL-QD-LGH32C32 by Sandvik Coromant is then used with the insert QD-NH-0400-0003-CR 1125 by Sandvik Coromant. This insert type features a straight cutting edge, but has two ridges at the corners of the cutting edge to bend the chip in a transverse direction to improve extraction

Parameter Symbol Unit Outer half Inner half Rep. n = 3 n = 3 Results Tool life T [min] 18.8 ± 0.3 20.3 ± 1.7

Table 4.2: Tool lives in tests performed with the reference parameter set, set A from Table 4.1, at the outer half of the faces and the inner half. The average value with the standard deviation of the samples is indicated 42 4. Turning Tests

Parameter Symbol Unit External Internal turning Internal turning Turning high pressure low pressure Pressure p [bar]70 70 16 Rep. n = 1 n = 3 n = 3 Results Tool life T [min] 21 26.5–27.8 22.2–24.4

Table 4.3: Comparison between external and internal turning tool life, using the reference parameter set, set A from Table 4.1. from the groove. This action is unwanted in this case, therefore the insert with a width of 4 mm is used to cut the fin with a width of 1.8 mm. In this manner, an orthogonal cut with the setup shown in Figure 4.6 is possible. Unfortunately, the grooving insert features a different micro-geometry than the insert used in the reference process, leading to different chip characteristics and a different wear behavior.

1.5 mm 1.8 mm

workpiece vc insert chip feed

Figure 4.6: Sketch of the internal orthogonal cutting setup.

4.4 Cutting Forces in Turning

Typical force measurement data is shown in Figure 4.7. At every inspection stop, the tool leaves contact with the workpiece and the forces drop to zero. Shortly after leaving the workpiece, the data acquisition is stopped, leading to a shorter representation of the interruptions than in reality. After every fourth segment, the outer diameter with the 4.5 Chip Flow Model 43

corner radius of the previous section is reached, leading to a peak in the forces due to the increase in cutting width and undeformed chip cross-section area. In the first 200 s, a run-in phase is observable with the forces first rising quickly but then saturating. In the stationary wear phase following afterward, the forces rise almost linearly. The cutting force only rises slowly as only a small portion of it is caused by the micro-geometry whereas most of the cutting force is caused by the chip shearing. The two other force components rise quicker, as they are determined to a large extent by rubbing of the primary and secondary flank face on the workpiece which is in turn strongly influenced by the micro-geometry of the cutting edge.

feed force 1200 cutting force passive force 1000

800

600

400 force components [N] components force 200

0 0 500 1000 1500 time [s]

Figure 4.7: Unﬁltered force data of a tool used with the reference parameter set, set A from Table 4.1.

4.5 Chip Flow Model

The insert’s macro-geometry and its oblique mounting position lead to a complicated chip formation situation, especially at the beginning of the cut when the undeformed chip cross- section is continuously changing. Semi-empirical cutting force models such as the Kienzle or Altintas model describe the degressive behavior with increasing feed in orthogonal cutting. They cannot describe the progressive behavior, which is observed when a tool 44 4. Turning Tests

with a large corner radius enters the workpiece. According to Figure 4.8, the cutting force rises faster than the undeformed chip cross-section area does. Tool run-in by wear can be excluded as possible reason for this increase, as can be proven by using the same tool twice, as shown in Figure 4.9. The forces diﬀer by less than 5% between the two cuts, whereas the force rises 20% more than the undeformed chip cross-section area does.

700 ]

0.5 2 600

500 0.4

400 0.3 300 0.2 200

cutting force [N] force cutting 0.1 100

0 0 [mm chip area undeformed

2.5 3 3.5 4 4.5 5 5.5 6 6.5 time [s]

Figure 4.8: Comparison between the cutting force measured (blue) and the calculated undeformed chip cross-section area (red) in the beginning of a cut.

The cutting forces could be modeled well with ﬁnite element methods; however, this is only possible with big computational eﬀort and does not reveal the reason for the increase in the cutting force. Therefore, a new model based on the Altintas model is developed. While the model would be identical to the Altintas model in the case of a straight cutting edge, Shi’s [134] quadratic term for force increase through restricted chip motion is added to describe the force increase with curved cutting edges. The variable term of the Altintas model is extended:

2 2 dFChip = kA · dA[1 + ku.r.(θ − θu.r.) + iu.r.(v − vu.r.) ] (4.1)

Equation (4.1) features FChip as the force exerted orthogonally to the rake face by the chip, a constant kA linking cross-section area and force, the weight ku.r. of the quadratic increase in force due to deviations from the optimal local chip ﬂow angle through restricted chip 4.5 Chip Flow Model 45

700 feed force cutting force 600 passive force 500 400 300 200 100 force components [N] components force 0 0 1 2 3 4 5 6 7 8 9 10 11 time [s]

Figure 4.9: Forces occurring when performing two short cutting tests with the same cutting edge.

motion, and the weight iu.r. of the quadratic increase in force due to deviations from the optimal local chip flow speed through restricted chip motion. Here, Shi’s model is applied to the oblique cutting situation in the corner radius of the insert. The situation in the beginning of a cut when the cutting tool enters the workpiece is studied. The different phases of the undeformed chip cross-section are depicted in Figure 4.10. The rake face is not planar in the corner part of the insert, as can be seen in the 3D rendering in Figure 4.11, preventing direct application of Shi’s model. The chip is assumed to behave like a thin shell element with no transversal stiffness. The chip is further assumed to conform to the rake face. The rake face is flattened geometrically without distorting the surface as shown in Figure 4.12. Because of the rake angle, the flattening increases the corner radius and angle. In the flattened view of the rake face, the chip behaves like a rigid body with three degrees of freedom: two translational and one rotational. The unrestricted chip’s motion directions, indicated in Figure 4.12 by blue arrows, are determined according to Stabler’s law [141] by calculating the local inclination angle and adding it to the normal direction indicated in Figure 4.12 by green lines. It is assumed that chip motion establishes in a way that the cutting force is minimal. Because of the quadratic nature of the model, this problem is solved by a least squares algorithm. The resulting three components of chip motion are used to calculate the local chip speed 46 4. Turning Tests

a) b) workpiece

tool feed

c) d)

Figure 4.10: Undeformed chip cross-section during different phases in the beginning of a cut. in every point on the cutting edge as shown in Figure 4.12 with red arrows. The local chip speed is then used with Equation (4.1) to determine the local force exerted by the chip on the rake face. The rake face is curved again to its initial shape to obtain the correct orientation of the local forces. A second force component is produced by friction on the rake face. It acts against the local chip flow direction. Knowing the amount of orthogonal force and the direction of chip motion for each element after curving the rake face again, the local friction force is estimated with Coulomb’s law using a coefficient of friction µ:

dFF ric = µ · dFchip (4.2)

The third force component is equivalent to the constant term in the Altintas model and summarizes the constant eﬀects of the cutting edge micro-geometry in cutting direction. It acts in the local plane containing the cutting direction and the cutting edge, orthogonally to the cutting edge. It has to be only considered for the engaged part of the cutting edge 4.5 Chip Flow Model 47

Figure 4.11: 3D-rendering of the cutting situation after cutting for 1.8 revolutions. and is calculated with the constant kc, which represents the cutting force per unit cutting edge length in the orthogonal case:

dFc = kc · dL (4.3)

The fourth force component summarizes the constant terms in feed direction, e.g. the eﬀect of the cutting edge micro-geometry. It acts orthogonally to the local plane containing the cutting direction and the cutting edge. It is as well only considered for the engaged part

of the cutting edge and is calculated with the constant kf , which represents the feed force per unit cutting edge length in the orthogonal case:

dFf = kf · dL (4.4)

As a result, the local distribution of all four force components, shown in Figure 4.13, is obtained. The four components are summed and integrated over the whole cutting edge to get the resulting force. The force is then split again into a component in cutting, feed, and passive direction to compare it to the measured force components. The process is repeated for every time step with the corresponding position of the cutting edge in the workpiece. 48 4. Turning Tests

1.2

0.8

0.6 [mm] 0.4

0.2

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 [mm]

Figure 4.12: Unwrapped insert rake face example for the situation shown in Figure 4.10d. The cutting edge is shown in gray. Normal vectors of each segment are shown in green. Blue arrows indicate the unrestricted chip ﬂow. Arrow thickness indicates the local chip thickness. The black arrow indicates the movement of the chip centroid. Red arrows show the rigid body movement in every point of the cutting edge.

The parameters ku.f. and iu.f. are not adjusted for each cutting edge, as they should be constants depending only on the material and the chip formation mechanism. ku.f. is set 2 to a value of ku.f. = 1.6/rad from preliminary tests. Since the cutting speeds along the cutting edge are nearly equal and the chip rotation is negligible, the influence of iu.f. in the presented turning setup is small. The value iu.f. is therefore set to zero. The influence of friction is significant, but hard to separate from the other force contributions. Therefore, the coefficient of friction of µ = 0.3 as determined by Wyen et al. [181] and Egaña et al. 2 [49] is used. kA = 1000 N/mm can be determined from the data of Wyen et al. [181]. kc and kf are determined by the actual micro-geometry of each individual cutting edge. 4.6 Reducing Variance in Turning Tool Life 49

tool cutting speed

feed

Figure 4.13: The four local force components. The chip deformation force dFChip is shown in yellow. The friction force dFF ric is shown in violet. The constant force dFc acting in cutting direction is shown in orange. The constant force dFf acting in feed direction is shown in green. For better visibility, the forces are shown in the direction they act on the workpiece.

They are determined with inverse ﬁtting. For the average cutting edge with a radius of

41.7 µm, the values are kc = 62 N/mm and kf = 101 N/mm. The transient forces are shown in Figure 4.14. The shape of the modeled forces ﬁts well to the measured curves. By adjusting the model parameters, the absolute values match as well. The ﬁrst abrupt change in the cutting force increase rate at 0.35 s corresponds to the time, when the main cutting edge reaches the corner of the workpiece shown in Figure 4.10b. After this point, the engaged cutting edge length increase rate is halved. The second abrupt change in the cutting force increase rate at 1.3 s corresponds to the time after one revolution shown in Figure 4.10c. After this time, the undeformed chip cross-section increases slower, as can also be seen in Figure 4.8.

4.6 Reducing Variance in Turning Tool Life

The repeatability of tool life tests causes major concerns, as the number of repetitions to obtain a mean tool life value with a given conﬁdence interval increases linearly with the variance of the tool life measurement. Furthermore, uncertainties stemming from the test setup or the wear assessment cover the inherent variances of the wear process, which may contain important information about the process itself. 50 4. Turning Tests

700 feed force 600 cutting force passive force 500

400

300

200

3D force components [N] components force 3D 100

0 0 0.5 1 1.5 2 2.5 3 3.5 4 time [s]

Figure 4.14: Simulation result of the forces occurring in the beginning of a cut with the reference parameter set from Section 4.2. The modeled forces are shown with bold lines, the measured ones with slim lines. The values for a cutting edge with an average radius of 43.9 µm around the corner is used. The parameters kc = 66.6 N/mm, and kf = 114.5 N/mm were determined for this cutting edge by inverse ﬁtting.

Using modern machine tools, variations in the process parameters can be neglected. The remaining factors contributing to the variance are: uncertainties when measuring the wear and variations in the workpiece material, tool geometry, surface, and substrate.

4.6.1 Illumination stage

Pictures of the flank face to evaluate the wear land width need to have a good contrast between the unworn and worn parts of the surface. The unworn part is usually dark and dull from the sintering or coating process, while the wear land is polished to a shiny appearance by the workpiece material. The highest contrast is therefore achieved by directing light to the flank face in a manner that the wear land reflects the light directly into the microscope. Conventionally, this is achieved using a ring light around the microscope. The ring light is mounted on an articulated stand and is adjusted individually for each insert. Depending on the illumination, the wear land can appear to be larger or smaller. To avoid this contribution to the measurement uncertainty, an illumination stage providing constant illumination is developed. A prototype with adjustable light positions and angles is shown in Figure 4.15. The final version would use fix positions. The flank face is illuminated using a single point high power LED on the left side. The flank face should have a good contrast to the black background and the bright wear land. The wear land is 4.6 Reducing Variance in Turning Tool Life 51

illuminated using a linear array of LEDs arranged in a cylindrical mirror on the right side. In comparison with the ring light, the light from below the corner allows the illumination of the shiny ﬂank wear land further around the corner. The higher light intensity reduces

the inﬂuence of the ambient light. The illumination stage leads to VBmax values up to 0.02 mm larger than the ones obtained with the ring light. The light coming from below is more dazzling for the user than the one from the ring light. Both reasons led to a poor acceptance in an industrial environment. The development of the illumination stage was halted and the ring light is kept in use.

point light linear light source source

clamping/ positioning

wear land to be measured

Figure 4.15: Prototype of the illumination stage.

4.6.2 Interpolation of the Wear Trend

The VBmax can only be measured optically at discrete times. Every measurement is only performed once and has an uncertainty. Assuming that the tool wear curve does not contain any high frequency components, a better estimate of the real tool life is obtained

by filtering VBmax over time. The filtered value can be used to estimate the time at which the wear reaches the critical value with a resolution smaller than the wear measurement intervals. The data is filtered by fitting a linear or quadratic function to the data of the last minutes. The lowest variations in tool life at constant conditions are achieved when the data of the last four minutes is used and a linear function is fitted. The intersection of this line with the tool life criterion is evaluated as the tool life. 52 4. Turning Tests

4.6.3 Automatic Wear Measurement

A further source of uncertainty comes from the manual measurement of VBmax. Con- ventionally, the cutting edge is manually aligned vertically before taking a picture. By clicking on a point of the wear land on both sides, where it seems to be the widest, the user determines the value. Thus, the measured value is greatly dependent on the accuracy of alignment and the individual assessment of the wear land border. The whole process can easily be automated using computer vision techniques, this being much quicker and the results being more reproducible. Python 3.4 is used with the libraries “tkinter” for the GUI, “numpy” for handling the data, “openCV” for image processing, “python docx” for report generation, and “pyinstaller” to freeze the application into a standalone executable. The algorithm works on grayscale images, therefore color images are converted to grayscale, as shown in the left side of Figure 4.16, by averaging the color channels. A histogram of the brightness values, shown in the right side of Figure 4.16, is calculated to dynamically adjust certain thresholds. The peak in the middle of the histogram corresponds to the brightness of the flank face. Any pixel brighter than 60% of the average flank face brightness is set to white, as shown in the left side of Figure 4.17, in order to avoid double peaks in the subsequent step. The image is differentiated in the horizontal direction, as shown in the right side of Figure 4.17. The edge position is determined as the maximum of a parabola that is fitted through the maximum point and its two neighbors, resulting in sub-pixel accuracy. The image is then adjusted roughly by fitting a straight line through the edge points in the lower half of the image. The image is shifted and rotated so that this line is positioned in the vertically and in the middle of an adjusted image. The vertical position of the insert is then detected by summing the pixel values in the right half of the adjusted picture in the horizontal direction, and detecting the jump from the background brightness to the flank face brightness. Now that the vertical position of the insert is known, the fitting range for the unused part of the cutting edge can be determined more accurately. Using a line fitted to all edge points in the unused range of the cutting edge, the adjustment of the image is improved. Using the histogram generated in the first place, the wear brightness threshold is set to a value at a given ratio between the peak of the histogram and the maximum brightness. The adjusted image is binarized using the wear brightness threshold. The region of interest, which is predetermined by the engagement conditions of the insert and the maximum expected wear, is selected and the pixel values in this region are averaged in the vertical direction. The average is shown in Figure 4.18. Coming from the right side, the first intersection of this curve above a predefined level is declared to be VBmax. The threshold value usually corresponds to a wear flank height of about four pixels. In the case of non-uniform wear, this can lead to a smaller value than 4.6 Reducing Variance in Turning Tool Life 53

the actual VBmax but is much more robust against dirt on the tool or noise from the image sensor. pixel count [-] pixel

brightness [-]

Figure 4.16: Grayscale image with corresponding histogram.

Figure 4.17: Image with dynamic threshold and corresponding derivative in the horizontal direction.

The algorithm is presented in a GUI for ease of use. Both the image adjustment and the wear evaluation can be manually overridden by the user in case the algorithms failed. Batch processing as well as report generation in .docx format are supported by the GUI.

4.6.4 Assessment of the Initial Cutting Edge Micro-Geometry with Optical Measurement

The following section is published in [92] in a similar form. 54 4. Turning Tests average pixel value [-] value pixel average

Figure 4.18: Bottom: Binarized image of the wear pattern. Top: Pixel value averaged in the vertical direction.

The insert type selected for the reference process features a cutting edge radius between 35 µm and 50 µm. It varies between different inserts as well as along a single cutting edge. The resulting variations in tool life are acceptable in the industrial environment and are met with conservative insert replacement intervals. However, in tool life testing, variations lead to substantially increased costs. Optical measurement is the current state of the art in cutting edge measurement but requires expensive equipment and the full measurement of a cutting edge takes up to 30 min with the InfiniteFocus G4 by Alicona. An Alicona InfiniteFocus G4 is used to measure the cutting edge prior to the quick-cutting test and the tool life test. It uses focus variation to obtain a 3D surface. For the first image, the main cutting edge is aligned horizontally. It is angled so that the flank and rake face are visible at the same time. A first profile is extracted by slicing the surface at a distance of 1.5 mm from the corner, therefore 0.3 mm before the end of the straight part of the cutting edge, as shown in Figure 4.19, profile a. This profile is then precessed with the algorithm of Wyen et al. [182], requiring no user input and therefore being very objective and reliable. The insert is then rotated so that the corner is facing to the microscope, 4.6 Reducing Variance in Turning Tool Life 55

making the whole flank and rake face visible along the corner. A second set of profiles is then collected along the curved corner part of the edge, as shown in Figure 4.19, profiles b. For this purpose, the position and direction of the cutting edge must be known, in order to slice the measured surface perpendicular to the cutting edge. Deviations from the perpendicular plane lead to a seemingly blunter profile in single profiles, as shown in Figure 4.20a. As this error is of second order, it is only relevant for large deviations. When averaging multiple adjacent profiles in order to reduce noise, smaller deviations lead to blunting as well in a first order effect, as shown in Figure 4.20b.

1.5 mm (0.3 mm)

r =1.2 mm cutting edge

ε a b2 feed

v b1 c x tool z y

Figure 4.19: Position of the slicing plane as seen in the direction of cutting speed. One profile in the straight part of the cutting edge (a) and multiple profiles along the corner radius (simplified, b).

a) b) Z (height) Z (height) Z

Figure 4.20: a) Apparent blunting effect for large deviations (red) from the orthogonal slicing plane (blue, dashed) b) Apparent blunting effect for small deviations (red) from the orthogonal slicing plane (blue, dashed) and by averaging over several neighboring profiles.

The cutting edge represents a 1D ridge of the 2D surface height scalar field in the mathematical sense. There are many different ridge definitions, but they only differ slightly in 56 4. Turning Tests

the given case of the very distinct cutting edge. In this case, the most common definition, the height ridge, is used. Height ridges are not invariant to monotonic transformation according to Norgard et al. [105], but this is not a drawback for the present application. A 1D height ridge of a 2D scalar field f is a line consisting of points where f has a local maximum in the direction of the maximum concave curvature at the same point. The direction of maximum concave curvature is given by the eigenvector corresponding to the smallest eigenvalue of the Hessian matrix H(f). However, this condition can be evaluated more computationally efficiently without solving the eigenvalue problem by checking for parallelism between the given gradient ∇f and the Hessian matrix H(f) times the gradient for each point p of the scalar field [113]:

H(f(p))∇f(p) = λ∇f(p) for some λ ∈ R (4.5)

Evaluating the cutting edge surface data leads to many points that belong to height ridges in a mathematical sense, but are caused by surface roughness and measurement noise and do not belong to the cutting edge. By using the MSAC algorithm [149], a modification of the RANSAC algorithm [53], points that do not fit well to a mathematical model, are discarded as outliers. In this case, the corner circle of the insert, with a fixed radius and inclination but with variable position, is used as mathematical model. The remaining inlier points are then used for the fitting of the cutting edge, as shown in Figure 4.21. The surface data is then sliced at 1800 equally spaced points perpendicularly to the identified cutting edge. Nine adjacent profiles are then averaged to obtain 200 profiles that are more robust. The profiles are then processed with the algorithm of Wyen et al. [182] shown in

Figure 2.5, resulting in the parameters Rn, R, ∆r and asymmetry S as a function of the position along the insert corner radius. The data is summarized in Table 4.4

0.45

0.4

0.35

0.3

0.25 y position [mm] position y 0.2

0.15

0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 x position [mm]

Figure 4.21: Each point denotes a point of a height ridge. Points identiﬁed to be on the cutting edge are represented by blue circles, outliers by red crosses. The resulting ridge-line is shown in blue. Iso-lines of the raw surface data are shown in the background. 4.6 Reducing Variance in Turning Tool Life 57

Parameter Unit Average Standard Minimal Maximum value deviation value value

Rn straight [µm] 39.3 2.2 35.2 44.5 n = 40 R straight [µm] 35.1 1.9 30.9 39.2 n = 40 ∆r straight [µm] 22.5 1.5 19.8 25.6 n = 40 S straight [-] 1.32 0.10 1.12 1.52 n = 40

Rn corner [µm] 41.7 3.0 37.1 47.6 n = 40 R corner [µm] 36.7 2.3 32.9 41.0 n = 40 ∆r corner [µm] 24.6 1.7 22.3 28.5 n = 40 S corner [-] 1.49 0.30 1.15 3.10 n = 40 T [min] 31.0 3.4 25.9 34.9 n = 8

Table 4.4: Overview of the micro-geometry values measured optically and tool life.

A tool life test is performed with 8 of the 40 cutting edges using the reference parameter set described in Section 4.2 with the internal MWF unit and the reference emulsion. The tool life is determined using the interpolation method described above. Pearson’s correlation coeﬃcient between the tool life and each of the eight optically measured values is calculated

and shown in Figure 4.22. The mean cutting edge Radius Rn of the corner part shows the strongest correlation with the tool life (r = −0.93), meaning that 87% of the tool life variance can be explained by the variance in the cutting edge radius. The shape of all cutting edge micro-geometries is very similar, with the difference only being the scale of the rounding. Since Rn, R and ∆r are all linearly dependent on the scale, they contain the same information and are all correlated similarly well with the tool life. The asymmetry parameter S however describes the shape itself which has a small variation and is therefore only slightly correlated with the tool life. Assuming that the variation in the cutting edge radii is only originating from variations in the cutting edge preparation process, it may be derived from the negative correlation that the optimal cutting edge radius in the corner for the given process is smaller than the measured average cutting edge radius of 41.7 µm from Table 4.4. Despite all the tools coming from the same production lot, the substrate properties may vary because of slight differences in pressing or sintering. It is well possible that the variations in the cutting edge radius are only an effect of variations in the substrate properties, e.g. a higher abrasion resistance leads to both a smaller cutting edge radius and longer tool life. Using the preliminary measurement data, two approaches to reduce the variance in tool life can be taken. On one hand, the tool life obtained with a specific cutting edge is compensated, meaning that tool lives obtained with cutting edges with a larger-than- average radius are multiplied by a bonus factor larger than one to compensate for the 58 4. Turning Tests

0.8 icient [-] icient �

0.6

0.4

0.2

Pearson correlation coef correlation Pearson Rn R r S Rn R r S

optically measuredΔ optically measuredΔ values straight part values corner

Figure 4.22: Pearson’s correlation coeﬃcient between the diﬀerent optically measured parameters on one hand and the tool life on the other hand. Positive values are shown in blue, negative values in red.

worse-than average cutting edge properties and vice versa. This approach relies on the assumption that the bonus factor is the same for all MWFs. On the other hand, the cutting edges with the average cutting edge radius deviating the most from the nominal value are discarded. The tool life distribution is asymmetrical, with a heavy tail towards short tool lives and a light one towards higher tool lives. It is therefore in most cases suﬃcient to discard the inserts with a low expected tool life, i.e. the ones with a large cutting edge radius.

4.6.5 Assessment of the Initial Cutting Edge Micro-Geometry with Process Forces

The same result as described above is sought without the need for long and expensive optical measurements of the cutting edge. Instead, the dependence of the process forces on the cutting edge radius is exploited. Preliminary measurement of the cutting force uses a portion of the tool life reserve. The preliminary force measurement duration is therefore chosen as short as possible. The scrap workpieces previously used with the internal reference parameter set for tool life tests are longitudinally turned to an internal diameter of 276 mm to achieve a smooth surface. The cutting edge to be measured is used to transversely turn to a diameter of 277.8 mm without holding time at the outer diameter. 4.7 3D Wear 59

The reference parameter set described in Section 4.2 is used. In this way, the same part of the cutting edge is subjected to chip load and the same mode of chip formation as in the full-scale cutting tests is expected. This quick test takes 3.6 s and is therefore not assumed to inﬂuence the remaining tool life measurably. The 40 cutting edges that were measured optically are tested with this quick test. The same eight tool life data points as in the previous section are used to check for correlation. In this process, the undeformed cross-section of the chip does not reach its steady-state. The relevant force component originating from the cutting edge micro-geometry is not easily separable from other force components in the measurement data because of the transient behavior and the non-trivial insert shape and angle. The model described in

Section 4.5 is used to inversely determine the parameters kf and kc separately for each cutting edge. An overview of the obtained parameters is given in Table 4.5. Multivariate linear regression is then used to obtain an estimator for the tool life based on the eight known tool lives:

min · mm min · mm T = 62.3 min + 0.248 · k − 0.457 · k (4.6) est N c N f

The tool life estimate Test correlates very well (r = 0.96) with the measured tool life. In the same manner as with the optical cutting edge measurement data, the force measurement data can be used to either compensate tool life measurement data or discard cutting edges with a low expected tool life. Under real tool life conditions, the variance can be reduced by a factor of two by discarding a third of the cutting edges.

4.7 3D Wear

If the 3D-distribution of wear is of interest, the region of interest with some adjacent areas is measured with a 3D-microscope before cutting. After cutting, the tool surface is

Parameter Unit Average Standard Minimal Maximum value deviation value value

kc [N/mm] 62.0 7.37 45.5 74.8 n = 40

kf [N/mm] 101 7.23 84.0 115 n = 40 T [min] 31.0 3.4 25.9 34.9 n = 8

Table 4.5: Overview of the micro-geometry values determined by inverse ﬁtting from cutting data and tool life. 60 4. Turning Tests

coated with adhering titanium workpiece material. The titanium is removed by etching with the method described in Section 7.1 to reveal the surface of the cemented carbide. The region of interest with the adjacent areas is then measured again. The unworn areas in the second measurement are manually selected and fitted to the first measurement to reveal the worn volume in three translational and one rotational degree of freedom, with the fixture keeping the remaining two angles constant. An example is shown in Figure

4.23.

μ [

m] [ μ

[ m]

Figure 4.23: 3D dataset of a worn turning tool (blue), after processing with the reference parameter set. The corresponding shape of the unused tool is shown as a green, semi-transparent overlay. The rake face is pointing to the left, the straight part of the cutting edge to the viewer.

In the illustration shown in Figure 4.23, the plastic deformation of the cutting edge is visible at the ﬂank face, where the used tool protrudes out of the shape of the new tool. The plastic deformation is however superposed with the ﬂank wear, rendering it impossible to quantify the deformation. 4.8 New Tool Life Criteria 61

A more easily graspable visualization as shown in Figure 4.24 is achieved by projecting the diﬀerence between the worn tool and the new tool in direction orthogonal to the tool reference plane. In this projection, quantitative results, such as the total worn volume on the rake face as well as the maximum crater depth, the position of the maximum crater depth and the crater width can be derived.

200 -5

400 m] μ

m] -10 μ [ 600 depth [ depth

-15 800

-20 200 400 600 800 1000 1200 1400 [ m]

Figure 4.24: Diﬀerence between worn and newμ tool shape, projected orthogonally onto the tool reference plane.

4.8 New Tool Life Criteria

New tool life criteria have to be more reliable, easier to measure or closer to the tool life in industrial applications. In roughing applications, the workpiece surface integrity is of minor importance. Therefore, the surface roughness or other properties of the workpiece are not considered as suitable tool life criteria. Instead, the force measurement data and the 3D wear shape are investigated as possible alternatives to VBmax.

4.8.1 Force Criterion

Force measurement data is measured and stored in all tool life tests anyway. The data can be evaluated automatically without the need for user input. In contrast to the optical wear 62 4. Turning Tests

data, the force data is available continuously during the whole cutting process, making it less sensitive to the influence of micro-spalling or other short events. From Figure 4.7 it can be derived that the feed force is the force component the most sensitive to wear. A tool life criterion based on a single force signal should therefore use the feed force. As the signals change slowly, the data can be simplified by averaging over one cut. The feed force data can be compared with the optically measured flank wear land width. If both wear measurements were equivalent, all curves would coincide. Instead, MWFs with a low lubricity and a high water content show lower forces at the critical flank wear land width. With the forces being similar for all products in the initial phase, it is clear that frictional forces are either too small to be relevant or are not influenced by the MWF. Therefore, the contra-intuitive behavior of the forces at higher wear levels can only be explained by a different worn tool micro-geometry dependent on the lubricity and the water content of the fluid. A tool life criterion based on the feed force would lead to longer tool lives for synthetic fluids while it would reduce tool life for oils. This does not reflect the actual tool life reserve. Tools used with grinding fluid sometimes failed even before reaching the optical

criterion of VBmax = 0.2 mm, whereas the tool life reserve of tools used with oil is much larger. This inaccuracy of the tool life criterion would be worsened by a tool life criterion based on the forces.

For all emulsions, the feed forcess in relation to VBmax are close to each other, shown in Figure 4.25. Therefore, if applied only for emulsions, the force criterion is a valid tool life criterion. Because the force is measured during the cutting process, tool life tests take a lot less time than with optical wear measurement. The experimental standard deviation of cutting force measurements is smaller than that of optical measurements. However, the range of variability between the products is narrower within the force measurements. When calculating the relative resolution as a ratio between the experimental standard deviation and the range of variability, the cutting force measurements are still advantageous. It is however a priori unknown, if a newly developed emulsion behaves similarly as the previous ones in terms of the force behavior. Furthermore, tool lives based on the force criterion are not fully comparable with tool lives based on optical measurements from the archive, rendering the data in the archive useless.

4.8.2 3D Wear Criterion

A tool life criterion could also be derived from the 3D wear data. The shape of the adhering titanium is not representative of the process, as it is generated in the last moment of the 4.8 New Tool Life Criteria 63

650 12.0 (oil)

22.6 (emulsion) 600 22.8 (emulsion) 23. 1 (emulsion)

550 21.2 (synthetic luid)

� 500 10.4 (grinding luid)

feed force [N] force feed 450 �

400

350 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 VBmax[mm]

Figure 4.25: Feed force plotted against the maximum ﬂank wear land width for diﬀerent MWFs. The curves are annotated with the average tool life [min] of the MWF and the type of MWF.

The conventional tool life criterion of VBmax = 0.2 mm is highlighted.

cutting process, with non-nominal conditions. Removal of the adhering titanium impairs the structural integrity of the cutting edge for reasons that will be shown in Section 7.1, leaving just one possible measurement per cutting edge. In this manner, the tool life cannot be measured sensibly. Instead of a full tool life test, the tool is used for a predefined cutting time and the wear is assessed only after this cutting time. The tool life cannot be estimated with this method, but the wear mechanisms with different MWFs are well comparable on tools with the same cutting time. With the reference parameter set the resulting crater shapes are typical for titanium turning. The crater follows directly behind the cutting edge, as can be seen in Figure 4.24, with the maximum depth near or even at the cutting edge. The position of the maximum depth is depending on the contact conditions between chip and tool and is not significantly influenced by the MWF. Furthermore, its determination is not accurate, as the crater is very shallow with a low curvature. The maximum crater depth is a valid wear criterion and can be determined automatically. The total worn volume would be an even more robust measure, but at a great measurement expense. 64 4. Turning Tests

4.9 Experimental Results

Typical commercially available representatives of each MWF class are tested in the turning setup. The results are summarized in Table 4.6. The variation of tool life results could be significantly reduced using the techniques described above. The differences between products of the same category remain small. They are smaller than the long-term variations, for example caused by aging of the MWF, degradation of the tool holder and machine tool, and changes in workpiece and tool properties. The chips shown in Figure 4.26 have pronounced shear bands with signs of cracks. No significant influence of the MWF on the chip shape with fresh cutting tools could be observed. This is in accordance with the initial machining forces being equal for all MWFs. The temperature measurement setup is validated by preparing two cutting edges of remaining wall thickness of 0.2 mm, 0.3 mm, 0.4 mm, and 0.6 mm each. The temperature is measured in five short cuts after the cutting edge has been used for two minutes. The cutting edge is used with pure mineral oil and the reference parameter set. The results are shown in Figure 4.27. Two possible extrapolations, one linear and one progressive are used to estimate the temperature at the rake face with a large uncertainty of 170 K. Depending on the extrapolation used, the minimal wall thickness of 0.2 mm leads to a temperature drop of 100 K to 270 K.

4.10 Conclusion

Not every MWF is suitable for titanium turning. For example, fully synthetic grinding ﬂuid, which is added as a negative reference, leads to tool lives half as long as those using suitable MWFs. All suitable emulsions and suitable fully synthetic ﬂuids show similar

MWF product Repetitions Tool life [min] Reference emulsion (mineral based) n = 5 28.7 ± 0.9 Emulsion (ester based) n = 3 28.1 ± 0.6 Universal emulsion (mineral based) n = 3 25.6 ± 0.6 Fully synthetic ﬂuid n = 5 30.2 ± 1.6 Neat oil (mineral based) n = 3 18.9 ± 1.0 Pure ester oil n = 3 19.7 ± 0.6 Fully synthetic grinding ﬂuid n = 3 13.4 ± 0.6

Table 4.6: Tool life results with diﬀerent MWFs tested with the reference parameter set and the internal MWF device. 4.10 Conclusion 65

Parameter Symbol Unit Synthetics Emulsions Neat oils Dry Temperature θ [◦C] 470 ± 20 480± 30 550 ± 20 640 ± 20 Tool life T [min]24 22 11 3

Table 4.7: Comparison between typical temperatures measured with a pyrometer for unworn tools and tool lives for diﬀerent MWF types. The temperatures are recorded at an oﬀset of 0.2 mm from the tool surface at a slightly elevated speed of 55 m/min to improve the signal to noise ratio, otherwise the reference parameter set is used.

500 m

Figure 4.26: Cross-section of a chip produced when cutting with oil. The microstructure is made visible by etching with Kroll’s reagent. μ results, despite their completely different composition. As will be shown in Section 7, suitable MWFs prevent the chemical degradation at the borders of the contact zone and leave the inherent dry wear at the cutting edge which is not susceptible to the composition of the MWF but strongly dependent on the local interface temperature. Temperature measurements at a distance of 0.2 mm from the interface are provided in Table 4.7 and correlate with the longest achievable tool life with each MWF class. Based on the turning tool life results with different MWFs of different types, including emulsions based on mineral oil and ester oil, fully synthetic fluids and neat oils, it can be concluded that there is little potential in developing improved MWFs for titanium turning with uninterrupted cuts. Because of the poor correlation between turning test results and the previously known performance in titanium milling, the turning test setup is not 66 4. Turning Tests

900 800 700 600 500 400 300

temperature [°C] temperature 200 100 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 remaining wall thickness [mm]

Figure 4.27: Inﬂuence of the wall thickness on the temperature with a pure mineral oil and the reference turning parameter set. The standard deviation of ten measurements is indicated with error bars. The region between progressive and a linear extrapolation is shown in red. suitable as an inexpensive and robust intermediate test method to predict the performance in milling either. 67

Chapter 5

Milling Tests

The results from the turning test bench shown in Section 4 can only partially be transferred to milling processes. Milling processes are the largest market for MWFs; therefore, they should be studied separately.

5.1 Milling Test Bench

5.1.1 Machine

Milling tests are performed on different three- or five-axis machines. Comparisons between different MWFs are always performed on the same machine to eliminate the influence of the machine type. Different machine types have different spindles, stiffnesses, and damping properties. These all have an influence on the tool wear. The tool is either completely flooded at no significant pressure or the tool is flushed internally with a pressure of 40 bar. The use of lateral nozzles placed on the spindle housing by the machine supplier is completely avoided, due to their poor efficacy compared to internal flushing and their effect being dependant on the adjustment by the machine user. The setup is shown in Figure 5.1

5.1.2 Tools

The tool assembly is shown in Figure 5.2. A square shoulder milling tool of type R390- 020A20-11L with two indexable cutting edges by Sandvik Coromant is used for the ﬁnish milling tests. A similar tool of type R390-020A20-11M by the same supplier with three indexable cutting edges is used for the rough milling tests. Both tools have a diameter of 20 mm. Inserts of type R390-11 T3 08M-MM S30T, optimized for medium cutting of 68 5. Milling Tests

measuring tool holder

milling tool

titanium workpiece

Figure 5.1: Milling test setup. titanium and with a twist angle of 6°, are used. They feature a corner radius of 1.2 mm and a 1.2 mm wiper section on the secondary cutting edge. The inserts are coated with an AlTiN layer in a PVD process.

Figure 5.2: Milling tools for ﬁnishing (right) and roughing (left) with the corresponding insert used for all tests. Image courtesy of Sandvik Coromant (modiﬁed).

5.1.3 Sensors and Data Acquisition

The milling tools are clamped in the tool holder with a “powRgrip” collet by REGO-FIX AG, Switzerland. The tool holder of type “Artis 4k-WISY” by Marposs AG, Switzerland measures bending moments in two transversal axes in its own, rotating coordinate system. The bending moments are almost entirely caused by radial forces and radial components 5.2 Reference Milling Process 69

of tangential forces acting on the tool, thereby allowing for the calculation of the radial forces through division by the tool overhang length. The position of the measurement point inside the tool holder is unknown. The overhang is calculated to be (107 ± 5) mm from a calibration with a dynamometer. The absolute values are not of interest as the data is used for comparative analysis only. In addition to the radial forces, the tool holder measures the axial force and the torque caused by tangential forces. The measurement data is obtained from strain gauges on a slightly weakened portion of the tool holder. It is transmitted to a computer in real time via Bluetooth, where it is stored and visualized. The maximum sampling rate is 2 kHz, equivalent to a sampling period of 0.5 ms. The tool holder is mounted on the spindle with an HSK-interface. After a track is ﬁnished, the machine is stopped, as is shown in Figure 5.1, and the cutting edges are measured with a light microscope.

5.2 Reference Milling Process

A straight shoulder down-milling process is chosen. It features a uniform uncut chip thickness along the whole main cutting edge, leads to the same conditions with fresh and used workpiece material blocks, and limits the influence of the machine tool. Preliminary tests showed that a single milling process parameter set is not sufficient to assess the performance of an MWF for titanium cutting. The ranking of MWFs depended strongly on the conditions at the cutting edge. Thus, two different reference processes are chosen. A finishing process, with a short engagement time per cut and a roughing process with a long engagement time per cut. The input parameters with the resulting engagement conditions are summarized in Table 5.1. Titanium blocks with a usable edge length of 250 mm, resulting in a track length of the same value, are used. The block is milled in horizontal layers using tracks from right to left. Once a track is finished, the tool is brought back to the right side without cutting.

It is then moved towards the back by the radial depth of cut ae and a new track is milled. After the last regular track of a layer is milled, the remaining material is removed with a diﬀerent tool. The axial position tool is then lowered by the axial depth of cut ap and the next layer is started in the front right corner.

5.3 Data Processing

Raw cutting force data of the measuring tool holder is shown in Figure 5.3. The raw data is hard to interpret as it is measured in a rotating coordinate system. To visualize it, the 70 5. Milling Tests

Parameter Symbol Unit Finishing Roughing Parameters Parameters

Cutting speed vc [m/min] 80 70 Tool Diameter D [mm] 20 20

Radial depth of cut ae [mm] 3 16

Axial depth of cut ap [mm]3 2 Number of teeth z [-] 2 3

Feed per tooth fz [mm] 0.1 0.1 Spindle speed n [min−1] 1273 1114

Feed rate vf [mm/min]255 334 3 Material removal rate Vtot [cm /min] 2.29 10.7

Engagement angle αeng [°] 45.6 126.9

Engagement time teng [ms] 5.97 19.0

Table 5.1: Different parameter sets used in milling tests. data is often shown in a scatter plot of the 2D radial forces shown in Figure 5.4. Despite the fact that two cartesian force coordinates are used, this type of plot is commonly called polar plot. In the case of finishing, only 14 data points are recorded on average per cut, rendering the peak cutting force Fmax and the exact shape of the cutting force uncertain. Assuming that the actual cutting force only changes minutely in consecutive cuts and further assuming that the sampling rate is not synchronized with the tool rotation, a more accurate force curve shape can be obtained using under-sampling approaches. A very simple and computationally inexpensive method, which loses temporal information of the signal, is presented here. It is only applicable if the angular coordinate of the cutting force in the rotating polar coordinate system of the tool holder is monotonically changing during a cut. This prerequisite is met by all cutting processes shown here. First, a train of several cuts is selected from the raw data. The raw data of radial cutting forces in two orthogonal axes is transformed to polar coordinates. The data is then grouped into bins according to the angle of the force. The number of bins is chosen according to the number of data points available. Usually, 200 to 400 bins are used. Each cutting edge has a characteristic force direction. The force mainly acts orthogonally to the rake face. Additional components from friction and the cutting edge micro-geometry lead to deviations from this direction. Using the number of data points per bin, a histogram is created. The number of cutting edges and their approximate direction is identified by the number of peaks and their position in the histogram. Borders between the cutting edge directions are defined to assign each data point to a cutting edge. This simple method 5.3 Data Processing 71

might lead to wrong assignment if the cutting force angle range of a single cutting edge is greater than 360° divided by the number of cutting edges. If two or more cutting edges are engaged at the same time, for example with the roughing parameters, the force distribution between the engaged cutting edges is not retrievable even with more advanced methods. The 95-percentile of absolute force values of every bin is computed. This value is set as the force value for the corresponding middle angle of the bin. The result is shown in Figure 5.5 and is a robust version of the envelope of the polar plot. The polar plot contains many outliers towards smaller forces caused by the start and end of the milling process, transmission problems, and under-sampling eﬀects. There are few outliers towards bigger forces as they are only caused by the measurement uncertainty, explaining why the 95-percentile is chosen instead of the median. A more elaborate but less robust method preserves the temporal aspect of the signal. It is depicted in Figure 5.6. The direction of the cutting edges and the borders between them are calculated as in the previous method. The signal is then cut into segments. A new segment is started when the magnitude of the force signal crosses the threshold value Fthr or when its angular coordinate crosses the border between two cutting edge angles. Segments within the threshold circle, shown in gray in Figure 5.6 are discarded and the other segments are assigned to the corresponding cutting edge. The segments can be analyzed for their maximum value, mean value, and length individually.

600

400

200

-200

radial force components [N] components force radial -400

-600

20 40 60 80 100 120 140 time [ms]

Figure 5.3: Radial raw force components measured in the rotating coordinate system. 72 5. Milling Tests

600

400

200

-600 -400 -200 200 400 600 radial forces [N] -200

-400

-600

Figure 5.4: Polar plot of the radial forces.

radial 0 200 400 600 800 1000 forces [N]

Figure 5.5: Polar plot of a new tool (blue) and the same tool after 16 m (red).

5.4 Wear Criteria

5.4.1 Automatic Wear Land Measurement

Similar to the wear assessment in turning, the maximum ﬂank wear land width VBmax can be measured. Because the cutting edge is curved, a measurement against a curve 5.4 Wear Criteria 73

cutting edge 1

s4 1

2D cutting → force

Fthr s3 NaN

→ O

s1 NaN

→ s5 2

s2 3 s6 3 → → → cutting edge 2 cutting edge 3

Figure 5.6: Method to analyze the raw 2D force data without losing the temporal information. The force threshold is indicated as a gray circle. The regions assigned to each cutting edge is shown with colors. The segments sn and their assignment to the cutting edge number is indicated. is required, as shown in Figure 5.7. The shape of this curve is strongly depending on the orientation of the milling tool. Therefore, the cutting edge offset through wear is neglected and VBmax is measured against the current, worn position of the cutting edge. The critical value is set to 0.2 mm. After at least one cutting edge reaches this value, the test is stopped and the distance milled to that point is considered the tool life under the given circumstances. To reduce the tediousness of tool wear measurement and the dependence on the user, the automatic tool wear evaluation algorithm for turning tools presented in Section 4.6.3 is adapted for milling tools. The cutting edge position is detected using an edge detection algorithm with sub-pixel resolution for every position along the vertical axis of the image. The cutting edge is then filtered with a low-pass filter to remove small imperfections of the cutting edge, such as small break-outs. The filtered cutting edge position is shown as thick red black in Figure 5.7 and is used as the base line for the flank wear land width measurement. The worn region is determined using the same thresholding techniques as used for the turning tools. The maximum width relative to the curved baseline is 74 5. Milling Tests

0.101

Figure 5.7: Picture of a cutting edge after milling 4 m, with a maximum flank wear land width of 0.101 mm. The automatically identified baseline is indicated with a thick black line, an offset line tangent to the wear land is shown with a thick red line. The offset corresponds to the flank wear land width VBmax calculated and presented as a measurement output. The same user interface as for the wear measurement in turning is used.

5.4.2 Force Criteria

As the milling forces are measured in each milling test, it is obvious to use them for a tool life criterion. Although milling forces are sometimes used in industrial applications as a criterion to exchange a tool, the threshold value is set separately for each process based on experience. There is no universally valid value as in the case of the ﬂank wear land width. 5.4 Wear Criteria 75

Nevertheless, the force is measured in the process, meaning that thousands of force values are obtained during a single track. Based on this data, the tool life feed distance can be determined with a resolution of less than 250 mm, the length of a single track, and the interval between optical measurements. If force values alone are used for tool condition monitoring, machine downtime during optical measurement is eliminated. Force data is insensitive to the environmental conditions, whereas the evaluation of the ﬂank wear land width relies on optimal illumination, focused images, and optimal tool orientation. Several indicators can be determined from the segments of force data. Each segment represents a single cut of the corresponding cutting edge. The following candidates can be calculated separately for each segment and are evaluated as possible tool life criteria:

• Maximum Force Fmax The maximum magnitude of the radial force vector

• Force Integral pcut The integral of the magnitude of the radial force vector over time, i.e. the impulse transferred. As the spindle speed is constant, this is also directly proportional to the energy transferred in one cut.

• Polar Integral Apolar The area enclosed by the radial forces in the polar plot. This value has the unit of force squared with no physical meaning.

• Rising Slope mrise The slope of the rising magnitude with respect to time of the radial forces in the beginning of a cut.

• Falling Slope mfall The slope of the falling magnitude with respect to time of the radial forces in the end of a cut.

• Engagement Time teng The time between the start and the end of a cut.

All of the above indicators are computed over the lifetime of a milling tool under ﬁnishing conditions. The results are shown in Figure 5.8. The values show a periodic drop, caused by the non-nominal engagement conditions of the milling tool in the beginning and end of each track. The limited sampling rate causes further variation towards lower levels.

Therefore, only the upper limit of the scatter band is analyzed. The engagement time teng is not depending on the wear of the tool, which is expected as long as no excessive burr formation occurs. The rising mrise and falling slope mfall show some sensitivity on wear, but do not allow for reliable quantification of wear due to the small change compared to the variation. The remaining three indicators, i.e. the maximum force Fmax, force integral p, and polar integral Apolar show a very good correlation with wear. pcut contains the same information as the Fmax, as teng is constant and the shape of the cutting force curve does not change significantly with progressing wear. pcut is much more difficult to compute, 76 5. Milling Tests

whereas Fmax and Apolar can be determined much easier and more robustly by discarding the temporal information and using the simpler method described in Section 5.3. Apolar is more sensitive to small disturbances, such as changes in the micro-geometry and cutting conditions than the maximum force. Fmax is smoother, allowing for an easier mapping to the continuous increase in wear. Both values are therefore selected as wear indicators based on the force measurement data.

5.5 Reducing Variance in Milling Tool Life

Reduced variation in the tool life results is required to obtain a reliable MWF rating with fewer repetitions.

5.5.1 Application of Force-Based Criteria

To compare the force-based indicators with the conventional optically measured ones, three milling tests under ﬁnishing conditions are performed with a universal emulsion. The tests are terminated based on the conventional criterion, using the maximum of the maximum ﬂank wear land width of all cutting edges. The optical wear progression results are shown in the bottom right of Figure 5.9. The tool with the highest life, shown in green, highlights the drawback of optical measurements. This tool nearly reaches the

critical value of VBmax = 0.2 mm after 16 m, like the tools from the other two tests. In the following 2 m however, no further increase was observed, leading to an exceptionally long tool life. With Apolar, the condition of tools can be monitored well near their end of life. However, this indicator is not suitable for reliable quantiﬁcation of small wear amounts.

Fmax shows a rather smooth, continuous increase with increasing wear. By deﬁning the tool life as distance, after which at least one of the cutting edges reaches a maximum force value of 900 N, as shown in the top left diagram of Figure 5.9, the distance by which the green tool outperforms the other two is reduced to 1 m. The initial maximum force

Fmax,0 does not have a big influence on tool life. Fmax,0 is therefore subtracted, as shown in the bottom left diagram of Figure 5.9, thereby limiting the influence of tool run-out and different micro-geometry. After the transformation, the two cutting edges belonging to the same tool are clearly grouped together, indicating mutual influence of one cutting edge’s wear state on the other. Although the maximum value of the two cutting edges are now spaced further apart than before the transformation, all six datasets can now be used to average the tool life, resulting in a more reliable value. 5.5 Reducing Variance in Milling Tool Life 77

Maximal Force Force Integral

4 800 3 [N] [Ns]

max 2 cut

F 600 p 1

400 0 0 5 10 0 5 10 number of cuts [-] ×104 number of cuts [-] ×104 ×105 Polar Integral ×106 Rising Slope 1.5

1 1 ] 2 0.5 [N

[N/s] 0

polar 0.5 rise A

m -0.5

0 0 5 10 0 5 10 number of cuts [-] ×104 number of cuts [-] ×104 1 ×106 Falling Slope Engagement Time 15

10 0.5 [N/s] [ms] fall eng

t 5 m

0 0 0 5 10 0 5 10 number of cuts [-] ×104 number of cuts [-] ×104

Figure 5.8: Overview of the indicator candidates for a milling test under ﬁnishing conditions. The value for cutting edge 1 is shown in blue, the value for cutting edge 2 in red.

5.5.2 Single-Flute Finish Milling Tests

In milling with multi-ﬂute tools, the cutting edges inﬂuence each other during cutting. A worn edge leaves more material on the workpiece and thereby increases the chipping load 78 5. Milling Tests

×104 1000 13

12 950 11 900 ] 2 10 [N [N] max

850 polar F

A 9

800 8

7 750

0 2.5 5 7.5 10 12.5 15 17.5 20 0 2.5 5 7.5 10 12.5 15 17.5 20 milling distance [m] milling distance [m] 200 ×104 4

150 3 ] 2 [N [N] 100 2 max polar F A Δ

Δ 1 50 0 0 -1 0 2.5 5 7.5 10 12.5 15 17.5 20 0 2.5 5 7.5 10 12.5 15 17.5 20 milling distance [m] milling distance [m]

0.25

0.2

0.15 [mm]

B,max 0.1 v

0.05

0 2.5 5 7.5 10 12.5 15 17.5 20 milling distance [m]

Figure 5.9: Force based indicators (top and bottom left) and conventional indicator (bottom right) for the same three ﬁnishing tests. The two cutting edges from the same test are shown in the same color. for the trailing cutting edge. For small, incremental wear, this process leads to a stable balancing of the wear of all cutting edges: More worn cutting edges have a lower chip load and thus wear slower and vice versa. However, bigger, sudden wear events, such as chipping of one cutting edge, lead to a sudden increase in the load of the trailing edge, which can induce even bigger chipping in the trailing edge. These events are especially common near the end of tool life, where they cause premature failure of the tool by a cascade of chipping events, induced by a single small chipping event on one cutting edge. 5.5 Reducing Variance in Milling Tool Life 79

Multi-fluted mills are industrially used because of their higher material removal rate at the same feed per tooth as single-fluted mills. Furthermore, vibrations are usually kept lower through the higher excitation frequency and the continuous engagement under roughing conditions. The requirements for tool life tests are different from the requirements in the industrial use of cutting tools. Therefore, single-fluted mills are suggested as a possibly advantageous solution for tool life tests. The total material removal per cutting edge at the end of tool life is not expected to depend strongly on the number of cutting edges. Therefore, the workpiece material use is expected to drop by a factor of two in the case of finishing and a factor of three for roughing. The information loss due to the absence additional, less worn cutting edges is marginal. The tests take the same amount of time, as the single cutting edge is not engaged for most of the time. To test the feasibility of single-flute milling tests as tool life tests, the finishing test is repeated five times with single-fluted mills. The test is again terminated using the conven-

tional tool life criterion of VBmax = 0.2 mm. The wear behavior in terms of the maximum force and the flank wear land width is shown in Figure 5.10. Despite filtering the signal with a moving-median filter, a characteristic waviness with the period corresponding to the track length is visible in the maximum force curve of the first two tests. In the middle of the third test, the used titanium block is exchanged for a new one. On the new block, the curves are very smooth. The waviness can therefore be attributed to inhomogeneous properties of the first titanium block, probably originating from a nonuniform heat treatment. Close to the end of tool life, some curves show sharp drops before rising again. This effect is caused by stochastic chipping of the cutting edge, which forms a new, sharper, although less stable, cutting edge. The flank wear land width increases much smoother than in the previous test as well. The average conventional tool life under the standard finishing conditions with double- fluted mills is 17.1 m. In the single-flute milling tests, an average of only 7.4 m is reached, less than half of the double-flute milling result. The most likely reason why the value is smaller than the expected 50% of the previous result is stronger vibration caused by the lower exciting frequency. Nevertheless, the empirical standard deviation is reduced from a value of 1.8 m or 10% with double-fluted mills to 0.5 m or 7% with single-fluted mills. To test whether the improvement in repeatability comes along with a reduced variability when testing different products, a high performance emulsion is tested with the single- fluted mills as well. The results of the two tests are shown in Figure 5.11. The high performance emulsion outperforms the universal emulsion by a factor of three. This value is even bigger than the performance improvement with double-fluted mills, where a factor of 1.8 is achieved. 80 5. Milling Tests

1000

800 0.2

600 0.15 [N] [mm] max 400 max 0.1 F VB 200 0.05

0 0 1234567 0123456789 number of cuts [-] ×104 milling distance [m]

Figure 5.10: Results of five single-flute milling tests under finishing conditions. The maximum force values are filtered with a moving median filter with a length of 161 cuts to improve readability. .

Vaughn [158] observed three stages of wear: A run-in phase, a uniform wear phase and a terminal phase with quickly increasing wear rate. In the previous tests with the universal emulsion, only the second and third phase are observed. With the high performance emulsion, the run-in phase is long enough to be visible in the maximum force data. In the uniform wear phase, the maximum forces are almost constant with both emulsions, while the flank wear width rises linearly. The forces in this phase are the same for both emulsions, suggesting that the high performance rather inhibits chemical deterioration of the tool rather than forming a stable film on the cutting edge. The uniform wear phase is much longer for the high performance emulsion. The terminal phase onset is therefore characteristic for the MWF’s quality, while the force increase in the terminal phase is determined by unknown factors. The two tests differ in the force increase rate in the terminal phase by a factor of two, despite they were run under the same conditions.

5.5.3 Single-Flute Rough Milling Tests

Single-flute milling tests are performed with the roughing parameter set. Two emulsions are used, the reference emulsion and a high-performance emulsion which is known to perform well under roughing conditions. Four tests are performed per emulsion, of which the results are shown in Figure 5.12. Similar as under finishing conditions, the maximum force stays at a nearly constant level, which is not influenced by the MWF. After a certain time, the maximum force increases suddenly as the cutting edge chips. The force does not increase smoothly before the first chipping occurs. Two or three tracks after the first chipping occurred, the test is stopped because the critical flank wear land width is reached. 5.5 Reducing Variance in Milling Tool Life 81

1000 0.3

800 0.25

0.2 600 [N] [mm] 0.15 max F 400 max

VB 0.1 200 0.05

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 0 5 10 15 20 25 number of cuts [-] ×105 milling distance [m]

Figure 5.11: Tests with a single-fluted mill and a high performance emulsion under finishing conditions. The results from the previous test are shown in gray as a reference. The maximum force values are filtered with a moving median filter with length 161 to improve readability.

Fitting an exponential wear model is not sensible in this case. However, the laborious measurement of the flank wear land width can be replaced with automatic cutting edge chipping detection leading to data with the same information about the tool condition. Single-flute milling tests do not come with the advantage of a smoother wear progression under roughing conditions. However, the workpiece material and insert use is reduced by a factor of three with negligible loss of information. Single-flute milling tests are therefore a valid option for tool life testing even under roughing conditions.

5.5.4 Robust Data Evaluation

The smooth wear progression with single-fluted mills under finishing conditions allows the thousands of force data-points to be simplified to a small set of numbers without information loss. These numbers can later be used in a tool life criterion. The force data from multi-flute cutting tests or cutting tests under roughing conditions are too erratic to be simplified with the following method. The run-in phase is neglected, as it is not critical in machining and is too short to give reliable information about tool life. The higher the wear is, the higher the cutting forces and therefore the temperatures get. This leads to a self-enforcing effect of the wear on the wear rate in the second and third phase. A generic exponential ansatz function is the solution of this linear positive feedback. An additional constant term is added to complement this model:

t Fmax(t) = Fmax,0 + Fc · e tc (5.1) 82 5. Milling Tests

800

700

600

500

[N] 400 max F 300

200

100

0 2 4 6 8 10 12 14 16 18 number of cuts [-] ×104 Figure 5.12: Single ﬂute milling tests under roughing conditions. Reference emulsion (red) and high-performance emulsion (blue).

Equation (5.1) uses three parameters: Fmax,0 is the maximum force offset. tc is a time constant determining the rate of positive feedback. Fc is a measure for the initial force increase, the initial wear rate at a given tc, or the time offset of the curve, which are all three equivalent. The model parameters are determined by fitting the model to the measured force data. The non-nominal conditions in the beginning and end of each track, small disturbances ( e.g. from jammed chips), and the sudden force drops caused by chipping of the cutting edge cause deviations from the ideal exponential behavior. A simple least squares fit to all data points therefore leads to resulting parameters which depend strongly on the stopping time of the test. The use of a more robust fitting algorithm is therefore necessary. The MSAC-Algorithm [149] is used to divide the dataset into a consensus set and outliers. The algorithm requires the determination of the three model parameters from a subset of three randomly chosen measurement values. The determination of all three parameters from randomly chosen points is inefficient. Fmax,0 is mostly determined by the measurement points in the beginning of the test for t ≪ tc when the exponential term is small. Fmax,0 is therefore determined as the average of the measured forces in the beginning of the test.

Fc and tc are determined analytically from two of the remaining points: 5.5 Reducing Variance in Milling Tool Life 83

t1 t −t Fmax(t1) − Fmax,0 2 1 Fc = F1 Fmax(t2) − Fmax,0 (5.2) t2 − t1 tc = − log Fmax(t2) Fmax,0 Fmax(t1)−Fmax,0

Equations (5.2) feature the measured maximum force Fmax(t1) at the time t1 and Fmax(t2) at the time t2. Outliers are defined as data points having a bigger distance from the model than a certain error threshold. After the parameters with the fewest outliers are found, the model is refined by using the least square method on the consensus set. The resulting three parameters are characterizing a single tool life test and are further processed to derive the quality of an MWF. The evaluation of force data with this method is shown in Figure 5.13. The algorithm successfully detects the chipping of the cutting edge and treats the following data points as outliers. Chipping of the cutting edge could be used as a tool life criterion as well, however, the variation is much bigger, which would require more tests to obtain reliable results. In Figure 5.13, an almost invisible irregularity at 7.2 · 104 cuts is detected as well and the corresponding points are discarded. An optimal value for the error threshold has to be determined for each process parameter set. Other than that, no user input is required. The values obtained for all single-flute finishing cutting tests are summarized in Table 5.2.

1100 900 1000

900 800 800 [N] [N]

max 700 max F F 700 600

500 600 400 0 1 2 3 4 5 6 7 8 4.5 5 5.5 6 6.5 7 7.5 8 number of cuts [-] ×104 number of cuts [-] ×104

Figure 5.13: Exponential model fitted to a single-flute milling test with the MSAC algorithm. Left: Measured force values (blue) and fitted model (red). Right: Detail of the same data, the detected outliers are indicated (red). 84 5. Milling Tests

MWF/Statistical value Fmax,0[N] Fc [N] tc [min] t1 N [min] T [min]

Universal emulsion 744 3.38 · 10−3 4.62 15.9 54.8 Universal emulsion 685 3.27 · 10−3 5.33 18.4 62.5 Universal emulsion 662 2.98 · 10−3 4.90 17.1 58.1 Universal emulsion 673 5.22 · 10−3 5.05 16.4 61.8 Universal emulsion 683 5.77 · 10−3 5.08 16.3 54.2 High-performance emulsion 682 9.06 · 10−6 10.0 60.3 180 High-performance emulsion 676 5.52 · 10−5 11.8 61.9 183 Repeatability 4% 28% 4.6% 5.2% 6.6% Variability 0.8% 98% 37% 57% 51% Resolution 538% 28% 12% 9.2% 13%

Table 5.2: Exponential force model parameters extracted from the maximum force data in comparison with the tool life T from VBmax for ﬁnishing conditions.

The observations made visually from the diagrams are confirmed. The initial maximum force Fmax,0 varies more among the tests with the same emulsion than it does between the emulsions. Even among the tests with the same emulsion, no significant correlation between Fmax,0 and the tool life T is observed. Fmax,0 is strongly influenced by the cutting edge micro-geometry, the workpiece hardness and small deviations from the nominal cutting parameters. On the other hand, Fc and tc are both correlated with tool life. tc indicates the time, after which the force increase reaches e times the initial force increase.

The initial force increase is not intuitively separable from the start force Fmax,0. The more

intuitive parameter t1 N is therefore calculated from Fc and tc and indicates the time, after which the force increase reaches 1 N. t1 N has a repeatability, defined by the standard deviation, of 5.2% in the five tests with the universal emulsion. The variability, defined as the quotient between the standard deviation of the mean values for the emulsions and the average parameter value is 57%, with the better emulsion leading to higher values. The resolution is estimated as the quotient between repeatability and variability, resulting in a value of 9.2%. t1 N is therefore superior to Fc and tc alone in every aspect to describe the quality of an MWF in terms of tool life. It is also more accurate than the conventional tool life criterion, which has a repeatability of 6.6%, a variability of 51%, a resolution of 13%, and is more time consuming to measure. 5.6 Conclusion 85

5.6 Conclusion

As a result of the investigations, milling test results could be made more reliable with the evaluation of the force data while saving at least 50 % of the workpiece material with the single-flute mills. The reduced variance in the tool life data allows for a reduction of the number of test repetitions while keeping the same confidence interval. The test duration of the milling tests themselves could not be reduced. With the finishing parameter set, the exponential wear behavior prohibits reliable tool life estimates until shortly before the actual end of life. Under roughing conditions, the end of tool life is reached even more abruptly. The cutting edges suddenly chip after a certain amount of unnoticeable damage is accumulated, without any precedent continuous wear. The wear behavior is strongly dependent on the cutting parameters as can be derived from the differences between the finishing and roughing parameter set. The MWF-ranking for both parameter sets is different. The tool life therefore cannot be extrapolated from quick tests at higher cutting speed or feed and tests have to be performed with industrially relevant parameters instead. 86 5. Milling Tests 87

Chapter 6

In-process Tribometer

The in-process tribometer as a model setup for the cutting process is regarded as the optimal compromise between simplification and relevance for industrial cutting. Simpler, commercially available tribometers can only work with oxidized titanium and most of them are not able to achieve severe friction conditions. Cutting tests, on the contrary, feature a large range of different conditions on a small scale, which renders the separation of the influence of temperature, pressure, and plastic deformation on the tribological system impossible. Furthermore, cutting tests consume an impressive amount of material, have a high variation, and are expensive. In cutting tools, wear has a feedback on the conditions that cause wear in turn, which complicates the evaluation. With an in-process tribometer, the severity of the friction condition can be tuned to match the most interesting conditions in the cutting process. The gradient of the relevant parameters, such as pressure and temperature in the contact zone can be smaller than in cutting, with a bigger area being subjected to similar conditions. This leads to more robust data. Compared to cutting tests, less material is consumed, the tests are shorter, and less expensive in general. An in-process tribometer has already been developed at IWF before the start of this work. It was tedious to use and prone to vibrations, but it provided valuable knowledge for a completely new design. The working principle stays the same as invented by Olsson et al. [107]. The newly designed tribometer is built and optimized, allowing for a new range of experiments.

6.1 History

The history of in-process tribometers at IWF started in 2009 with a ﬁrst design by Tobler [147] under the supervision of Wyen, shown in Figure 6.1. The pin was guided by a long 88 6. In-process Tribometer

swinging lever mounted on ball bearings. This lead to a reasonably straight motion of the pin and a low friction in the preload mechanism. The pin guidance was very stiﬀ in the direction of the lever (radial) but compliant in the direction of the friction force (tangential), increasing the tendency for vibrations. The preload was governed by a preload mass, acting over a pulley system. The low stiﬀness combined with the high preload mass lead to a low resonant frequency. The pin holder occupied a space of 4.5 mm radially inwards from the pin center axis, as visible in Figure 6.1b. 0.5 mm are needed radially outwards for the half track width plus another 0.5 mm as safety distance. This sums up to a total necessary shoulder width of 5.5 mm, of which the track occupies less than 1 mm.A depth of cut ap of 5.5 mm was not achievable with the given narrow tool at the necessary overhang. Therefore, a tube with a wall thickness slightly thicker than the track width had to be prepared by face grooving.

a b

Figure 6.1: Drawing of the ﬁrst in-process tribometer at IWF after [147]. 1: Force sensor, 2: Cutting insert, 3: pin, 11: lever.

Later, the preload mechanism was replaced by a coil spring, improving the dynamics of the pin guidance and reducing the tendency of vibrations. Tackling the other problems required a major rebuild. In a first version of the new in-process tribometer, shown in Figure 6.2, emphasis is put on low to medium normal forces. Stick-slip phenomena lead to a deviation from the desired normal force especially at low normal forces. Flexible material joints have therefore been used for the pin guidance, with the further advantage of easy and accurate stiffness predictions using FE-models. The pin holder is miniaturized, a small force sensor is used and all moving parts are weight-optimized to reduce the deviation from the desired normal force due to inertial forces on uneven sliding surfaces. The adapter on the machine side can be easily replaced to us the tribometer on different lathes. The standard cutting 6.1 History 89

tool holder can be replaced by a force measurement platform, allowing for tool condition monitoring or simple workpiece material monitoring.

Figure 6.2: New in-process tribometer design, with solid-state material joints.

The flexible material joints proved to be unreliable and the small usable stroke is tedious when setting the preload spring. In a reversible modification, they were replaced by two linear bearings with low friction. The linear bearings lead to a higher stiffness in the radial direction and reduce the tendency of vibration. In the last major redesign shown in Figure 6.3, experience from the older versions together with the established use-cases leads to a slight loss of universality but better performance. The preload nut of the force sensor protrudes the pin in axial direction, sacrificing the spiral-on-disc operating mode described in Section 6.6 for higher performance in the in- process mode. The distance between the linear bearings was increased and the overhang reduced. A redesigned pin holder was produced with the SLM process, allowing for the MWF duct and a coolant channel to be integrated. The preload spring is replaced with a pneumatic cylinder, making force adjustments more convenient allowing for new modes of operation. The tribometer should be usable on a turn-mill machine without turning turrets. It is therefore modified to the current version, which is mounted in the tool spindle and meets the tool weight requirements. 90 6. In-process Tribometer

MWF valve

linear bearings force sensor

pin pneumatic cylinder

cutting tool

Figure 6.3: Tribometer after major redesign.

6.2 Design

6.2.1 Requirements

The design process of an in-process tribometer is shaped by many different conflicting needs. Engineering decisions are made to find reasonable compromises. The conflicts can be understood best when following the workpiece material from the cutting to the friction zone. The cutting process should be as stable as possible. It should produce a surface with low roughness and waviness at a low tool price and a large range of cutting speeds. The lower

the depth of cut ap, the lower the tendency for vibration and waviness of the surface are.

The minimal ap is however given by the space needed for the pin in the friction zone. A large corner radius would increase the lifetime of the indexable insert but increases the tendency for vibration as well. A small entry angle would be beneficial as it leads to machining forces in the stiff longitudinal direction of the tool shank. This would however lead to a conical surface, which would require an adjustable angle of the pin-holder in turn to keep the preload force in the normal direction to the surface at every possible workpiece diameter. After being cut, the fresh surface moves past the flank face to the friction zone. The shorter the distance between the cutting edge and the friction zone, the weaker adverse oxidation effects become. This limits the possible tool holder thickness below the insert, reducing its stiffness and increasing the chance of vibrations. In addition, an MWF orifice 6.2 Design 91

needs to be placed in the place between the cutting tool and the friction zone, if lubricated contacts should be investigated. The closer the orifice is to the workpiece surface, the less MWF is spoiled for cooling and the closer the achievable temperature in the friction zone is to an industrial cutting process. In order to adapt to different workpiece diameters, the pin position must be adjustable in relation to the cutting edge. A short distance between the cutting edge and the friction zone reduces the necessary range and allows for smaller initial workpiece diameters. All other parts of the tribometer should stay outside of the envelope of all possible workpiece diameters in order to allow for an unlimited axial workspace, as shown in Figure 6.6. In the friction zone, the conditions should usually be as constant as possible. The pin should follow the surface and move over imperfections without a change in the normal force or lateral movement. This requires low stiffness in the normal direction and high stiffness in the tangential direction. The fluctuations in the normal force can be further minimized by reducing the waviness of the surface, reducing friction in the moving parts, and reducing mass of the moving parts. On the other hand, damping forces are advantageous to prevent the pin from bouncing. Vibrations in the direction of the pin’s sliding velocity can be caused by the stick-slip phenomenon. The pin-holder assembly is simplified to a single mass oscillator with a single degree of freedom. The equation of motion is derived by Brecher et al. [26]:

m · x¨ − C1 · x˙ + kf · x = kf · v · t − Fs (6.1)

Equation (6.1) features the equivalent mass of the pin-holder m, the coordinate x, the local derivative of the Stribeck curve in the point x˙ C1, the stiffness kf and the stiction force Fs. From the equation, it can be derived that the tendency for stick-slip effects can be reduced by reducing the mass or increasing the stiffness in the transversal direction. The shape of the Stribeck curve and the stiction cannot be influenced, as they are inherent properties of the tribological system to be investigated. The heat generated in the friction zone causes temperature changes in the pin-holder and the force sensor, leading to a drift in the measurements. The further away the sensor is placed from the pin, the better thermal insulation is achieved. The bigger the pin-holder mass is, the higher its heat capacity is. This is however conflicting with the dynamic requirements. Vibrations can only be observed with the force sensor up to a certain frequency, related to the resonance frequency. The mass between the friction zone and the force sensor is especially critical. At high frequencies, this mass leads to a difference between the actual 92 6. In-process Tribometer

contact forces and the measured forces, making the contact forces unobservable near and above the resonance frequency.

As ap is minimized, the safety clearance between the pin-holder and the workpiece shrinks to less than 0.1 mm. This is in the range of the chip thickness. As the chip is formed in front of the friction zone, actions have to be taken to prevent a chip from entering into the gap between the pin lateral area and the workpiece.

6.2.2 Solutions

First, the force sensor is selected. The 9027C piezo-electric sensor by Kistler Instrumente AG, Switzerland, is used because of its high stiffness, compact dimensions, and high dynamic range. It is used as a load-sensing washer and is preloaded with a tension bolt. The sensor is used with its longitudinal axis in the direction of the preload force. In this way, the thermal drift, which is mostly acting in longitudinal direction, does not affect the measurement of the friction forces. The drift in the preload force can be separated from the relevant data later during data evaluation, as the static behavior of the preload mechanism is well known. To allow free passage of the workpiece material, the pin is placed off-axis with a pin- holder, as shown in Figure 6.4. The clamping mechanism occupies more space than the sensor and is therefore limiting the minimal offset. The pin-holder is angled downwards to bring the pin as close as possible to the cutting edge, resulting in a 14 mm distance in the horizontal direction. The tribometer is adjusted to different workpiece diameters by clamping the shank tool at the corresponding overhang. In order to reduce the offset between the friction zone and the sensor in the axial direction, the pin-holder has the minimum required thickness given by the force sensor specifications and extends backwards around the sensor to securely clamp the pin. The pin clamp wraps around the pin with 225°, leaving the remaining 135° free to get as close to the workpiece at different diameters as possible. The design is prone to thermal drift, as there is no thermal insulation between the pin and the sensor. Therefore, a closed coolant channel is integrated into the design, as shown in Figure 6.5. Next to it, the channel for the MWF is integrated with an orifice just in front of the pin. These channels can only be produced by additive manufacturing, in this case the SLM-process was chosen. The pin-holder is made from 1.2709 tool steel in its as-built state. The functional surfaces are built with an allowance, which is removed by grinding, drilling, reaming and tapping. The sensor is mounted with its cable pointing to the right at an angle of 25°, thereby avoiding collisions between the cable and the workpiece while providing enough room for the upper linear bearing to be placed close to the friction zone as seen in Figure 6.6. The 6.2 Design 93

Figure 6.4: Partitioning between pin-holder (1) with pin (3) and cutting tool (2).

Figure 6.5: Translucent rendering of the pin-holder with a clamped pin shown in gray. The MWF channel is shown in red, the coolant channel is shown in blue. upper linear bearing is placed as close to the front as possible to take up the lateral forces with high stiffness. The lower linear bearing is set back 18 mm so that the indexable insert can quickly be replaced without unclamping the shank of the cutting tool. The two linear bearings are placed at a large distance of 68 mm in the vertical direction to be able to counteract the torque generated by the offset of the pin. Cross roller bearings are used because of their small installation space and high stiffness. An overlong roller cage is used, which limits the travel to 5 mm but further increases the stiffness. 94 6. In-process Tribometer

Figure 6.6: left: Front view of the tribometer, envelope of all workpiece diameters after cutting (red), linear bearings (green), pin-holder (yellow) with pin (blue); right: Side view of the tribometer.

All the moving parts are connected by a carriage. The carriage has a T-shaped cross- section when viewed from above, which is created by screwing the sensor mounting plate to the back plate. The T-profile is terminated with mounting panels for the linear bearings at the top and bottom. The otherwise poor torsional stiffness is thereby improved via the roller bearings and the solid stationary body. A shank tool with an entry angle of 90° for ISO-inserts of the type CC.. − 09T 3.. is chosen because of the wide availability of inserts. The tool is modified as shown in Figure 6.4 to allow the pin-holder to move freely. The shank is clamped with a double-wedge unit, which is the stiffest available option. The preload force is provided by a pneumatic cylinder with a 32 mm diameter. The cylinder is equipped with low friction seals. By connecting it to a pressure regulator, the cylinder provides the same force over the whole travel range, i.e. its static stiffness is zero. Both the pressure regulator as well as the valve connected to the other side of the cylinder are operated electronically, allowing for changing the normal forces during an experiment, automatic presetting of the normal force, and continuous control of the normal force.

6.3 Speciﬁcations

The diameter of the cut workpiece material must be larger than 30 mm as shown in Figure

6.6. With a depth of cut ap of 2.5 mm, this is equivalent to an uncut diameter of 35 mm. Smaller workpieces would collide with the sensor mounting plate. There is no upper limit for the workpiece radius; it is even feasible to use the tribometer in a shaping setup. 6.3 Speciﬁcations 95

The pin clamping can sustain normal forces of 500 N in friction-lock, allowing reliable measurements up to 400 N. Above that, a form-locking device at the rear of the pin-holder must be used. 563 N can be achieved with the pneumatic cylinder at a standard supply pressure of 7 bar. By increasing the pressure to the limit of the cylinder of 10 bar, the force can be further increased to 800 N. The pin geometry can be chosen freely, as long as there is a 3 mm shaft. However, too small and too large radii tend to vibrate at higher loads. A misalignment of 2° between the surface normal and pin axis has to be considered, therefore spherical pin shapes are preferred. The relative speed is limited by the spindle speed at smaller workpiece diameters or by the cutting process at higher workpiece diameters. The cutting process needs to be stable and produce a surface with the required properties. Usually, the lowest achievable roughness of the workpiece surface is desired to achieve a well-defined contact situation. Cutting insert wear should be negligible during one test in order not to change the surface properties during the test. As the cutting edge is poorly cooled, maximum cutting speeds for dry cutting may be assumed. In titanium, with cemented carbide inserts, cutting speeds of 160 m/min are possible. The measurement uncertainty in the coefficient of friction is dominated by the uncertainty of the force measurement. The uncertainty of the charge amplifier is 0.3% per axis, which would result in a combined error of 0.42% for the ratio used in the calculation of the coefficient of friction, assuming that the error is not systematic. The force sensor adds another 0.25% per axis. This results in a total uncertainty of 0.56%. In reality, at least a part of the error is going to be systematic, which would cross out in the calculation of the quotient. The friction in the preload mechanism changes the preload force by up to 3 N, meaning that the preload force can only be selected with an accuracy of 3 N. However, the effect of friction in the preload mechanism is fully observable with the force sensor and must therefore not be considered in the error budget. Under high loads, the head tilts slightly because of elastic deformation. This leads to an additional systematic load dependent error of 0.4%. In any case, the uncertainty is irrelevant due to the repeatability of the coefficient of friction, which is seldom lower than 10%. 96 6. In-process Tribometer

6.4 Heat Flow Model

6.4.1 Model Setup

The temperature distribution plays a major role in tribological processes. Bulk material properties, such as the yield strength are temperature dependent. Lubricants may even change phase. Surface reactions, such as adsorption, chemical bonding, or metallic adhesion, are strongly temperature dependent. It is argued whether the cutting process in front of the pin significantly increases the temperature in the friction zone of the pin. The temperature increase through the cutting process is even discussed as a possibility to deliberately increase the temperature in the friction zone of the pin to reach conditions closer to the cutting process. To check the feasibility of this approach, it is desirable to model the temperature distribution of the whole tribometer setup. A simple stationary heat flow model is set up for this purpose. Heat sources are the cutting process and the friction zone. Heat sinks are all exposed surfaces, which lose heat by convection and the clamped surface, which conducts heat away into the tribometer. Advective heat transfer due to the workpiece material motion is modeled with a rotational and an axial component. The heat input from the cutting tool is simplified as a line heat source. This leads to exaggerated temperatures in the near field but does not affect the far field near the friction zone. The input power is estimated based on the cutting forces in [181] from which the

cutting power is calculated. The fraction ηwp of the cutting power which is conducted into the workpiece as heat is estimated to be roughly 20% in [132]. The speciﬁc power of the ′ line source Pcut,wp therefore calculates as:

′ vc · Fc · ηwp Pcut,wp = (6.2) ap

Equation (6.2) further uses the cutting speed vc, the cutting force Fc, and the depth of

cut ap from an experimental data set or a cutting simulation. Plastic deformation is neglected. For the pin, this is a reasonable assumption. How- ever, the dissipation of the complete frictional power in the ﬂat contact zone leads to an overestimation of the contact temperature, compared to the real case, where the heat is generated in a shear zone in the workpiece with a certain thickness. The contact pressure for an elastic contact has a spherical distribution according to the Hertzian theory. The tangential shear stress is assumed to be proportional to the contact pressure according to 6.4 Heat Flow Model 97

the Coulomb law, therefore leading to a spherical distribution of the heat source density ′′ Ppin(r) over the contact radius as well:

2 2 −1 1 − ν 1 − ν 3 3FR E∗ = 1 + 2 a = E E 4E∗ 1 2 r 2 3F − r p0 = 2 p(r) = p0 1 2 (6.3) 2πa r a

′′ · · Ppin(r) = µ p(r) vc

Equations (6.3) feature the Young’s modulus E1 of cemented carbide and E2 of titanium,

Poisson’s ratio ν1 and ν2 for the same materials, the load force F , the pin radius R, the coefficient of friction µ, and the speed vc. ′′ Heat flow density Pconv on exposed surfaces due to convection and heat flow into the pin holder is modeled with a linear model:

′′ · − Pconv = h (θ θenv) (6.4)

Equation (6.4) features the heat ﬂow coeﬃcient h, the surface temperature T and the temperature Tenv of the environment. The heat transfer constant of the clamped surface is calibrated with temperature measurements in the center of the pin, 0.2 mm above the contact zone. The resulting value is 2000 W/(m2K). The heating of the whole workpiece on a longer timescale is neglected because of the short test duration. Computational costs can thus be reduced by simulating the heat distribution only on the relevant sector of the workpiece. Quadratic Lagrangian elements are used for the discretization. Further reduction in computation time is achieved by adapting the mesh size to the expected temperature gradient, resulting in a computation time of 1 min.

6.4.2 Model Results

The model is used to estimate the temperature of the workpiece surface which arrives at the pin. The results at ﬁve feeds and three speeds are shown in Figure 6.7. The temperature increase through the cutting tool is negligible in most cases. For example, when using

a cutting speed vc = 100 m/min and a normal force FN = 200 N in dry conditions with the standard feed of 0.04 mm, the cutting process leads to a temperature increase of 24 K. Most of the heat is conducted away from the surface by self-quenching. The temperature 98 6. In-process Tribometer

90 v = 100 m/min 80 c

60 vc = 60 m/min

T [K] T 40 Δ 30 vc = 20 m/min 20

0 0 0.05 0.1 0.15 0.2 feed per revolution [mm]

Figure 6.7: Simulated temperature increase at the pin through the prior cutting. Cutting power is calculated based on the data from Wyen et al. [181]. increase is even smaller if the convection on the surface is increased by the application of a fluid. Compared to the temperature increase in the friction zone of more than 700 K at a normal force FN = 200 N as seen in Figure 6.8, the influence of the cutting process is insignificant. However, if a higher temperature is desired, the feed can be deliberately increased, leading to a maximum temperature increase of 82 K at 0.2 mm/rev, which may lead to significant changes in sensitive tribo-systems.

6.5 Workpiece Surface Properties

The workpiece surface which arrives at the pin contact deviates from a ground and polished surface, as it is usually used in tribological tests. The cutting action leads to strain- hardening of the surface. This effect is minimized by using a sharp cutting edge with a cutting edge radius under 5 µm. The surface roughness should be as low as possible to obtain predictable contact conditions. The roughness transversal to the pin track is mostly influenced by the cutting edge shape, including BUE. Titanium cutting does not cause BUE-formation with the given conditions. The transversal roughness is therefore minimized by choosing a tool with a finely ground cutting edge. The roughness in direction of the pin movement is caused by the material separation and unstable BUE. As titanium is ductile and no BUE is observed, this part of the roughness is expected to be negligible. The roughness is therefore expected to be dominated by the cutting edge shape with no influence of the cutting speed or lubrication. 6.5 Workpiece Surface Properties 99

1400

1200

1000

800 temperature [K] temperature 600

400

Figure 6.8: Simulated stationary temperature ﬁeld in the in-process tribometer. The cutting edge is situated to the right. The workpiece material then moves to the left, where it is heated in the friction zone, leaving a heated track on the surface. vc = 100 m/min, FN = 100 N, µ = 0.5, dry.

To test this hypothesis, the roughness of a titanium workpiece is measured three times at 20 m/min, 60 m/min, and 100 m/min respectively without lubrication. Three additional measurements are performed at 60 m/min with a neat oil. As the surface of interest on the shoulder is not accessible with a stylus or a microscope, imprints of the surface are made with a 2-component silicone rubber. The surface is then measured with focus variation on an Alicona G4 and the values Sa and Sz are calculated with a cut-off wavelength of 100 µm. The measurement results are summarized in Figure 6.9. The difference in Sz for the different parameter sets is less than 300 nm, which is well within the scattering range of the data. This difference is equalized by plastic deformation through the pin under all reasonable load conditions. The influence of the roughness can therefore be neglected, especially if the tribometer is used for comparative analysis at the same speed. 100 6. In-process Tribometer

Sa Sz 0.4 3.5 0.35 3 0.3 2.5 0.25 2 0.2

[µm] [µm] 1.5 0.15 0.1 1 0.05 0.5 0 0

vc [m/min] 20 60 100 60 20 60 100 60

Dry Oil Dry Oil

Figure 6.9: Measured Roughness parameters of the workpiece at diﬀerent parameter sets. Three measurements are performed per set. The minimum and maximum value for each set are indicated with error bars.

6.6 Operating Modes

The in-process tribometer can be used in a variety of different modes. Some of them are designed to get the conditions as close as possible to the process which is to be modeled and others are designed to validate the tribometer design. The standard in-process configuration shown in Figure 6.10a uses the in-process tribometer at a constant speed and a constant normal force. In this configuration, the pin is gliding on a freshly generated and nearly unoxidized surface. Fluid is provided at a constant flow rate if required. After 5 s, the friction reaches a steady state. The test is usually carried on until a total duration of 10 s to 20 s is reached to get a more reliable average value. Usually, the wear of the pin can be neglected in this short time frame. If the pin wear is to be investigated, longer test durations of 4 min are chosen. Depending on the expected differences between different parameters, the test has to be repeated three to five times. The absolute values of the aforementioned test are the most scientifically meaningful and should be used when needing absolute values for the coefficient of friction. However, when testing new MWFs, a quick comparison over a whole range of parameters is necessary. To achieve this goal, the velocity is reduced in steps of 10 m/min from 100 m/min to 10 m/min. The holding time at the first step is 10 s and 3 s at each subsequent step. While the steady state of the coefficient of friction is not reached in this manner and absolute values are not useable, the results still allow for a reliable comparison of MWFs with minimal material usage and short test duration. 6.6 Operating Modes 101

In temperature dependent tribological systems, it can make sense to keep the frictional power constant. The frictional power is calculated using the formula for mechanical work:

Pfriction = µ · FN · vc = FT · vc (6.5)

Equation (6.5) reveals the three parameters of inﬂuence: the coeﬃcient of friction µ, the

normal force FN , and the velocity vc. The coefficient of friction is not a suitable control variable, as it is usually inherent to the tribosystem to be investigated. The velocity is controlled by the CNC, which can be influenced close to real-time using a work-around based on the Fanuc Focas Interface. The normal force can be controlled via the air pressure. In order to adapt to changes in the coefficient of friction, the tangential force is measured and fed into a PI controller. In order to compare the results to conventional closed tribometers, pin-on-disc tests can be performed by removing the cutting tool and setting the feed to zero. In this way, the pin is gliding over the same surface over and over, as shown in Figure 6.10b. The open tribometer configuration is similar to the tribometer by Rech et al. [122], but uses the bottom rather than lateral area of a cylindrical workpiece. A smooth, flat face is turned on the workpiece. The tribometer is then dragged from the outer diameter to the center, as shown in Figure 6.10c, while keeping the gliding speed constant by increasing the rotational speed. In this way, the pin is gliding on an unused, but strongly oxidized surface. a) b) c)

vf 1 1 1 3 vf 2 2 2

Fn Fn Fn

vc vc vc

Figure 6.10: The three diﬀerent main tribometer conﬁgurations. A: In-process, B: pin-on-disc, C: spiral pin-on-disc. 1: workpiece, 2: pin, 3: cutting tool.

Using the solenoid valve for the ﬂuid dosage, interrupted ﬂuid supply to the friction zone is possible. Frequencies up to 0.5 Hz are feasible. This emulates interrupted cuts as in 102 6. In-process Tribometer

milling, but at a much lower frequency. The behavior after turning off the fluid supply can provide information about the durability of an anti-wear or EP-layer on the pin. The behavior after opening the fluid valve gives information of the penetration ability of the fluid into the friction zone as similarly tested by Claudin et al. [35]. Spherical pin shapes are the easiest ones to use. They are unaffected by the angle at which they are clamped and the lateral angle due to the tribometer deformation. The elastic case at low loads is well described by the Hertzian model. The ultrafine-grained cemented carbide grade K40-UF with 10% of cobalt is chosen as the standard material, as it is often used for producing cutting tools for titanium. Small pin-on-disc tribometers often use a 6 mm ball point. The standard tip is therefore ground to a spherical shape with a radius of 3 mm to allow for the comparison of results. A coating with Balinit Latuma, an AlT iN-based hard coating by Oerlikon Balzers, Liechtenstein, is chosen for the standard pin to achieve a similar situation as in cutting. In some cases, a line contact is more favorable than a point contact. A solid pin with the lateral surface of a cylinder ground to the front is not feasible, as an even load along the contact line could not be assured. A compliant axis at a right angle to the contact line has to be introduced. The easiest way to do so proved to be the jigsaw-puzzle-like joint shown in Figure 6.11. The shaft is made out of steel to prevent the protruding part from breaking off. The moving part is disposable and can be made of any material of interest. A stop (not visible in the figure) counteracts the friction force.

Figure 6.11: Pin with line contact.

The temperature in the friction zone is of major interest. The small contact zone and high temperature gradients make the temperature measurement diﬃcult. Surface temperature measurement with an infrared camera is unsuitable in case of ﬂuid application. Subsurface 6.7 Experimental Results 103

measurements in the workpiece are unsuitable because of the surface removal by the cutting tool and the workpiece rotation. The remaining possibility are subsurface measurements in the pin. The pin can be EDM-drilled in the center down to a remaining wall thickness of 200 µm. A ﬁber from a pyrometer or a mantle thermocouple can be inserted into the hole to measure the temperature.

6.7 Experimental Results

In the following section, results regarding the friction forces are shown. The evaluation of the pin wear as a secondary use is described in Section 7.3.

6.7.1 Comparison of Diﬀerent Setups

In order to justify the additional effort to measure the friction in the process compared to a conventional pin-on-disc setup, the coefficient of friction of the three setups shown in Figure 6.10 is compared. The standard pin specifications are used. The preload force is set to 100 N and velocities of 20 m/min, 60 m/min, and 100 m/min are used. Dry friction, as well as friction lubricated with a pure ester oil with a viscosity of 10.2 mm2/s are used. The coefficient of friction is evaluated at steady state, i.e. after 5 s. Every combination of parameters is tested six times for each setup.

0.6

0.5

0.4 icient of friction [-] of friction icient � 0.3 in-process

coef pin-on-disc spiral pin-on-disc 0.2 20 60 100 velocity [m/min]

Figure 6.12: Coeﬃcient of friction compared for the three basic setups with dry friction. One standard deviation of the sample is shown as error bar.

The results for dry friction are summarized in Figure 6.12. The coeﬃcient of friction in the in-process setup is high, which can be traced back to the adhesive nature of unoxidized 104 6. In-process Tribometer

titanium. In the spiral pin-on-disc setup, the coefficient of friction is much lower, because the titanium oxide layer prevents the titanium from adhering to the pin surface. Whereas in all other tests, a strongly adhering titanium layer is observed. The pin appears slightly polished but otherwise unchanged after the spiral pin-on-disc test. The pin-on-disc setup reaches similar friction values as the in-process setup, despite the fully oxidized surface. However, the pin-on-disc setup is unstable. Tiny surface irregularities are amplified over the dozens of passes on the same surface. At some point, the surface is irregular enough to cause the pin to intermittently leave the surface. The maximum load exceeds the nominal load, the pin is able to penetrate the oxide layer, and a built-up layer is formed. The built-up layer is eventually sheared off from the pin and sticks to the workpiece. In the subsequent passes, the built-up layer is sheared off at this exact location, leading to a periodic change in the coefficient of friction as can be seen in Figure 6.13b. The coefficient of friction is similar to the one measured in-process at the beginning, although with the variation being much bigger due to the vibrations. In the course of the experiment, vibrations increase and lead to a lower coefficient of friction due to non-linear effects.

a) 0.8 b) 0.8

0.7 0.7

0.6 0.6

0.5 0.5 icient of friction [-] of friction icient [-] of friction icient � � 0.4 0.4 coef coef

0.3 0.3

0.2 0.2 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 time [s] time [s]

Figure 6.13: Unﬁltered dry coeﬃcient of friction at 100 m/min and 100 N. a) in-process measurement b) pin-on-disc setup.

The results for lubricated friction are summarized in Figure 6.14. The coefficient of friction in the in-process setup is smaller than in the same setup without lubrication. This proves the ability of esters oils to at least partially prevent adhesion. With higher speeds, the friction rises, which can be traced down to a partial film breakdown. The in-process setup shows higher variations in the lubricated situation, probably due to hydrodynamic effects in correlation with the irregularly shaped built-up layer. The oil does not have a significant influence in the spiral pin on disc setup, as no adhesion occurs. The coefficient of friction is slightly lowered at 100 m/min, probably due to hydrodynamic effects. At lower coefficients of friction, the resulting temperature in the interface is lower as well, thereby shifting film 6.7 Experimental Results 105

breakdown to higher speeds. The pin-on-disc setup is unstable again, making it impossible to evaluate a steady state.

0.6

0.5

0.4 icient of friction [-] of friction icient � 0.3 in-process

coef pin-on-disc spiral pin-on-disc 0.2 20 60 100 velocity [m/min]

Figure 6.14: Coeﬃcient of friction compared for the three basic setups lubricated with pure ester oil. One standard deviation of the sample is shown as error bar.

6.7.2 Inﬂuence of the Load

The standard pin is used to investigate the influence of the preload force on the coefficient of friction. A speed of 60 m/min is chosen, as this relative speed is reasonable for the relative velocity between the chip and the cutting tool as well. Preload forces of 5 N, 20 N, 100 N, 250 N, and 400 N are selected to cover the whole operating range of the in- process tribometer. The friction force is averaged in the steady state. The experiment is performed three times. Figure 6.15 shows the results with a clear trend to a lower coefficient of friction at higher load. In other words, the friction force shows a degressive behavior with increasing preload force. Furthermore, a slight trend to a lower variation at higher preload forces is observable. The higher the preload force, the bigger the contact zone gets, thereby the micro-geometry of the pin has less influence. The experiment is repeated with a pure ester oil at preload forces of 20 N, 100 N, and 400 N. The results are shown in 6.15 as well. The behavior of the coefficient of friction is similar, with the biggest difference from the dry friction observed at a medium load of 100 N. This represents the optimal range for the friction-modifying abilities of the ester oil. 106 6. In-process Tribometer

0.65

0.6 dry ester oil 0.55

0.5

0.45 icient of friction [-] of friction icient

� 0.4 coef 0.35

0.3

0.25 0 50 100 150 200 250 300 350 400 normal force [N]

Figure 6.15: Coeﬃcient of friction at diﬀerent preload forces using the standard pin at 60 m/min.

6.7.3 Transient Behavior

The above approach using steady-state friction is not suitable for the quick assessment of an MWF, especially if the exact velocity at which ﬁlm breakdown occurs needs to be determined. A whole velocity range is therefore tested with a single pin and with short holding times at the speciﬁed velocities. To validate the approach, results from a test with pure ester oil going up from 10 m/min to 100 m/min in steps of 10 m/min and going down from 100 m/min to 10 m/min are compared. The holding time at each velocity is 3 s. The results are shown in Figure 6.16. Both strategies result in a similar behavior with a minimum in friction at 40 m/min. The transition to the higher friction zone at speeds above 40 m/min is less stable in the sweeping up direction, leading to larger standard deviations in this region, than with the sweeping down direction. The sweeping down strategy is therefore chosen for subsequent tests.

6.7.4 Inﬂuence of the MWF Composition

As the in-process tribometer is expected to be used in preliminary testing of MWFs, its ability to differentiate between different products is crucial. To test this ability, four different neat oils are prepared. Each of the four oils features a viscosity of 10.2 mm2/s at 40 ◦C which is achieved by replacing a small portion of the base oil in the fully formulated 6.7 Experimental Results 107

0.6

0.5

0.4 icient of friction [-] of friction icient � 0.3 coef

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.16: Transient coefficient of friction when sweeping velocity up (blue, six tests) and down (orange, five tests) at 100 N preload. The standard deviation of the sample is indicated with error bars. neat oils with a base oil of lower viscosity of the same type to compensate for the increase in viscosity through the additives. The first oil is a pure ester oil. The second oil is a fully formulated metal working fluid based on the first oil, which is known for its high performance in titanium cutting. In addition to the base oil, it contains AW- and EP- additives. The third oil is a pure mineral oil. The fourth oil is a fully formulated MWF based on the third oil. It contains FM-, AW-, and EP-additives and is a universal product with medium performance in titanium cutting. The oils are tested with standard pins and with sweeping the velocity from 100 m/min to 10 m/min. Normal forces of 50 N, 100 N, 200 N, and 300 N are used and each test is performed three or four times. The results are summarized in Figure 6.17. In light friction conditions, i.e. at a normal force of 50 N and velocities below 50 m/min, the fluids behave similarly. The energies are too low to cause activation of AW- or EP- additives. Ester base oils form a weak adsorption layer on metal surfaces, which causes a lower coefficient of friction. They work as friction modifiers themselves, without any FM-additives needed. The only oil without FM-abilities is therefore the pure mineral oil, which leads to a slightly higher coefficient of friction in this region. At the same load but higher speeds, the adsorption film is not stable anymore. In the fully formulated products, the action of AW- and EP-additives kicks in. AW- and EP- additives do not necessarily lower the coefficient of friction. In this case, however, with the strong adhesive action of the fresh titanium surface, any interlayer with a lower shear 108 6. In-process Tribometer

50 N 100 N 0.6 0.6

0.5 0.5

0.4 0.4 icient of friction [-] of friction icient [-] of friction icient � 0.3 � 0.3 coef coef

0.2 0.2 0 20 40 60 80 100 0 20 40 60 80 100 velocity [m/min] velocity [m/min] pure ester oil neat oil based on ester oil pure mineral oil 200 N neat oil based on mineral oil 300 N 0.6 0.6

0.5 0.5

0.4 0.4 icient of friction [-] of friction icient [-] of friction icient � 0.3 � 0.3 coef coef

0.2 0.2 0 20 40 60 80 100 0 20 40 60 80 100 velocity [m/min] velocity [m/min]

Figure 6.17: Comparison of four different oils used with the standard pin at different normal forces. The maximum and minimum of three to four tests are indicated with error bars. strength than titanium lowers the coefficient of friction. Therefore, the coefficient of the fully formulated neat oils stays low up to the maximum speed of 100 m/min, whereas the coefficient of friction for the two pure oils rises. In medium friction conditions at normal forces of 100 N and 200 N, the coefficient of friction rises even for the fully formulated products at high speeds. At those conditions, the film generated by EP-additives is not stable enough and adhesion increases again. The velocity where the coefficient of friction rises again is dependent on the oil. In severe friction conditions at 300 N, the absolute differences between the oils are smaller, which is partly explained by the lower absolute value of the coefficient of friction at higher loads. In addition, at high velocities, the film generated by the EP-additives is not stable anymore for any product at high velocities, leading the same coefficient of friction for every oil at 100 m/min. 6.7 Experimental Results 109

A qualitative model to explain the behavior of the coefficient of friction in greater detail is presented in Section 6.8.2. The best separation of the products is observed at 100 N and 100 m/min or at 200 N and 50 m/min, with the same ranking obtained at both points. The variations at 100 N and 100 m/min relative to the absolute values are smaller and the ranking is therefore more robust at this point. The method does not account for chemical deterioration of the surface, which may be the dominant cause of tool failure. It can therefore only be used in conjunction with the other methods described in Chapter 7. Water miscible MWFs are economically much more significant than neat oils. Therefore, the procedure for neat oils described above is adapted for them. The same range of MWFs, previously used in the turning tests shown in Table 4.6, is used in the tribometer tests. Deionized water is added as a negative reference. The fluids are tested with standard pins and with sweeping the velocity from 100 m/min to 10 m/min. The normal force is set to 100 N and each test is performed three to five times. The results are summarized in Figure 6.18.

0.6

0.5 reference emulsion (mineral based) deionized water 0.4 universal emulsion (mineral based) fully synthetic grinding luid icient of friction [-] of friction icient � 0.3 fully synthetic luid �

coef emulsion (ester� based)

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.18: Comparison of six diﬀerent water-miscible MWFs and deionized water used with the standard pin at 100 N. The maximum and minimum of three to ﬁve tests are indicated with error bars.

Deionized water leads to a significantly higher coefficient of friction than the other tested fluids. The other tested fluids are all similar, with no clear distinction between them. Even the synthetic grinding fluid, with no tribological additives, appears in the middle of the other fluids. Water-miscible fluids have a significantly better cooling action than 110 6. In-process Tribometer

neat oils, thereby leading to lower temperatures in the contact zone. The temperatures are presumably too low to form a stable reaction layer and an adsorption layer is formed instead. The observable difference between the deionized water and MWFs can also be explained by differences in the physical parameters, such as the viscosity. Several options to increase the provided energy for the formation of a reaction layer exist. The normal force or the velocity can be increased to increase energy input, whereas the concentration of the MWF can be increased to enhance the lubricity while at the same time reducing the cooling action. A slightly better separation of the products is achieved at 400 N, however the ranking does not correlate with the tool life associated with the MWFs in either turning or milling. The in-process tribometer is therefore not suitable to estimate the performance of a water-miscible fluid.

6.7.5 Inﬂuence of the Pin

The in-process tribometer setup is designed to emulate the conditions in cutting as closely as possible. The standard pin specifications are chosen based on the experience with conventional pin-on-disc tests, but they are still arbitrarily chosen in many aspects. To assess the sensitivity of the setup to deviations from the standard pin specifications, the following tests are performed. The pin surface is ground to a surface roughness Ra of 0.4 µm. Polishing the pin to a surface roughness Ra of smaller than 0.1 µm increases the cost by five times and is therefore avoided if possible. Three pins are polished and the steady-state coefficient of friction is measured for each of them at 20, 60, and 100 m/min at dry conditions. The test is repeated with another three polished pins at 400 N. The results are shown in Figure 6.19. Unpolished and uncoated pins are used to obtain a reference value. No significant difference is observed between the unpolished and the polished pin, therefore justifying the decision to use the ground pin directly for friction measurements. The standard pin is coated with the AlTiN-based coating Balinit Latuma by Oerlikon Balzers, Liechtenstein. AlCrN-based coatings present another reasonable coating technology with a supposedly lower affinity to titanium. Six pins are coated with Alcrona by Oerlikon Balzers and tested with the same procedure described above. As a reference, six standard pins, coated with Balinit Latuma, are tested with the same procedure as well. The results are shown in the Figure 6.19. No significant difference in the coefficient of friction is observed between uncoated, AlTiN-coated, and AlCrN-coated pins. The missing influence of the pin surface roughness and coating leads to the hypothesis that the tribological contact actually consists of two titanium surfaces, the titanium workpiece and the titanium built-up layer. The titanium built-up layer is formed immediately under 6.7 Experimental Results 111

0.6

100 N 0.5 uncoated

AlTiN-coated 0.4 uncoated, polished

icient of friction [-] of friction icient AlCrN-coated � 0.3

coef 400 N

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.19: Influence of the pin surface on the steady-state coefficient of friction in dry conditions. The maximum and minimum of three tests are indicated with error bars. all in-process test conditions, rendering the actual pin surface irrelevant with respect to the coefficient of friction. To test this hypothesis, a titanium pin with the same geometry as the standard pins is turned from Ti6Al4V. To limit pin wear in the beginning of the test, the cutting speed is swept up from 10 m/min to 100 m/min at a preload force of 100 N. A pure ester oil is used to lubricate the contact. The resulting coefficient of friction is compared with the data for coated cemented carbide pins from Section 6.7.3 in Figure 6.20. It is similar to the one obtained with the standard pins, indicating similar adhesive conditions in both cases. At lower speeds, the coefficient of friction with the titanium pin is lower, as the formation of a strain-hardened layer on top of the pin is not possible and the pin material is sheared off instead. At higher speeds, the coefficient of friction of the titanium pin is higher. The pin is substantially worn at this stage and eventually the whole cross section is in contact. A bigger contact area causes a lower contact pressure. The coefficient of friction increases at lower loads as is shown in Section 6.7.2.

Under the Hertzian contact assumptions, the maximum pressure pmax scales with the force F and the pin radius R:

F 1/3 p ∼ (6.6) max R2 The spatial pressure distribution is elliptical and therefore all pressure distributions are

geometrically similar and fully deﬁned by pmax and the contact zone radius a. The normal 112 6. In-process Tribometer

0.6

titanium pin cemented carbide pin

0.5

0.4 icient of friction [-] of friction icient � 0.3 coef

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.20: Comparison between friction with a titanium pin and standard coated cemented carbide pins on a titanium workpiece. force is deﬁned as the integral of the local pressure over the contact area. Therefore, the normal force F is described as follows:

2 F ∼ pmax · a (6.7)

Combining Equations (6.6) and (6.7) leads to the following relation for the radius of the contact zone:

a ∼ (F · R)1/3 (6.8)

2 Assuming that F/R is kept constant to keep pmax constant, the contact area radius a scales linearly with the pin radius R. However, nonlinear effects, such as plastic material deformation, friction, and lubrication effects, cause deviations from this behavior. To quantify the deviations, pins with different spherical tip radii are compared: A pin type with 1.5 mm radius, which corresponds to a hemisphere and the 3 mm radius standard pin. Both pin types are coated with the AlTiN-based coating. The normal force is adjusted to reach the same maximum Hertzian pressure pmax in the compared contacts. The pins with a radius of 1.5 mm are therefore 6.7 Experimental Results 113

tested at 25 N and the standard pin with a radius of 3 mm at 100 N. Three velocity sweeps from 100 m/min to 10 m/min are performed, each one with a new pin. The results are shown in Figure 6.21. 0.6

0.5

0.4

r = 1.5 mm, 25 N icient of friction [-] of friction icient � 0.3 r = 3 mm, 100N coef

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.21: Inﬂuence of the pin radius on the coeﬃcient of friction with dry friction. The experiments are performed at equal Hertzian pressures. The maximum and minimum of three tests are indicated with error bars.

Similar experiments are performed with lubrication. Pure ester oil with a viscosity of 10.2 mm2/s is used. Medium friction is studied using the 1.5 mm pins at 25 N and the 3 mm pins at 100 N, again ensuring that the maximum Hertzian pressure is equal. The results are shown in Figure 6.22. Severe friction is tested with the 1.5 mm pins at 100 N and the 3 mm pins at 100 N, as shown is Figure 6.23. The dry results show a good conformity between measurements with the same maximum pressure, indicating that the pin radius can be adjusted to reach a desired contact stress while staying in the optimal range of the normal force between 30 N and 300 N. In the case of lubricated friction, this procedure is not possible, as the coefficient of friction shows a different behavior at the same maximum pressure but different pin radius. For example, the typical film breakdown observed at severe friction conditions with a 3 mm pin at about 40 m/min could not be observed with the 1.5 mm pin as can be derived from Figure 6.23. Therefore, both the preload force and the pin radius have to be carefully chosen to observe a certain phenomenon. Luckily, the usable range of optimal preload forces is extended up to 400 N because of the smaller susceptibility to tribometer vibration in lubricated contacts. 114 6. In-process Tribometer

0.6

r = 1.5 mm, 25 N 0.5 r = 3 mm, 100N

0.4 icient of friction [-] of friction icient � 0.3 coef

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.22: Inﬂuence of the pin radius on the coeﬃcient of friction, lubricated with pure ester oil, at medium conditions. The experiments are performed at equal Hertzian pressures. The maximum and minimum of three tests are indicated with error bars.

0.6

r = 1.5 mm, 100 N 0.5 r = 3 mm, 400N

0.4 icient of friction [-] of friction icient � 0.3 coef

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.23: Inﬂuence of the pin radius on the coeﬃcient of friction, lubricated with pure ester oil, at severe conditions. The experiments are performed at equal Hertzian pressures. The maximum and minimum of three tests are indicated with error bars. 6.7 Experimental Results 115

A pin with a cylindrical joint is prepared as described in Section 6.6. The frictional contact consists of a cylinder with a radius of 3 mm and a flat surface, forming a line contact. The coefficient of friction is measured by sweeping down the velocity from 100 m/min to 10 m/min at a normal force of 100 N and 200 N and using an oil with a viscosity of 10.2 mm2/s. The test is performed three times with the same pin. The results are shown in Figure 6.24. The contact area increases significantly compared to the standard pin, thereby reducing the contact pressure at a given preload force. Assuming Hertzian contact conditions, the contact pressure in the line contact is 930 N/mm2 at 100 N and 1310 N/mm2 at 200 N. The Hertzian pressure for the point contact with the standard pin is 2950 N/mm2 at 100 N and 3720 N/mm2 at 200 N. The calculated pressure in the point contact is about three times larger than in the line contact, but the assumption of purely elastic deformation is clearly not met in the point contact, leading to a smaller difference in reality. The lower contact pressure leads to a higher coefficient of friction in the line contact and an absence of the typical film breakdown at 40 m/min and 200 N with the standard pin. However, contrary to the behavior in dry contacts, the coefficient of friction at 200 N is higher than at 100 N, indicating a higher portion of fluid friction.

0.8

standard pin 100 N

0.7 line contact 100 N

standard pin 200 N

0.6 line contact 200 N

0.5

0.4 icient of friction [-] of friction icient �

coef 0.3

0.2 0 20 40 60 80 100 velocity [m/min]

Figure 6.24: Coeﬃcient of friction with line contact in comparison to the standard pin. The contact zone is lubricated with a pure ester oil. The maximum and minimum of three tests are indicated with error bars. 116 6. In-process Tribometer

6.7.6 Temperature in the Friction Zone

The temperature in the friction zone is of great interest, as it influences the mechanical and wear properties of the tool materials as well as the rate of chemical reactions on the surface. Measuring the temperature distribution near the contact zone poses major difficulties, as temperature gradients in the order of 1000 K/mm are expected. The temperature is measured with a fiber pyrometer in an axial blind hole in the middle of the pin with a remaining wall thickness of 200 µm, similar to the measurements in the cutting edge in Section 4.1.2. This allows dry and lubricated measurements with the same setup. The high aspect ratio of the hole for the fiber causes higher variations in the remaining wall thickness and therefore in the measured temperature than in the turning inserts prepared with the same method. A number of different tests are performed at normal forces from 50 N to 500 N and with velocities ranging from 40 m/min to 140 m/min. The measured temperatures are shown as a function of the frictional power, which is calculated by multiplying the measured tangential force with the velocity, in Figure 6.25. It can be seen that most of the temperature change can be explained by the frictional power. The temperature is therefore mainly governed by heat conduction in the pin and radiation. If forced convection through the workpiece rotation or heat transport with the workpiece rotation had a major influence, lower temperatures would be expected at higher velocities but at the same frictional power. From the same measurement data, the coefficient of friction can be plotted against the temperature, as is shown in Figure 6.26. A slight negative correlation between the coefficient of friction and the temperature is visible; it may however be an effect of the higher loads, which had to be used in order to reach higher temperatures. Other means to raise the temperature independently of the load and velocity, such as preheating the pin or workpiece were not successful. It is therefore not possible to separate the effects. At high loads and high speeds, the pin wears quickly. It develops a wear land of undefined geometry, which leads to non-reproducible results. The wear land leads to a lower contact pressure, which in turn increases the coefficient of friction. A higher coefficient of friction leads to higher tangential forces, more frictional power, and therefore a higher temperature. As soon as the wear depth reaches 0.2 mm, semisolid titanium is squeezed into the now open hole. It destroys the fiber or pushes it away from the surface, rendering further measurements impossible. Shortly before the break-through occurs however, the best measurements of the real interface temperature are possible. In one case, at a preload force of 463 N and a velocity of 140 m/min, sustained temperatures of 1105 ◦C were measured shortly before the wear reached the hole at a depth of 0.18 mm. 6.7 Experimental Results 117

1000 900 800 700 v = 40 m/min 600 v = 60 m/min 500 v = 80 m/min 400 v = 100 m/min temperature [°C] temperature 300 v = 140 m/min 200 100 0 0 100 200 300 400 500 600 700 friction power [W]

Figure 6.25: Measured temperature at a depth of 200 µm in the pin of the in-process tribometer as a function of the friction power at diﬀerent velocities v without lubrication.

0.8

0.7

0.6 v = 40 m/min 0.5 v = 60 m/min 0.4 v = 80 m/min

icient of friction [-] of friction icient 0.3 v = 100 m/min �

coef v = 140 m/min 0.2

0.1

0 0 200 400 600 800 1000 temperature [°C]

Figure 6.26: Coeﬃcient of friction as a function of the temperature measured at a depth of 200 µm in the pin of the in-process tribometer at diﬀerent velocities v without lubrication.

Similar experiments are performed with a lubricated contact as well. The situation in lubricated contacts is more complicated, as the MWF does not only lubricate the contact but does lead to a stronger convective cooling as well. A pure ester oil with a viscosity of 10.2 mm2/s is used. The measured temperature as a function of the frictional power is shown in Figure 6.27. It is within the uncertainty range from the dry results, meaning that no signiﬁcant cooling eﬀect of the oil could be detected with this method. The pin 118 6. In-process Tribometer

is lubricated only very locally in the contact zone. A large part of the heat flows into the cool pin holder through the large clamping area. This cooling effect is not affected by lubrication. The influence of the temperature on the coefficient of friction is shown in Figure 6.28. A negative trend of the coefficient of friction with increasing temperature is only observed within the sub-sets at the same speed, where the effect is more likely caused by an influence of the load on both the temperature and the coefficient of friction.

1000 900 800 700 v = 40 m/min 600 v = 60 m/min 500 v = 80 m/min 400 v = 100 m/min temperature [°C] temperature 300 v = 140 m/min 200 100 0 0 100 200 300 400 500 600 700 friction power [W]

Figure 6.27: Measured temperature at a depth of 200 µm in the pin of the in-process tribometer as a function of the friction power at diﬀerent velocities v with ester oil.

0.8

0.7

0.6 v = 40 m/min 0.5 v = 60 m/min 0.4 v = 80 m/min

icient of friction [-] of friction icient 0.3 v = 100 m/min �

coef v = 140 m/min 0.2

0.1

0 0 200 400 600 800 1000 temperature [°C]

Figure 6.28: Coeﬃcient of friction as a function of the temperature measured at a depth of 200 µm in the pin of the in-process tribometer at diﬀerent velocities v with ester oil. 6.8 Friction Models 119

6.8 Friction Models

The Coulomb-model is the simplest friction model and is widely used. It assumes a linear dependency of the normal force on the friction force. The influence of the velocity is not covered. Due to its linear nature, superposition of different friction zones is easy. The total friction force can be calculated from the total load by simply multiplying with the coefficient of friction without knowing the local load distribution. The results shown in Section 6.7 clearly deviate from the Coulomb-model. Two approaches, one analytical for dry friction and one empiric for lubricated friction are developed to address the differences.

6.8.1 Analytical Dry Friction Model

The in-process tribometer delivers friction data integrated over the whole contact area of the pin. This information is useful to compare different situations; however, the absolute value has no significance, as it originates from a variety of different local conditions. There- fore, the value depends on the pin shape, normal force, and velocity. For the numeric simulation of cutting processes, the local coefficient of friction needs to be determined, given as a function of contact pressure instead of normal force and calculated as the quotient of shear stress and normal stress rather than the quotient of tangential force and normal force. To allow for a simple transformation between the experimental data and the local coefficient of friction, a number of simplifications are made. The influence of the velocity is not considered. However, the process could be repeated at multiple velocities to cover this dependency. Hertzian pressure distribution is assumed, neglecting plasticity effects normal to the surface that cannot be treated analytically. The shear strength of titanium is assumed to be independent of the hydrostatic pressure. Furthermore, the influence of the temperature is neglected. The Hertzian contact theory [62] provides the contact radius a for a sphere and a half plane in dependence of the spherical radius R, the combined equivalent elastic modulus E∗, which is calculated from Poisson’s ratio ν and elastic modulus E of both materials, and the normal load F :

− 1 − ν2 1 − ν2 1 E∗ = 1 + 2 (6.9) E E 1 2

3 3FR a = ∗ (6.10) r 4E 120 6. In-process Tribometer

The maximum pressure is calculated:

3F p = (6.11) max 2πa2

The pressure distribution is elliptically decreasing towards the edge of the contact zone:

2 − r p(r) = pmax 1 2 (6.12) r a

Coulomb-Orowan friction [109] is used as the simplest model with saturation of the tangential force:

µ · p p < τmax/µ τ = (6.13) τ otherwise  max  Equation (6.13) features the two unknown parameters µ and τmax. The Hertzian pressure distribution is used as the pressure input in the Coulomb-Orowan model. At low loads, the critical pressure τmax/µ is not reached, therefore no saturation occurs. The radius of the saturated region, rsat is therefore zero. At loads above a critical Force Fcrit, a region concentric to the contact zone but smaller with saturated shear stress establishes:

0 pmax < τmax/µ rsat = 2 (6.14) τmax a · 1 − 2 2 otherwise  µ pmax q  Both situations are shown in Figure 6.29. Using Equations (6.10) and (6.11), Equation (6.14) is solved for F :

3 2 3 π R

p, p,

τ τpmax

pmax µ pmax

max max

µτ pmax τ

r r -rsat rsat

Figure 6.29: Sketch of the stresses in the two regimes. Left: unsaturated shear stress at low loads. Right: saturated shear stress at higher loads.

contact radius a radius

saturation radius rsat

F crit load

Figure 6.30: Size of the saturated zone approaching the size of the contact zone with increasing load.

The total tangential force Ftan is obtained by integration of τ over the whole contact area. The result is shown in Equation (6.16) and in Figure 6.31. Finally, the coeﬃcient of friction is obtained by a division by the load F , as shown in Figure 6.32.

µF F < Fcrit Ftan = (6.16) 2 a r πτmax + 2π µ · p(r) · rdr otherwise  sat rsat R The model is validated using the data presented in Section 6.7.2. The cemented carbide is assumed to have a Young’s Modulus of 600 GPa and a Poisson ratio of 0.31. On the other hand, titanium is assumed to have a Young’s modulus of 115 GPa and a Poisson ratio of 0.35, resulting in an equivalent modulus E∗ of 217 GPa. The pin’s radius is 3 mm. The two unknown parameters of the Coulomb-Orowan-model are determined by ﬁtting 122 6. In-process Tribometer

Coulomb model

Coulomb- Orowan model tan F

F crit load

Figure 6.31: Total tangential force as a function of the normal load. The linear Coulomb model is added as a reference.

Coulomb model μ

Coulomb- Orowan model icient of of friction icient � coef

Fcrit load

Figure 6.32: Coeﬃcient of friction as a function of the load. The linear Coulomb model is added as a reference.

the modeled coeﬃcient of friction to the measurement data. The resulting local coeﬃcient

of friction µ is 0.57 and the maximum shear stress τmax is 1800 MPa. Figure 6.33 shows the ability of the model to describe the decreasing coeﬃcient of friction with increasing

load. Although in the same order of magnitude, τmax is higher than expected from the tensile strength of the titanium (900 MPa to 1200 MPa). The contact area estimated by

the elastic model is smaller than in reality. Therefore, τmax needs to be chosen bigger to match the measured tangential forces. The stress state in the titanium near the surface may lead to an increase in its shear strength. Furthermore, the surface is work-hardened by the cutting process to some extent, increasing its shear strength. 6.8 Friction Models 123

0.6

0.5

0.4

0.3 icient of friction [-] of friction icient

� 0.2

coef 0.1

100 200 300 400 load [N]

Figure 6.33: Comparison between measured data (blue) and the model (red).

6.8.2 Lubricated Friction Model

The analysis of the friction data obtained with different neat oils leads to a simple empirical model. Its main aim is to simplify the evaluation of the coefficients of friction curves and compare the results of different MWFs. It is based on three different friction states: a lubricated state, a transient state, and a quasi-dry friction state. The model is only valid for boundary friction conditions, which are observed in the whole parameter range of the tribometer. In the lubricated state, a film is developed. Due to the longer exposure time, it is expected to form mainly on the surface of the pin but formation on the titanium surface is possible as well. The film supports part of the load and possibly lowers the coefficient of friction with a lower shear strength than the titanium. The film is not to be confused with the fluid film, which establishes under hydrodynamic lubrication conditions and shows a completely different behavior. The properties of the film vary with the provided energy and with the rate at which the film is sheared off; therefore, they vary with the load and velocity. Furthermore, the coefficient of friction is measured in a transient state after only a short holding time at each velocity, leading to deviations from the steady-state coefficient of friction. Pure mineral oils are non-polar and have a minimal tendency to adhere to metal surfaces. They present a baseline for a weak film formation while still being able to physically lubricate. Figure 6.34 shows the expected coefficient of friction in the lubricated state for a pure mineral oil and other MWFs. The lubricated state can therefore be used as an indicator for the friction modifying ability of an MWF. The dependence on the load is 124 6. In-process Tribometer

small and is not included in the diagram for better readability. The shape of the curves to the far right is only theoretically estimated. No MWF is able to operate in the lubricated state at the highest speeds. However, the shape of the lubricated state curves in this region does not inﬂuence the predictions.

lubricated state

run-in phase mineral oil

baseline experiment start } icient of of friction icient � increasing coef friction-modifying properties theoretical progression

velocity

Figure 6.34: Sketch showing the hypothetical coeﬃcient of friction curves in the pure lubricated state.

A mathematical representation of the curves for µlub shown in Figure 6.34 is determined with manual ﬁtting. The base functions are chosen so that no negative values for the coeﬃcient of friction result in the extrapolation region for higher loads. Only two free

parameters, the speed vc and the normal force FN , were investigated and are considered as variables. The resulting equation is therefore only valid for this exact experimental setup with the same in-process tribometer, pin material and shape, workpiece material, ﬂuid viscosity, and a stepwise decrease in velocity:

0.55 µlub(vc, FN ) = − ∆µFM (6.17) − − FN 0.28 1 exp 200 N vc 1 m/min

Equation (6.17) features the FM-abilities ∆µFM , i.e. the reduction in the coeﬃcient of friction through additives compared to a mineral oil. 6.8 Friction Models 125

If the load or the velocity are high enough, the film cannot establish itself. The metal is directly rubbing on its metal counterpart, leading to quasi-dry friction. In this state, the composition of the MWF does not have any influence on the coefficient of friction anymore. Instead, the coefficient of friction is only determined by the load and the velocity. Figure 6.35 shows the expected coefficient of friction in quasi-dry friction for different loads. At low loads, the coefficient of friction is increasing with higher velocity because of stronger adhesion and less oxidation of the titanium surface. At high loads, the coefficient of friction is decreasing because of thermal softening of the titanium. Results from the quasi-dry state cannot be used for the development of MWFs, as the MWF’s composition has no influence. quasi-dry state run-in phase }

increasing experiment start

icient of of friction icient load �

theoretical progression coef

velocity

Figure 6.35: Sketch showing the hypothetical coeﬃcient of friction curves in the quasi-dry state.

In the same manner as with Equation (6.17), a mathematical representation of the curves

for µdry shown in Figure 6.35 is derived:

F v µ (v , F ) = 0.4 + 0.007 exp − N − 0.2 · c (6.18) dry c N 125 N 1 m/min Equations (6.17) and (6.18) are plotted in the left side of Figure 6.39.

Between the two extreme states of µlub and µdry, a transition region exists. In the transition state, also called film breakdown state, the film is partially formed and reduces friction but is not able to fully establish. The coefficient of friction takes a value in between the 126 6. In-process Tribometer

lubricated state and the quasi-dry state. Figure 6.36 shows the behavior of the coefficient of friction in the transition state. Two example cases are shown. Case 1 shows a fluid with medium friction-modifying abilities and a low transition speed, which is used at a high load. In contrast, case 2 shows a fluid with optimal friction-modifying abilities and high transition speed, which is used at a moderate load. transition state

2 icient of of friction icient � 1 coef

velocity

Figure 6.36: Sketch showing the transition between the lubricated state at low velocities to the quasi-dry state at high velocities for two example cases (black).

Mathematically, the transient state is modeled with a sigmoid function, which gradually reduces the weight of the lubricated coefficient from 1 to 0 and at the same time increases the weight of the quasi-dry coefficient of friction from 0 to 1. The resulting combined coefficient of friction µcomb for the whole velocity range is:

1 v − v 1 v − v µ (v ) = · 1 − tanh c trans ·µ + · 1 + tanh c trans ·µ (6.19) comb c 2 v lub 2 v dry range range

Equation (6.19) features the transition velocity vtrans and the transition range vrange, which are properties of the ﬂuid. Using the qualitative curves, the results obtained in Section 6.7.4 can be analyzed in more detail. Figure 6.37 shows the friction modifying ability and transient state of three examples. To further validate the model, the velocity range of the measurement data with pure ester oil at 100 N is extended up to velocities of 160 m/min. Figure 6.38 shows the 6.8 Friction Models 127

result of the test, which extends the previously measured data on the quasi-dry curve as expected and conﬁrms the model.

0.6 0.6 0.6

0.5 0.5 0.5

0.4 0.4 0.4

0.3 0.3 0.3 icient of friction [-] of friction icient �

coef 0.2 0.2 0.2 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 velocity [m/min] velocity [m/min] velocity [m/min]

Figure 6.37: Three exemplary evaluations with the presented model. Measured data is indicated with dots. Left: Pure ester oil at 200 N. Middle: Neat oil based on ester oil at 100 N (the quasi- dry state is not reached). Right: Pure mineral oil at 300 N.

0.6

0.5

0.4 icient of friction [-] of friction icient � 0.3 coef

0.2 0 20 40 60 80 100 120 140 160 velocity [m/min]

Figure 6.38: Measured coeﬃcient of friction as function of the velocity in the range from 160 m/min to 100 m/min (blue) lubricated with pure ester oil of a single test. The average of previous measurements in the range from 100 m/min to 10 m/min is shown in gray.

The mathematical representation can be used to provide values for ∆µFM , vtrans and

vrange of a ﬂuid automatically. In the case of the pure ester oil at 100 N, which is shown in the right side of Figure 6.39, the following values are determined by minimizing the

least-squares error: ∆µFM = 0.037, vtrans = 43 m/min, and vrange = 8.8 m/min. The model represents a useful way to qualitatively and quantitatively compare MWFs. It helps to get an overview of the high number of data points, which are obtained in the experiments. Each curve can be described with just three parameters. If data from more than one load value is available, the inﬂuence of the load on the transition speed or the transition range can be discussed as well. 128 6. In-process Tribometer

0.8 FN = 0.45 50 N 0.7 0.4 icient � 0.6 100 N coef of friction [-] friction of 0.5 0.35 FM = 0 0.03 200 N Δ 0.06 0.3

0.09 [-] of friction icient � 0.3 400 N

coef 0.25 0.2

0.1 0.2 10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100 velocity [m/min] velocity [m/min]

Figure 6.39: Left: Families of curves resulting from the evaluation of µlub (black - blue) at different values of ∆µFM and µdry (red) at different normal loads FN . For readability, the influence of FN on µlub is omitted. Right: Empirical model (black) fitted to the measured data (green). µlub is shown in blue, µdry in red.

6.9 Conclusion

The in-process tribometer proves to be a valuable tool for the investigation of tribo-systems involving titanium. The in-process principle is the only way to observe nascent titanium surfaces in an industrial environment. The current design of the in-process tribometer has a high stiffness, is easy to use, uses small amounts of workpiece material, and can easily be adapted for new measurement tasks. With the help of the in-process tribometer, anti-adhesive abilities of different neat oils are compared. The friction heat allows reaching temperatures close to those observed in cutting and the force range allows to reach severe plastic deformation if necessary. Differences in the coefficient of friction between water-miscible MWFs tend to be smaller than those between neat oils because the activation temperature for surface-active additives is not reached with water-miscible MWFs. The coefficient of friction deviates from the constant Coulomb model in many ways. For dry friction, the coefficient of friction approaches zero asymptotically with increasing load. A simple local shear stress saturation model combined with the Hertzian contact model can explain the behavior in this case. In lubricated friction, a transition from a well lubricated to a quasi-dry friction state is observed above a certain speed, leading to sigmoid coefficient of friction curves as a function of speed. 129

Chapter 7

Wear Mechanisms

7.1 Analysis of Cutting Tools

7.1.1 Experimental Setup

During cutting, only indirect effects of tool wear are observable. To assess the worn geometry of a tool, however, the process is interrupted. The tool is removed from the machine tool and a picture is taken with an optical microscope. The illumination is adjusted so that the wear mark is clearly distinguishable from the unworn tool surface and a picture of the flank face is taken. If necessary, a picture of the rake face is taken as well. This process is recommended by both ISO 3685 and ISO 8688 for cutting tool life tests. During optical wear measurement, the tool is not in the same state as during the cutting process. After a defined metal removal volume or after reaching a tool life criterion, the process is interrupted by leaving the workpiece, leading to a decreasing chip thickness and a different adhesive behavior. Built-up material is therefore not representative for the material that is adhering to the tool under nominal cutting conditions. The geometry of the worn cutting tool can be measured non-destructively with a 3D- microscope. In the case of titanium cutting, it is however often impossible to see the cemented carbide cutting edge due to built-up titanium. Hartung et al. [59] use concentrated hydrofluoric acid to remove the titanium. This leads to problems when an inhomogeneous layer is to be removed, as the cobalt binder is edged by hydrofluoric acid as well. Furthermore, hydrofluoric acid is highly toxic and can only be handled in special laboratories. Alternative etchants are identified by evaluating available corrosion data for titanium [58]. The amount of unwanted cobalt-leaching is quantified by measuring the concentration of cobalt ions in the etchant after etching by X-ray fluorescence (XRF) spectroscopy. Phosphoric acid proves to etch titanium quickly, but leaches cobalt, leading 130 7. Wear Mechanisms

to a very brittle tungsten carbide skeleton, which often cracks spontaneously. Trying to reduce the cobalt leaching rate by a priori increasing the cobalt ion content in the etchant leads to no improvement. The oxidative effect of the etchant is increased by the addition of hydrogen peroxide in an attempt to passivate the cobalt surface. Then, the reducing effect of the etchant is increased by using phosphorous acid in an attempt to prevent the formation of a passivation layer on titanium. Both attempts did not result in a reduction of the cobalt-leaching rate in relation to the titanium etching rate. A very different candidate, aluminum chloride, does not behave as documented in [58] for unalloyed titanium and does not lead to Ti6Al4V removal. Oxalic acid slowly etches titanium but does not react with cobalt at a significant rate. Oxalic acid tends to leave insoluble titanium oxalate residues on the surface of the tool, slowing down the etching process and making 3D-measurements impossible. These residues can be partially avoided by addition of a small amount of ethylenediaminetetraacetic acid (EDTA). EDTA forms stronger complexes with titanium than oxalic acid and therefore leads to a conversion of titanium oxalate to pure oxalate, which is soluble in water. The titanium-EDTA complex formed in this process is soluble in water as well. The result of the etching process is shown in Figure 7.1.

Figure 7.1: View of the same indexable insert after cutting titanium (left) and after etching with oxalic acid (right).

The coating of the tool is slightly more resistant to etching than the adhering titanium. However, ﬂaking of the thin decorative golden T iN top-layer cannot be prevented. A change in the properties of the tool bulk material cannot be ruled out; therefore, etched tools are never re-used in cutting tests, even if their end of life is not yet reached. Another approach is to cross-section the tool at the desired position with wire EDM or with a diamond grinding disc, remove the damaged layer from the cross-sectioning operation with gentle grinding, and polish the surface with abrasive polishing or with a broad ion 7.1 Analysis of Cutting Tools 131

beam (BIB). BIB is the preferred method when brittle surface layers are expected and good edge retention is necessary. Depending on the structure to be analyzed, light microscopy, 3D-microscopy, or scanning electron microscopy are used. Although being the most expensive technique, SEM allows for the resolution of single tungsten carbide grains, provides a good contrast between diﬀerent phases, and can provide information about the elemental composition via energy- dispersive X-ray spectroscopy (EDX).

7.1.2 Results

Figure 7.2 shows the comparison between a cutting edge used with an emulsion and one used with an oil. Although the diﬀerences are small after 4 min, the cutting edge used with a neat oil shows more pronounced wear with a similar pattern. This is in accordance with the tool life results shown in Section 4.9.

0.5 mm

Figure 7.2: Cutting edges used for 4 min with the reference parameter set. The titanium built- up layer is completely etched to reveal the cemented carbide. Left: High performance emulsion. Right: High performance neat oil.

Figure 7.3 shows a comparison between a high performance and a standard oil. The crater wear and cutting edge micro-geometry is similar for both oils, however, the standard oil leads to strong notch wear. The notch wear protrudes far into the crater wear zone. A second notch-type wear zone, not connected to the primary notch zone in the middle of the cutting edge, is revealed after etching the adhering titanium. To study the progression of notch wear, the process is stopped after diﬀerent cutting times. As the cutting edge is not useable after etching anymore, a new cutting edge is used for every step, therefore the wear progression seen in Figure 7.4 is not strictly monotonic. It can be seen that the 132 7. Wear Mechanisms

notch wear starts at the cutting edge and progresses further along the edge of the contact zone.

1mm

Figure 7.3: Cutting edges used for 2 min with the reference parameter set. The titanium built- up layer is completely etched to reveal the cemented carbide. Left: High performance neat oil. Right: Standard neat oil.

3 s 41 s 83 s 166 s 290 s

Figure 7.4: Cutting tests with a standard neat oil, interrupted after the indicated time. Ad- hering titanium is completely etched. Top row: View of the rake face. Bottom row: Isometric view.

The results further support the hypothesis that no MWF is reaching the crater wear or the cutting edge in a continuous cut. In contrast, the notch wear is heavily inﬂuenced by the composition of the MWF. It is therefore advisable to develop new MWFs with a 7.2 High Temperature Oxidation Test 133

special emphasis on preventing notch wear. No conventional quick tests for notch wear are available; therefore, new test possibilities are investigated.

7.2 High Temperature Oxidation Test

According to the literature presented in Section 2.5, notch wear is caused mainly by tribo- oxidative eﬀects in the case of titanium cutting. A quick test is developed to estimate the notch wear behavior of a newly formulated MWF before producing large amounts and performing expensive cutting tests.

7.2.1 Experimental Setup

Titanium, cemented carbide, and MWF as well as the temperature and pressure conditions are identified as the main drivers for notch wear. The quick test should reproduce them as closely as possible in a simple setup. A titanium cup is prepared with a small volume at the bottom, a chamfered section to act as a seal and a fine thread to receive the plug. A cross-section of the setup is shown on the left side of Figure 7.5. The volume at the bottom is filled with 60 µl of the fluid to be tested. A small disc with 6 mm diameter and a thickness of 2 mm is polished and placed upside down into the fluid in the cup. The cup is then closed with a plug. The plug presses the cemented carbide disc into the chamfered titanium section to create a hermetic seal. At the same time, the circumference of the cemented carbide disc is exposed to unoxidized titanium and a point, where titanium, cemented carbide and MWF meet, is created, which is shown in the right side of Figure 7.5). The whole setup is designed to be easy to manufacture, to allow the whole setup to be cross-sectioned or to be replaced if oxidized. The assembly is put in an annealing furnace at temperatures between 780 ◦C and 1000 ◦C, the temperatures which are expected to occur in cutting according to the results described in Section 4.9. The MWF evaporates quickly. The excess vapor lifts the cemented carbide disc, passes by the thread and burns in the furnace. The test duration may be chosen between 4 min, just long enough to heat up the carbide disc, and 1 h, as long as the normal tool life. After heating, the cemented carbide disc can be retrieved by removing the plug. The cemented carbide disc can be cross-sectioned or inspected directly. After re-cutting of the conical seal area in the container, it can be reused several times. 134 7. Wear Mechanisms

1 cemented carbide 3 titanium MWF

Figure 7.5: High Temperature Oxidation Test Setup. 1: Cup, 2: MWF, 3: Cemented Carbide Disc, 4: Plug. Right: Close-up of the circumference of the cemented carbide disc.

7.2.2 Results

Two diﬀerent neat oils are tested. One universal product and one optimized for titanium cutting. After opening the container, the space below the cemented carbide disc is usually ﬁlled with soot with both oils. Sometimes, the carbide disc is covered with a carbon-rich soot layer as well, as shown in Figure 7.6. The soot layer is easily removable with ethanol in an ultrasonic bath.

1 mm 1 mm

Figure 7.6: Soot layer on top of the cemented carbide disc directly after test.

The investigation of the surface with a light microscope reveals irregularities on the surface that was exposed to the universal neat oil. Typically, the polished surface is interrupted 7.2 High Temperature Oxidation Test 135

by small particles shown in Figure 7.7a. 3D-microscopy data shown in 7.7b reveals that the originally even surface has formed small hills with a particle sitting on each bump.

10 a) b) 0.8 9

0.7 8

7 0.6 6

0.5 5 Y[mm]

4 µm in height 0.4

3 0.3 2

0.2 1

0 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 X [mm]

Figure 7.7: Cemented carbide surface exposed to the universal neat oil after cleaning. a) Light microscopy image. b) Height data from 3D-microscopy.

The resulting disc surfaces vary strongly during the lifetime of a chamber. The smallest leak in the seal leads to a pressure drop, allows the MWF to leave the chamber and oxygen from the environment to enter the chamber. In this case, no change of the cemented carbide can be observed, as is seen in Figure 7.9c. Furthermore, the oxidation state of the surface of the chamber has an influence. A freshly machine container, as shown in Figure 7.9a, leads to irregular particles. The steadiest results are obtained after a run-in phase and before the container becomes untight which corresponds to Figure 7.9b. The strong dependence on the container conditions leads to unfavorable variations in the test results. However, measures to monitor the critical parameters, e.g. the pressure and oxygen concentration, are too expensive to meet the demand of a simple quick test. A series of tests, all performed at different temperatures using the universal neat oil is shown in Figure 7.8. At 780 ◦C, the surface is inert and only a soot layer is observed after the test. At 840 ◦C, small particles are formed. At 900 ◦C, surface deformation is first observed. At 1000 ◦C, the surface deforms into plateaus. Some of the cemented carbide discs are cross-sectioned to get an insight into the sub-surface changes. The discs are cut by wire EDM and polished with a BIB to retain the fragile 136 7. Wear Mechanisms

a) 740°C b) 840°C c) 900°C d) 1000°C

1 mm

Figure 7.8: Cemented carbide after exposure to ester-based neat oil vapor for 1 h at diﬀerent temperatures.

a) b) c)

1 mm

Figure 7.9: Different results of the cemented carbide surface depending on the state of the container. All tests are performed at 900 ◦C for 1 h with the universal neat oil. a) Freshly machined container surfaces. b) Second container use. c) Container untight. surface structures. A homogenous redeposited layer on top of the original surface and ripples in the vertical direction are formed as artifacts of this preparation method. While the sample exposed to the optimized oil does not show any sub-surface transformation as seen in Figure 7.10a, Figure 7.10b reveals cracks parallel to the surface, which are the reason for the bump formation with the universal oil. The particle is homogenous apart from a few inclusions most probably stemming from loose tungsten carbide grains. It consists mainly of cobalt, which leads to the conclusion that the binder was liquefied, and then resolidified when it could not wet the carbide grains anymore. The particle is loosely stuck to the surface with no visible interlayer. The surface of the degraded sample is more irregular than the undegraded surface but the tungsten carbide grains do not show any 7.2 High Temperature Oxidation Test 137

signs of degradation. The grains seem to be slightly rearranged. The top 5 µm have a slightly diﬀerent appearance, as shown in Figure 7.10b1.

a1) b1)

30 µm 30 µm a2) b2)

3 µm 3 µm

Figure 7.10: Cross-section of the cemented carbide disc after corrosion test. a) Optimized oil. b) Universal oil. 1) Overview. 2) Zoomed-in on surface 138 7. Wear Mechanisms

The EDX linescan orthogonal to the exposed surface shown in Figure 7.11 reveals an almost total absence of cobalt in the top 5 µm, clearly indicating severely impaired mechanical properties of the cemented carbide.

W Counts [-] Co m μ 5 5

Figure 7.11: EDX linescan orthogonal to the exposed surface. Tungsten counts (red) and cobalt counts (blue) are shown overlaid onto an SEM image.

The cemented carbide’s integrity below a bump is even more compromised by horizontal sub-surface cracks. They are visible in Figure 7.10b1, which shows a cross-section through a bump. The surface layer would easily shear off when coming into contact with a moving chip. Therefore, the results from the high temperature oxidation quick test correlate with the notch wear to some extent. Five of the tested experimental formulations show a strong degradation with binder particles forming on the surface. Three of these products are tested in cutting and all show significant notch wear, while the inert products do not lead to notch wear. A comparison between the two test results is shown in Figure 7.12. However, one anti-oxidative additive added to an active formulation leads to a different particle appearance in the oxidation quick test as shown in Figure 7.13. The particles are less contracted and have a dendritic shape. Although this may seem like a less severe degradation, with an average tool life of 6.2 min, the tool wears even slightly quicker with this oil in cutting tests. For comparison, the universal neat oil leads to an average tool life of 7.5 min. It may therefore be concluded that the anti-oxidative additive only changes the wettability of tungsten carbide while not changing the dissolution mechanism. 7.3 Pin Wear Oil 139

Liquefaction of the binder alloy is a posteriori only detectable if the binder alloy changed its morphology. Reversible melting of the cobalt binder without changes in the carbide- wettability or bump formation would not leave any traces in the quick test, but would lead to quick tool wear nevertheless. Thus, the quick test is only capable to detect cobalt dissolution if the dissolved binder alloy has a reduced carbide-wettability or bumps are formed. inhibited product active product

oxidation quick test

cutting test

1 mm

Figure 7.12: Comparison between the oxidation quick tests (top) and turning tests (bottom) for two diﬀerent neat oils (left and right).

7.3 Pin Wear Oil

To bridge the gap between the oxidation quick test and machining tests, wear is investigated with the tribometer presented in Section 6. The tribometer is used at a high speed of 100 m/min and with a high preload force of 400 N. Through frictional heat generation, this leads to similar temperatures as they occur in cutting. After the test, as shown in the top of Figure 7.14, the pins are coated with titanium, making it impossible to distinguish between diﬀerent wear forms. The pins are therefore etched to reveal the surface of the cemented carbide, as shown in the middle of Figure 7.14. The pins are then measured with a 3D microscope and the original spherical shape is subtracted to obtain the wear pattern, as shown in the bottom of Figure 7.14. 140 7. Wear Mechanisms

100 m

Figure 7.13: SEM image of the surface after using a product with an antioxidative additive. μ The pins exhibit a similar wear pattern as the cutting tools shown in Figure 7.3. Lateral to the central wear zone, two notches form. In these regions, the temperature is high, access for the MWF is provided and a shearing action removes weakened cemented carbide. In the center of the pin, similar conditions as in the crater wear region on a cutting tool occur. The temperatures are even higher than in the notch wear regions, but this central area is not accessible for MWFs or oxygen from the environment. The crater wear is asymmetrical because the ﬂow of titanium stagnates in front of the pin which reduces wear in the left side of the pins shown in Figure 7.3. Although the test is nearly as costly as a cutting test, the conditions in the contact zone are much more controllable and the gradients are smaller than in actual cutting, making it ideal to investigate wear patterns. In contrast to the oxidation quick test results, no exception in the correlation between notch wear in cutting and the test results could be observed. All MWFs that lead to 7.3 Pin Wear Oil 141

Ti 0.5 mm 0.5

0 depth [ m] 20

Figure 7.14: Pin wear with a standard oil (left) and a high performance oil (right). The evaluation chain is shown from top to bottom withμ the pins after the test (top), the pins after etching (middle), and 3D wear measurement at the bottom. The sliding direction of titanium over the pins is indicated with an arrow. [145] 142 7. Wear Mechanisms

notch wear on the pin lead to notch wear on the cutting tool as well. This holds true even for the MWF with an anti-oxidative additive, for which the pin wear result is shown in Figure 7.15.

0 m] μ depth [ depth

50 Ti 0.5 mm

Figure 7.15: Pin wear with an oil with an antioxidative agent. The sliding direction of titanium over the pin is indicated with an arrow.

7.4 Pin Wear Emulsion

Slight differences in the crater wear rate could be observed in turning tests with different emulsions described in Section 4.9. Friction measurements are a bad predictor for the differences in tool life as described in Section 6.7.4. Therefore, pin wear measurements are performed with severe friction and minimal emulsion flow rate to reach similar temperatures as in metal cutting. Two emulsions with known behavior in titanium cutting are compared. Emulsion 1 is a universal product and emulsion 2 is a specialized product for titanium cutting. Emulsion 2 is known to outperform emulsion 1 in terms of tool life by a factor of 3 to 10, depending on the cutting parameters. A first set of experiments is performed at 400 N and by keeping the speed at 100 m/min. The emulsions are delivered at a rate of 15 ml/min. Four samples are tested for each emulsion. The results are shown in Figure 7.16, on the left hand side. Emulsion 1 leads to lower wear rates than emulsion 2 in the tribometer setup, contrary to the experience in cutting. Emulsion 1 leads to a lower coefficient of friction than emulsion 2, which in turn 7.5 Dissolution of the Cobalt Binder 143

leads to a lower temperature in the friction zone. The wear progression is therefore mostly dependant on the temperature and, to a much smaller extent, on the anti-wear properties of the MWF. In cutting, most of the heat is generated through shearing of the chip and unlubricated friction in the crater zone. Contrary to the situation in the tribometer, the temperatures in cutting therefore do not significantly depend on the coefficient of friction in the lubricated friction regions. To eliminate the difference in temperature in the tribometer as well, constant-power measurements are performed. The normal force is kept at the same level of 400 N, while the velocity is adjusted in real-time to keep the friction power at 300 W. The result of four tests per emulsion are shown in Figure 7.16, on the right hand side. The wear behavior is inversed by controlling the power and matches now the ranking in cutting tests. The set- point of 300 W leads to slightly less severe conditions on average than in the uncontrolled reference; therefore, the worn volume is lower on average as well.

8 emulsion 1 emulsion 2

6 worn volumeworn [nl] 4

0 reference uncontrolled power controlled

Figure 7.16: Pin wear with two diﬀerent emulsions. Constant speed as uncontrolled reference (left, n=3) and power controlled at 300 W (right, n=4). The maximum and the minimum values are indicated with error bars.

7.5 Dissolution of the Cobalt Binder

In the absence of water, i.e. when using neat oils, tungsten carbide grains behave inertly at temperatures up to 1000 ◦C in the oxidation test. The part of wear that can be inﬂuenced by the MWF can therefore be traced back to the cobalt binder. Pure cobalt has a good heat resistance in air and a melting point of 1495 ◦C. The heat resistance stems from the formation of a dense self-passivation layer consisting of cobalt oxides. Chromium doping, as it is used in the inserts to inhibit grain growth during sintering, further increases the 144 7. Wear Mechanisms

resistance by forming chromium oxides on the cobalt alloy [145]. The high wear rate with certain MWFs, therefore, cannot be explained with the behavior of cobalt in air. Odelros et al. [106] suggest the formation of an alloy containing cobalt, titanium carbide, and titanium with a melting point lower than 1000 ◦C. Together with constituents of the MWF, chromium and vanadium from the cemented carbide, as well as aluminum and vanadium from the workpiece material, the melting point of this alloy drops to a value below 900 ◦C, as can be derived from the traces of liquid drop movements in Figure 7.17. 900 ◦C is often reached during cutting, as was shown in Section 4.9. The liquid phase can dissolve cobalt until reaching the solidus line. At this point, the alloy is saturated with cobalt and the cobalt dissolution seizes unless the other elements of the alloy are replenished. Since the mass transport in liquids is orders of magnitude faster than in solids, the other elements are transported quickly, therefore leading to the rapid progression of the liquid front and to tool wear.

1 mm

Figure 7.17: Cemented carbide after exposure to mineral-based neat oil vapor for 1 h at 900 ◦C.

The lack of wettability of the tungsten carbide with the molten alloy causes the liquid to leave the tungsten carbide grid and contract into droplets on the surface. The droplets move and collect smaller droplets, as can be seen in Figure 7.17. The alloy undergoes secondary changes, probably due to oxidation, changing the droplet structure with time. After the test, only cobalt is detectable in the droplets with EDX. The alloy in interaction with the solid binder therefore cannot be assumed to be in an equilibrium state. The long-term behavior of the droplets is, however, of no practical interest. Contrary to the situation in the oxidation quick test, any liquid leaving the cutting insert’s surface is 7.6 Oxidation of Tungsten Carbide 145

quickly removed by the flow of titanium under cutting conditions. Droplet formation is therefore an artifact of the oxidation quick test. MWF additives have a great influence on the described wear mechanism, as they can inhibit or promote the dissolution of cobalt. In the best case, they only leave the Ti-TiC- Co system described by Odelros et al. [106] with melting points over 1000 ◦C and much slower and controlled wear rates. In the worst case, they quickly oxidize cobalt and form an alloy with a melting point lower than 1000 ◦C. A phase change in the newly formed binder alloy is able to explain the drastic reduction of tool life at high feed with increased cutting speed or insufficient cooling described in Section 4.2. At low feeds or low cutting speeds, the melting point is not reached, whereas in the case of high feed and high speeds the cutting tool rapidly deteriorates. When the same MWFs are used in finish milling, a more uniform wear is observed, as the MWF is able to wet the whole cutting edge in an interrupted cut.

7.6 Oxidation of Tungsten Carbide

Tungsten carbide oxidizes by forming a tungsten oxide layer. In the presence of water vapor, i.e. when using water-miscible MWF, the oxide layer can be volatilized, leaving to a quicker oxidation. This is in accordance with the results obtained by Warren et al. [163]. Without any mechanical removal of the weakened grains, the surface forms a sponge-like structure as shown in Figure 7.18a. Oxidation of the tungsten carbide is slower than the dissolution of cobalt described above. The cobalt dissolution affects the top 5 µm, whereas tungsten carbide oxidation only affects the top 0.5 µm after one hour of exposure. Therefore, oxidation of tungsten carbide is not of large relevance for tool wear. Possibly, this mechanism leads to the formation of the characteristic ripples at the border of the contact zone indicated in Figure 7.18b with arrows, occuring only when cutting with a water-containing MWF. They do not influence tool life and are, therefore, not further investigated.

7.7 Decarburization of tungsten carbide

Titanium has a high aﬃnity to carbon. Carbon atoms constitute about 42% of the substrate material by number in the form of tungsten carbide WC. Carbon atoms leaving the surface leads to a reduction of the tungsten carbide to tungsten semicarbide W2C while oxidizing the titanium to titanium carbide T iC. Further reduction of tungsten leads to the 146 7. Wear Mechanisms a) b)

2 µm 0.5 mm

Figure 7.18: WC degraded by water vapor in the high temperature oxidation test. a) Back- scattered electron image of the cross-section of a cemented carbide surface exposed to an emulsion. b) Etched tool after turning with the same emulsion. Characteristic ripples are indicated with arrows. formation of elemental tungsten W . According to Zakharova et al. [185], tungsten semicarbide is even harder and has a higher wear resistance than tungsten carbide. However, the tungsten semicarbide is loosely attached to the surface and can easily be transported away with the titanium flow. Elemental tungsten is ductile and can be removed easily as well. A broad ion beam (BIB) is used to prepare the cross-section in the crater region of the rake face of a turning cutting tool shown in the right side of Figure 7.19. The interface between the adhering titanium layer and the substrate surface is investigated. The coating is completely worn in the crater area. Backscattering electron imaging reveals the density of the different phases, with heavier phases resulting in a stronger, brighter signal. A phase brighter and, therefore, denser than the tungsten carbide grains is identified at the interface. The only possible substances are tungsten semicarbide and tungsten. The reduced phase has a flaky texture and is therefore readily sheared off. The same effect can also be observed with emulsions, as a cross-section through a used pin seen in the left side of Figure 7.19 shows. 7.8 Crack Formation 147

1 m 1 m

Figure 7.19: Backscattered electronμ images of BIB cross-sections. Left: Pinμ of the in-process tribometer after pin wear test with emulsion. Right: Crater on the rake face of a turning cutting tool used with high performance oil.

7.8 Crack Formation

Hard tool materials tend to be brittle. They form fatigue cracks when loaded repeatedly with thermal and mechanical stresses. Usually, this is only an issue when using an interrupted cut. The tendency to form cracks is mainly determined by the toughness of the substrate, the MWF category, the cooling strategy and the process parameters. After crack formation, the cutting edge chips nearly symmetrically with similar chipping damage extended on both the rake and ﬂank face. Irregular comb-cracks at the surface, as seen in the left side of Figure 7.21, are promoted by corrosive cobalt degradation by the MWF through corrosion fatigue as shown by Wildner [177]. Sub-surface cracks, on the other hand, as seen in the right side of Figure 7.21, can only be inﬂuenced with the physical properties of the MWF. Lateral cracks cannot be observed in any tested case.

7.9 Combined Wear Model

Three main wear contributions are applied to the situation at a cutting edge in titanium cutting and a qualitative phenomenological tool wear model is derived from the results presented above. A uniform, continuous wear comprising crater wear, edge blunting, and ﬂank face wear forms to the same extent when turning with all suitable MWFs. This is a strong indication that the MWF is not able to penetrate the contact zone down to the cutting edge in a continuous cut. Instead, this type of wear is supposed to be strongly temperature depen- 148 7. Wear Mechanisms

0.240 mm

1 mm

Figure 7.20: Cutting edge after milling for 8 m with the roughing parameter set. The ﬂank face is shown on the left side, the rake face on the right side.

200 µm 100 µm

Figure 7.21: Backscattered electron microscopy of cracked cutting edges. Left: Comb cracks seen from the rake face after milling. Right: Cross-section of a cutting edge after milling, revealing sub-surface cracks. dent and therefore is sensitive to the cooling strategy and the MWF’s cooling ability. The wear is governed by the decarburization of the surface, forming W2C which is subsequently removed with the titanium flow. The remaining cobalt binder is probably sheared off due to its low strength or diffuses into the titanium as well. 7.10 Conclusion 149

The second wear portion includes notch wear and, in severe cases, pitting in the crater region in turning and increased uniform wear in finish milling. This type of wear is dominant when using unsuitable MWFs. This mechanism requires temperatures of at least 800 ◦C, contact of the tool surface with reactive species, including the oxygen in the atmosphere and additives of the MWF, and a periodic mechanical removal of the damaged parts. Under these conditions, with lower temperatures as well as slow and intermittent titanium flow, the tungsten carbide is inert. However, cobalt forms low melting alloys with unoxidized cobalt, doping elements in the cemented carbide and constituents of the MWF. As a liquid binder does not provide any stability, the brittle tungsten carbide skeleton can easily be mechanically sheared off. The third wear mechanism is due to chipping of the cutting edge. This is the dominant wear mechanism when milling with parameters that cause a high temperature amplitude in the region of the cutting edge. First leading to comb and subsurface cracks, the unstable cutting edge then quickly chips. Neat oils, with their lower cooling ability lead to lower temperature amplitudes and therefore reduce the chipping tendency. The chemical effect on the crack initiation and propagation remains unclear; however, an influence of the composition of the MWF is present. Possible explanations include wetting effects, crack corrosion, and crack initiation through pitting wear. Although these three mechanisms can act independently, they influence each other as they all act on the tool’s micro-geometry. For example, in the beginning, the temperature may be too low to cause melting of the newly formed cobalt alloy and, therefore, notch wear may not be pronounced. With progressing decarburization wear, the cutting edge gets blunter, thereby increasing the temperature in the notch region to the melting point of the oxidized cobalt alloy. This interaction makes it impossible to deduce a quantitative model from the given data. Nevertheless, the current qualitative model is sufficient to compare different MWFs and derive new formulations, as it is able to predict the effect of additives and the interactions between them. The plastic tool deformation through high temperature creep is not considered as a wear mechanism, as no tool material is removed in the process. The altered cutting edge shape can, however, lead to increased rubbing at the flank face, locally decreasing the rake angle and increasing the cutting edge radius, all leading to increased tool load and temperatures.

7.10 Conclusion

Dissolution of the cobalt binder, decarburization of the tungsten carbide, and crack formation are identiﬁed as the main tool wear mechanisms. The oxidation of tungsten carbide 150 7. Wear Mechanisms

only plays a minor role. With the exception of crack formation, all wear mechanisms can be studied in a quick test. The dissolution of the cobalt binder and tungsten carbide can be observed in the high-temperature oxidation test. Decarburization can be studied on worn pins from the in-process tribometer. Although abrasion is usually regarded as the reason for flank face wear, it plays a minor role in titanium cutting. Ti6Al4V does not contain hard precipitates. Instead, the same mechanism as in the crater wear zone may explain flank face wear as well. All relevant wear mechanisms are temperature dependent with higher wear at higher temperatures. The dissolution of the cobalt binder even leads to a sharp transition in the behavior above a certain temperature due to melting of the newly created alloy. It is, however, not possible to derive a quantitative wear law, because the conditions in the quick-tests cannot be quantified exactly and it is impossible to split the total worn volume quantitatively for each mechanism due to unknown interactions between the mechanisms. 151

Chapter 8

Summary and Outlook

This work had the aim to improve the MWF development process. The largest potential gains in terms of reducing the number of tests, time-to-market and understanding of the working mechanisms of MWFs are identified. Corresponding methods are developed and tested to shed some light on the previously unexplored action of MWFs in titanium cutting. Even fast and accurate MWF testing procedures cannot guarantee that the right MWF is chosen for a given application. Therefore, in the future, more focus should be put on the prediction of the tribo-system at the cutting edge. From the conditions at the cutting edge, the wear mechanisms can be estimated, from which the requirements for the MWF can in turn be derived. A suitable MWF can then be selected or developed. If a particular MWF is given, an estimation of how it influences the tribo-system based on its composition is necessary to efficiently develop new formulations. In parallel to the MWF, the machines, processes, tools, workpiece alloys, and regulations are continuously developed further, requiring adapting the MWFs. The biggest advances in the near future are expected in tool substrate and coating technology.

8.1 Wear Mechanisms

Several different wear mechanisms of cemented carbide are identified by analyzing worn tools as well as with the help of simplified test setups. The most basic mechanism is dissolution of the tungsten carbide grains by decarburization. This is a special case of diffusion wear. The process is strongly temperature dependent and is not influenced by the MWF. Cobalt as the binding agent is not crucial in this process, but is shown to have an accelerating effect by other researchers [106]. This effect is responsible 152 8. Summary and Outlook

for crater wear in titanium turning and for the largest portion of the uniform wear in milling when using a high performance MWF that suppresses the other wear mechanisms. The cobalt binder can be selectively degraded if exposed to an inappropriate MWF at high temperatures, which is a type of chemical wear. In general, this leads to notch wear, but can also lead to uniform wear in milling, where the MWF can wet the entire rake and flank face. This type of wear is most influenced by the MWF and is supposed to be the reason for the differences in tool life when using different MWFs for finish milling operations. Under roughing conditions, crack formation and crack growth overshadow this type of wear. The change in the MWF ranking in roughing compared to finishing indicates that a different mechanism is involved in notch wear than in crack initiation and growth. Tungsten carbide grains can decompose when they are exposed to MWF-vapor with a high water content.This is a type of chemical wear. This mechanism is not relevant in turning as it is expected to only lead to small ripple formation at the border of the crater zone, where it does not have a considerable influence on tool life. In milling, the effect is not distinguishable from other uniform wear mechanisms. Comb crack formation due to of thermo-mechanical fatigue causes stochastic tool failure, especially under roughing conditions. The crack growth is possibly enhanced by fatigue corrosion with inappropriate MWFs or inhibited with suitable ones. Plastic substrate deformation decreases the local rake angle and leads to a disadvanta- geous micro-geometry. As the bulk of the tool substrate is affected, this effect cannot be influenced by the MWF directly and is therefore not relevant in the MWF development. However, the MWF indirectly affects the plastic deformation via its influence on temperature and mechanical load due to friction. While the basic wear mechanisms could be discovered, the interactions between MWF and the tool surface on a molecular scale remain unclear. In future research, the influence of new additives on the different wear mechanisms and ultimately on the tool will be predictable. Conventional equipment to study additives on a molecular level is not suitable for use in cutting processes and MWFs. Therefore, special setups have to be designed.

8.2 Metal Working Fluids in Continuous Cutting

There are no signs of the MWF reaching the cutting edge in a continuous cut. The additives cannot act at the points of highest wear rate: close to the cutting edge. The conditions would certainly activate any present EP-additives, leading to a reduction in wear. Since such a reduction in wear is not observable with EP-additives, it can be concluded that EP-additives are not present in the areas of highest wear. Further evidence comes from 8.3 Metal Working Fluids in Interrupted Cutting 153

the cutting forces that are not significantly influenced by the MWF. It is very unlikely that a tribological film has the same adhesive properties as the bare titanium-tool contact. The cutting forces would change, if a significant portion of the contact between tool and workpiece would be lubricated. If the MWF acted on the ductility and crack formation of the workpiece material instead, a change in the cutting forces would be expected as well. A last piece of evidence comes from the microscopic analysis of the tool surface and tool cross-sections. They show a very strong bonding of titanium to the tool surface. There are no micro-channels, where a cutting fluid could creep into the contact zone. Therefore, as long as the MWF inhibits notch wear, no further improvement in tool life seems possible through further development of the MWF. This does not render the development of new MWFs for continuous cutting easy, because secondary quality aspects gain importance. For example, part surface staining, durability, universality, health and environmental aspects, machine cleaning abilities, smell, and look come more in focus. The tool life can be strongly influenced with the MWF application strategy. With no risk of thermo-mechanical fatigue in continuous cuts, the main goal is to keep the tool as cool possible. This reduces the chemical reaction rates, diffusion rates, and adhesion that govern wear and reduces the tool’s creep deformation rate. A clear correlation between the tool temperature and tool life is shown. Dry tools have the shortest tool life, followed by oil-cooled and emulsion-cooled tools. The longest tool life is achieved with fully synthetic fluids with the best cooling abilities. As the heat can only be transported away by the MWF at the edge of the contact zone, the tool is best cooled with one or more small, accurately targeted high-pressure jets. Similar effects are expected when machining stainless steel, nickel alloys, tantalum, or other materials known for strong adhesion to tool materials. However, these findings are by no means transferable to workpiece materials which are easier to cut, such as mild steel or brass. Multiple reviews [7, 43, 80, 111] highlight changes in cutting forces and thereby prove a significant penetration of the MWF into the gap between tool and chip or an influence of the MWF on chip shearing.

8.3 Metal Working Fluids in Interrupted Cutting

When cutting titanium with an interrupted cut, e.g. in milling, the influence of the MWF is more significant. In shoulder milling, the cutting edge is wet by the MWF more than half of the time. The hot cutting edge leaving the workpiece readily reacts with MWF constituents, which can lead to a deterioration of the surface in the worst case or the formation of a protective film in the best case. No change in the initial cutting force is observed when changing the MWF. Therefore, if a protective film is formed at all, it 154 8. Summary and Outlook

cannot prevent adhesion during the single cuts. The cutting edge is quenched rapidly by the MWF after leaving the workpiece, leading to thermo-mechanical stresses. Over the lifetime of a tool, a cutting edge under the investigated conditions experiences a number of load cycles in the order of magnitude of 104. Together with the mechanical stresses from the process force, this causes fatigue of the cutting edge and eventually leads to sudden chipping. The exact moment of chipping is stochastic; nevertheless, the average tool life is strongly influenced by the MWF. Two mechanisms are feasible. A good MWF could reduce the thermo-mechanical stress by causing a weaker thermal gradient. The bulk properties, such as density, viscosity, boiling point, heat capacity and thermal conductivity of different emulsion are similar. The difference in cooling properties, therefore, has to originate in a different evaporation behavior in contact with the hot cutting edge. The second, more likely reason is that the MWF does not reduce the stress level in the tool but instead inhibits crack initiation or growth at a given stress level. This mechanism has of course only an effect on comb cracks and does not affect the sub-surface crack formation. The mechanism of fatigue corrosion in cemented carbide exposed to MWFs is unexplored and the necessary MWF properties to mitigate crack initiation and growth remain unknown. With further research in this field, significant improvements in tool life seam feasible. Depending on the process parameters and cutting tools used, different wear modes and therefore different requirements for the MWF arise. The roughing and finishing parameters sets are summarized in Table 8.1. Some MWFs perform well under finishing conditions but lead to short tool lives in roughing conditions and vice versa. By preventing dissolution of the binder, oxidation, and decarburization of the carbide grains on one hand, and preventing crack initiation and growth on the other hand, high performance products that excel under both finishing and roughing conditions can be designed.

8.4 Wear Analogy Tests

The results from conventional tribometers deviate significantly from the results that are obtained under cutting conditions; therefore, conventional tribometers are not further investigated. An in-process tribometer, which removes the oxide layer in front of the pin, is designed and built. It is successfully used to measure the coefficient of friction under different conditions with different MWFs. Different neat oils lead to significantly different results depending on their base oil and additives. Water-miscible MWFs do not show significant difference in the coefficient of friction when the pins is flood-cooled. When the MWF flow is restricted and the temperature rises, differences become apparent. In all cases, the coefficient of friction does not correlate with tool wear. However, the wear of the pins turns out to correlate well with the wear in cutting tests. In-process tribometer 8.4 Wear Analogy Tests 155

Finishing Roughing Chosen process parameters Engagement angle [°] 45.6 126.9 Axial depth of cut [mm]3 2 Cutting speed [m/min] 80 70 Feed per tooth [mm] 0.1 0.1 Tool diameter [mm] 20 20 General tendencies Engagement angle small big Axial depth of cut small big Cutting speed high low Feed per tooth low high Cutting edge radius small big Rake angle big small Tool diameter small big Resulting conditions Mechanical tooth load low high Temperature amplitude low high Thermo-mechanical stress low high Wear Secondaryphasewear exponential ﬂank wear irrelevant land growth Ultimate tool failure catastrophic breakage of chipping of cutting edge cutting edge Tool failure mode mechanical overload due to crack growth weakens the changes in micro-geometry cutting edge Dominant wear mechanism s ubstrate dissolution thermo-mechanical fatigue MWF requirements Cooling strongest possible gentle Suitable MWF high water content oil or MQL Chemical properties prevent cobalt leaching prevent crack initiation and carbide dissolution and stress crack corrosion

Table 8.1: Overview of the ﬁnishing and roughing parameter set, combined with the general tendency for each parameter in the industrial application, the wear behavior, and the requirements for the MWF. 156 8. Summary and Outlook

tests are quicker, need less workpiece material, less MWF, and use cheaper consumables. The conditions caused by material deformation under the pin are much less complex than in metal cutting with chip formation and can be more accurately simulated. The ability to control the in-process tribometer in real-time added a whole range of new possibilities which need to be further investigated. A very simple and inexpensive quick test is developed with the high-temperature oxidation test, where a cemented carbide surface is exposed to MWF vapor. The lack of mechanical shearing action helps to isolate the effects of the MWF. The test is ideal to investigate the binder dissolution and carbide oxidation. The poor quantitative reproducibility limits the application in the qualitative investigation of previously unknown effects. The newly developed high temperature oxidation test and in-process wear tests are only validated against a small number of samples to check if the modeled wear mechanism re- sembles the one in cutting. More diverse samples will accumulate over time in an industrial test phase. With a larger sample size, the limit of applicability of each quick test can be defined better. Although only titanium cutting processes are investigated, some results and methods are expected to work similarly well with other materials. The extent to which this is true remains to be tested and gives insight into the working mechanism of MWFs with other materials.

8.5 Tool Life Testing

Full-scale cutting tests are and remain the most important but also the most expensive test when qualifying new MWFs. One of the main contributions to the high costs is the large number of test repetitions required to reach reliable results. In turning tests, the initial cutting edge radius correlates well with the tool life. Cutting edge radius measurements, either directly with a 3D Microscope or indirectly via the initial cutting force can therefore be employed to compensate the measured tool life and thereby reduce its standard deviation by a factor of three. Initially thought to serve as quick test for newly developed MWFs, the turning process proves to be insensitive to improvements of the MWF. Nevertheless, the turning process serves well to study crater wear and the underlying mechanisms as well as notch wear in a more realistic environment than in the high temperature oxidation quick test. The milling process is much more sensitive to the MWF than the turning process, because the cutting edge is intermittently exposed to the MWF. In the case of ﬁnishing conditions, the tools exhibit a smooth exponential force increase, which can be used to deﬁne a reliable 8.6 Developing Metalworking Fluids 157

tool life criterion. In the previously used optical tool life criterion, stochastic chipping events determined the tool life with optical measurement, which led to large variations. Using a robust fitting algorithm, the force data is evaluated without user influence. Under roughing conditions, this is not possible, as the cutting force stays nearly constant until the cutting edge chips and the force rises suddenly. In the case of milling, the information in the cutting force signal is equivalent or even superior to the information of the flank wear land width, because it is continuosly available and is less dependent on the environment. The evaluation algorithms are all used a posteriori on measurements that were terminated by an optical criterion. If the optical measurement is to be abolished to save time and equipment cost, a reliable test termination criterion must be developed based on the force data. It does not have to be equivalent to a tool life measurement. Instead, it has to detect unsafe conditions and terminate the test in case of catastrophic tool failure to prevent sensor and spindle damage. An automatic flank wear land measuring algorithm is developed. It is successfully applied whenever the flank wear land is optically well distinguishable from the unworn tool surface and the original position of the cutting edge is known. The automatic measurement disburdens the operator and leads to more reliable results than a manual measurement. The measurement algorithm is integrated in an intuitive user interface with additional functions for batch processing and report generation. With the elimination of the dominant disturbance factors, previously masked effects suddenly become apparent. For example, when eliminating the disturbance from stochastic cutting edge chipping in milling under finishing conditions, a slightly negative trend for tool life with every test repetition appears after exchanging the MWF. The reason for this behavior is unknown as additive consumption happens on much longer time-scales. Further investigation of this phenomenon can improve the repeatability and therefore the economy of milling tests.

8.6 Developing Metalworking Fluids

The findings can be condensed into a recommendation for the procedure when developing new MWFs for titanium. The procedure is shown in Figure 8.1. In a first step, appropriate additives have to be selected. If possible, additives known to suppress the binder degradation should be selected. If a new additive or even a new class of additives is to be added, its effect can be tested in the oxidation quick test before preparing a fully formulated MWF. New, fully formulated neat oils can be tested in a tribo-wear tester on a nascent titanium surface, similar to the in-process tribometer setup. The binder degradation can be assessed 158 8. Summary and Outlook

in such a setup with higher reliability than in the high temperature oxidation test. Tribo- wear testers are not able to emulate the temperatures in industrial applications in the case of water-miscible fluids due to their superior cooling ability. Water miscible fluids are therefore directly tested by cutting. Depending on the expected effort to correct issues with secondary requirements, the respective tests should be performed sooner or later in the development process. These tests include foaming, stability, flash points, and oxidation stability. The parameter set in the cutting tests has to be carefully chosen to match the later application. If a universal product is to be developed, at least one parameter set with continuous wear and one with crack formation should be tested. The measurement of the flank wear land width during the test is not necessary, as the same information about the tool condition can be extracted from the process forces without interrupting the process. If the wear mechanism deviates from the known effects, an in-depth analysis using surface analytics, such as electron microscopy with EDX is recommended. If possible, a model of the wear mechanism should be devised. If the effect cannot be predicted with sufficient accuracy from the model, a quick and inexpensive analogy test can be designed to test for the occurrence of the effect on a lab scale.

High-Temperature Oxidation Test

New Additives

De inition of Wear Mechanisms Water-Miscible Single Flute Foam Application Requirements Know-How Milling Test Stability Test � Neat Oils Flash Point Pin-Wear Test Unexpected Wear Wear Analytics

Figure 8.1: Suggested development process for MWFs for titanium.

8.7 Other Inﬂuences on the Cutting Productivity

Once the MWFs are operating near their optimal capacity, the optimization of other aspects of the cutting process gains importance. In other workpiece materials, the development of better coatings led to a small revolution with massively improved productivity in recent years. Up to now, titanium cutting could 8.7 Other Inﬂuences on the Cutting Productivity 159

not profit from these improvements, as no coating with suitable durability is available. If such a coating is found, the requirements for the MWF may change significantly, rendering the wear mechanisms specific to cemented carbide unimportant. Cemented carbide substrates are developed with a strong focus on hardness and toughness over a large temperature range. According to the findings, an improvement in titanium cutting tool life seems likely if the chemical resistance of the substrate is increased, even if the mechanical properties are slightly compromised. The recommended tool geometry for titanium cutting varies greatly from manufacturer to manufacturer, indicating optimization potential. Up to now, recommended tools from a single manufacturer are used in cutting tests. The same machine tool is used for comparison of MWFs. While the underlying wear mechanisms are expected to stay the same when using different tools or machines, the relative wear rates as well as transition points are expected to change. 160 8. Summary and Outlook Bibliography 161

Bibliography

[1] Abdalla H. S., Baines W., McIntyre G., Slade C., 2007, Development of novel sustainable neat-oil metal working ﬂuids for stainless steel and titanium alloy machining. Part 1. Formulation development, International Journal of Advanced Manufacturing Technology, 34(1-2):21–33, doi:10.1007/s00170-006-0585-4.

[2] Abele E Fr¨ohlich B, 2008, High-speed milling of titanium alloys, Ad- vances in Production Engineering & Management, 3(3):131–140, doi:10.3103/ s1068798x08100237.

[3] Ackroyd B., Chandrasekar S., Compton W. D., 2003, A Model for the Contact Conditions at the Chip-Tool Interface in Machining, Journal of Tribology, 125(3):649, doi:10.1115/1.1537747.

[4] Altinta¸sY. Budak E., 1995, Analytical Prediction of Stability Lobes in Milling, CIRP Annals - Manufacturing Technology, 44(1):357–362, doi:10.1016/S0007-8506(07) 62342-7.

[5] Amiril S. A. S., Rahim E. A., Talib N., Kamdani K., Rahim M. Z., Syahrullail S., 2017, Performance Evaluation of Palm- Olein TMP Ester Containing Hexagonal Boron Nitride and an Oil Miscible Ionic Liquid as Bio-Based Metalworking Fluids, Journal of Mechanical Engineering, SI 4(1):223–234.

[6] Amiril S. A.S., Rahim E. A., Embong Z., Syahrullail S., 2018, Tribological investigations on the application of oil-miscible ionic liquids additives in modiﬁed Jatropha- based metalworking ﬂuid, Tribology International, 120(September 2017):520–534, doi:10.1016/j.triboint.2018.01.030.

[7] Amiril S. A.S., Rahim E. A., Syahrullail S., 2017, A review on ionic liquids as sustainable lubricants in manufacturing and engineering: Recent research, performance, and applications, Journal of Cleaner Production, 168:1571–1589, doi:10.1016/j. jclepro.2017.03.197. 162 Bibliography

[8] Archard J. F., 1953, Contact and rubbing of ﬂat surfaces, Journal of Applied Physics, 24(8):981–988, doi:10.1063/1.1721448.

[9] Arsecularatne J. A., Zhang L. C., Montross C., 2006, Wear and tool life of tungsten carbide, PCBN and PCD cutting tools, International Journal of Machine Tools and Manufacture, 46(5):482–491, doi:10.1016/j.ijmachtools.2005.07.015.

[10] Astakhov V. P., 2012, Environmentally friendly near-dry machining of metals, In Metalworking Fluids (MWFs) for Cutting and Grinding: Fundamentals and Recent Advances, pages 135–200. doi:10.1016/B978-0-85709-061-4.50004-9.

[11] Axinte Dragos A., Belluco Walter, De Chiffre Leonardo, 2001, Reliable tool life measurements in turning - An application to cutting fluid efficiency evaluation, In- ternational Journal of Machine Tools and Manufacture, 41(7):1003–1014, doi:10. 1016/S0890-6955(00)00110-3.

[12] Bagchi H. Basu S.K., may 1974, SELECTION OF CUTTING FLUID, Industrial Lubrication and Tribology, 26(5):181–182, doi:10.1108/eb053073.

[13] Bagchi H., Mukherjee N.P., Basu S.K., Tresidder C.G., may 1972, Investigation of Metal Cutting Using Molybdenum Disulphide as a Cutting Fluid, Industrial Lubri- cation and Tribology, 24(5):239–258, doi:10.1108/eb053024.

[14] Banerjee Nilanjan Sharma Abhay, 2014, Identiﬁcation of a friction model for minimum quantity lubrication machining, Journal of Cleaner Production, 83:437–443, doi:10.1016/j.jclepro.2014.07.034.

[15] Barker D. C., Johnston G. J., Spikes H. A., Bunemann T. F., 1993, Ehd ﬁlm formation and starvation of oil-in-water emulsions, Tribology Transactions, 36(4):565–572, doi:10.1080/10402009308983196.

[16] BARLOW P. L., 1966, Rehbinder Eﬀect in Lubricated Metal Cutting, Nature, 211(5053):1076–1077, doi:10.1038/2111076a0.

[17] Bassett E., K¨ohler J., Denkena B., 2012, On the honed cutting edge and its side eﬀects during orthogonal turning operations of AISI1045 with coated WC- Co inserts, CIRP Journal of Manufacturing Science and Technology, 5(2):108–126, doi:10.1016/j.cirpj.2012.03.004.

[18] Bassett Edmond, 2014. Belastungsspeziﬁsche Auslegung und Herstellung von Schneidkanten fur¨ Drehwerkzeuge. Bibliography 163

[19] Bataller H., Lamaallam S., Lachaise J., Graciaa A., Dicharry C., 2004, Cutting ﬂuid emulsions produced by dilution of a cutting ﬂuid concentrate containing a cationic/nonionic surfactant mixture, Journal of Materials Processing Technology, 152(2):215–220, doi:10.1016/j.jmatprotec.2004.03.027.

[20] Belluco W. De Chiﬀre L., 2004, Performance evaluation of vegetable-based oils in drilling austenitic stainless steel, Journal of Materials Processing Technology, 148(2):171–176, doi:10.1016/S0924-0136(03)00679-4.

[21] Belluco Walter, 2001, Performance Testing of Cutting Fluids, PhD thesis, Technical University of Denmark.

[22] Bierma Thomas J Waterstraat Frank L, 2004, WMRC Reports, Technical Report April, Illinois State Universit.

[23] Blaser Swisslube AG, 2018. Our Commitment. https://www.blaser.com/ download/5ecc5fbe-b80d-4a57-8917-dbb65ea18314.

[24] Bordin A., Bruschi S., Ghiotti A., Bariani P. F., 2015, Analysis of tool wear in cryogenic machining of additive manufactured Ti6Al4V alloy, Wear, 328-329:89–99, doi:10.1016/j.wear.2015.01.030.

[25] Bork Carlos Alberto Schuch, Gon¸calves Janaina Fracaro de Souza, Gomes Jeﬀerson de Oliveira, Gheller Jord˜ao, 2014, Performance of the jatropha vegetable-base soluble cutting oil as a renewable source in the aluminum alloy 7050-T7451 milling, CIRP Journal of Manufacturing Science and Technology, 7(3):210–221, doi:10.1016/j. cirpj.2014.04.004.

[26] Brecher Christian Weck Manfred, 2017, Werkzeugmaschinen Fertigungssysteme. Springer Berlin Heidelberg, Berlin, Heidelberg, doi:10.1007/978-3-662-46567-7.

[27] Brinksmeier E, Meyer D, Huesmann-cordes A G, Herrmann C, 2015, CIRP Annals - Manufacturing Technology Metalworking ﬂuids – Mechanisms and performance, CIRP Annals - Manufacturing Technology, 64(2):605–628, doi:10.1016/j.cirp. 2015.05.003.

[28] Brown R.H. Armarego E.J.A., jun 1964, Oblique machining with a single cutting edge, International Journal of Machine Tool Design and Research, 4(1):9–25, doi:10. 1016/0020-7357(64)90006-X.

[29] Burton Geoﬀ, Goo Chan Seo, Zhang Yanqiao, Jun Martin B G, 2014, Use of vegetable oil in water emulsion achieved through ultrasonic atomization as cut- 164 Bibliography

ting ﬂuids in micro-milling, Journal of Manufacturing Processes, 16(3):405–413, doi:10.1016/j.jmapro.2014.04.005.

[30] Byers Jerry P., editor, sep 2017, Metalworking Fluids, Third Edition. CRC Press, Third edition. | Boca Raton : Taylor & Francis, CRC Press, 2017., doi:10.4324/ 9781351228213.

[31] Cambiella A., Benito J. M., Pazos C., Coca J., Hernández A., Fernández J. E., 2007, Formulation of emulsifiable cutting fluids and extreme pressure behaviour, Journal of Materials Processing Technology, 184(1-3):139–145, doi:10.1016/j.jmatprotec. 2006.11.014.

[32] Canter Neil, 2018, Bridging the gap between metalworking ﬂuid bench tests and machining operations, Tribology & Lubrication Technology, (March 2018):46–58.

[33] Cassin C. Boothroyd G., 1965, Lubricating Action of Cutting Fluids, Journal of Me- chanical Engineering Science, 7(1):67–81, doi:10.1243/jmes_jour_1965_007_012_ 02.

[34] Chandrasekaran H. Johansson J. O., 1994, Chip Flow and Notch Wear Mechanisms during the Machining of High Austenitic Stainless Steels, CIRP Annals - Manu- facturing Technology, 43(1):101–105, doi:10.1016/S0007-8506(07)62174-X.

[35] Claudin C., Mondelin A., Rech J., Fromentin G., 2010, Eﬀects of a straight oil on friction at the toolworkmaterial interface in machining, International Journal of Machine Tools and Manufacture, 50(8):681–688, doi:10.1016/j.ijmachtools. 2010.04.013.

[36] Cook N. H., 1973, Tool Wear and Tool Life, Journal of Engineering for Industry, 95(4):931–938.

[37] Cook N. H. Lang A. B., 1963, Criticism of Radioactive Tool-Life Testing, Journal of Engineering for Industry, 85(4):381–385, doi:10.1115/1.3669899.

[38] De Chiﬀre L. Belluco W., 2000, Comparison of methods for cutting ﬂuid performance testing, CIRP Annals - Manufacturing Technology, 49(1):57–60, doi:10. 1016/S0007-8506(07)62895-9.

[39] De Chiffre L Belluco W, 2002, Investigations of cutting fluid performance using different machining operations, Lubrication Engineering, 58(10):22–29.

[40] De Chiﬀre L., Lassen S., Pedersen K. B., Skade S., 1994, A reaming test for cutting ﬂuid evaluation, Journal of Synthetic Lubrication, 11(1):17–34, doi:10.1002/jsl. 3000110103. Bibliography 165

[41] De Chiﬀre Leonardo, 1980, Mechanical testing and selection of cutting ﬂuids, Lubr. Eng, 36:33–39.

[42] Dearnley P.A. Grearson A.N., 1986, Evaluation of principal wear mechanisms of cemented carbides and ceramics used for machining titanium alloy IMI 318, Materials Science and Technology, 2(1):47–58, doi:10.1179/mst.1986.2.1.47.

[43] Debnath Sujan, Reddy Moola Mohan, Yi Qua Sok, 2014, Environmental friendly cutting ﬂuids and cooling techniques in machining: A review, Journal of Cleaner Production, 83:33–47, doi:10.1016/j.jclepro.2014.07.071.

[44] Deiab Ibrahim, Raza Syed Waqar, Pervaiz Salman, 2014, Analysis of lubrication strategies for sustainable machining during turning of titanium ti-6al-4v alloy, Pro- cedia CIRP, 17:766–771, doi:10.1016/j.procir.2014.01.112.

[45] Denkena B. Biermann D., 2014, Cutting edge geometries, CIRP Annals - Manu- facturing Technology, 63(2):631–653, doi:10.1016/j.cirp.2014.05.009.

[46] Denkena B, Reichstein M, Brodehl JLGL, de Le´on Garc´ıa L, 2005, Surface preparation, coating and wear performance of geometrically deﬁned cutting edges, In Proceedings of the 5th international conference the coatings in manufacturing engineering, pages 5–7, 2005.

[47] Denkena Berend, Koehler Jens, Rehe Michael, 2012, Inﬂuence of the Honed Cutting Edge on Tool Wear and Surface Integrity in Slot Milling of 42CrMo4 Steel, Procedia CIRP, 1:190–195, doi:10.1016/j.procir.2012.04.033.

[48] Eckstein V M, Lebkuechner G, Blum D, 1992, Schaftfr¨asen von Titanlegierungen mit hohen Schnittgeschwindigkeiten, VDI-Z, 134(6):61–67.

[49] Ega˜na A., Rech J., Arrazola P. J., sep 2012, Characterization of Friction and Heat Partition Coeﬃcients during Machining of a TiAl6V4 Titanium Alloy and a Ce- mented Carbide, Tribology Transactions, 55(5):665–676, doi:10.1080/10402004. 2012.692007.

[50] Evans R, 2012, 2 - Selection and testing of metalworking ﬂuids, In Astakhov V P Joksch S, editors, Metalworking Fluids (MWFs) for Cutting and Grinding, Wood- head Publishing Series in Metals and Surface Engineering, pages 23–78. Woodhead Publishing, doi:https://doi.org/10.1533/9780857095305.23.

[51] Ezugwu E. O., Bonney J., Yamane Y., 2003, An overview of the machinability of aeroengine alloys, Journal of Materials Processing Technology, 134(2):233–253, doi:10.1016/S0924-0136(02)01042-7. 166 Bibliography

[52] Ezugwu E. O., Da Silva R. B., Bonney J., MacHado A´ R., 2005, Evaluation of the performance of CBN tools when turning Ti-6Al-4V alloy with high pressure coolant supplies, International Journal of Machine Tools and Manufacture, 45(9):1009–1014, doi:10.1016/j.ijmachtools.2004.11.027.

[53] Fischler Martin A Bolles Robert C, 1981, Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography, Commun. ACM, 24(6):381–395, doi:10.1145/358669.358692.

[54] Ghani Jaharah A., Che Haron Che Hassan, Hamdan Siti Hartini, Md Said Ah- mad Yasir, Tomadi Siti Haryani, 2013, Failure mode analysis of carbide cutting tools used for machining titanium alloy, Ceramics International, 39(4):4449–4456, doi:10.1016/j.ceramint.2012.11.038.

[55] Goindi G. S., Sarkar P., Jayal A. D., Chavan S. N., Mandal D., jan 2018, Investigation of ionic liquids as additives to canola oil in minimum quantity lubrication milling of plain medium carbon steel, The International Journal of Advanced Manufacturing Technology, 94(1-4):881–896, doi:10.1007/s00170-017-0970-1.

[56] Goindi Gyanendra Singh, Chavan Santosh Namdeo, Mandal Debaprasad, Sarkar Prabir, Jayal Anshu Dhar, 2015, Investigation of ionic liquids as novel metalworking ﬂuids during minimum quantity lubrication machining of a plain carbon steel, Procedia CIRP, 26:341–345, doi:10.1016/j.procir.2014.09.002.

[57] Grzesik Wit Zak Krzysztof, 2013, Friction quantiﬁcation in the oblique cutting with CBN chamfered tools, Wear, 304(1-2):36–42, doi:10.1016/j.wear.2013.04.020.

[58] Gustoc. Die Korrosionsbest¨andigkeit des Titans. http://www.gustoc.de/de/ Vorteile/Titan_Korrosion.html (Accessed: 2017-08-28).

[59] Hartung P. D., Kramer B. M., Turkovich B. F. von, 1982, Tool Wear in Titanium Machining, CIRP Annals - Manufacturing Technology, 31(1):75–80, doi:10.1016/ S0007-8506(07)63272-7.

[60] Hartung Paul Dudley, 1981, Tool Wear in Titanium Machining, PhD thesis, Mas- sachusetts Institute of Technology.

[61] Hedenqvist Per Olsson Mikael, 1991, Sliding wear testing of coated cutting tool materials, Tribology International, 24(3):143–150, doi:10.1016/0301-679X(91)90020-A.

[62] Hertz H., 1881, H. Hertz, Uber¨ die Beruhrung¨ fester elastischer K¨orper, Journal fur¨ die reine und angewandte Mathematik 92, 156-171 (1881), Journal fur¨ die reine und angewandte Mathematik, 92:156–171. Bibliography 167

[63] Huang Chihyung, Lee Seongeyl, Sullivan John P., Chandrasekar Srinivasan, aug 2007, In Situ Measurement of Fluid Film Thickness in Machining, Tribology Letters, 28(1):39–44, doi:10.1007/s11249-007-9245-6.

[64] Huesmann-Cordes A. G., Meyer D., Brinksmeier E., Schulz J., 2014, Inﬂuence of additives in metalworking ﬂuids on the wear resistance of steels, Procedia CIRP, 13:108–113, doi:10.1016/j.procir.2014.04.019.

[65] Jagadish Ray Amitava, 2015, Green Cutting Fluid Selection Using Moosra Method, International Journal of Research in Engineering and Technology, 03(15):559–563, doi:10.15623/ijret.2014.0315105.

[66] Jain V. K. Shukla D. S., 1996, Study of the EP activity of water-soluble inorganic metallic salts for aqueous cutting ﬂuids, Wear, 193(2):226–234, doi:10.1016/ 0043-1648(95)06774-4.

[67] Jawaid A., Sharif S., Koksal S., 2000, Evaluation of wear mechanisms of coated carbide tools when face milling titanium alloy, Journal of Materials Processing Tech- nology, 99(1):266–274, doi:10.1016/S0924-0136(99)00438-0.

[68] Jayal Anshu D. Balaji A. K., 2009, Eﬀects of cutting ﬂuid application on tool wear in machining: Interactions with tool-coatings and tool surface features, Wear, 267(9- 10):1723–1730, doi:10.1016/j.wear.2009.06.032.

[69] Jhita A S, Jain V K, Pandey P C, 1982, On the tool wear during face turning, In Proceedings of the Twenty-second International Machine Tool Design and Research Conference, pages 247–253. Springer, 1982.

[70] John Jacob, Bhattacharya Mrinal, Raynor Peter C., 2004, Emulsions containing vegetable oils for cutting ﬂuid application, Colloids and Surfaces A: Physicochemical and Engineering Aspects, 237(1-3):141–150, doi:10.1016/j.colsurfa.2003.12.029.

[71] Joksch Stefan, Leistungsbewertung von wassermischbaren Kuhlschmierstoﬀen¨ durch Triboteste und reale Bearbeitungsprozesse.

[72] Kadmer E.H., Danninger H., Reichert H., 1958, Ueber die Messung des Last- Trageverm¨ogens von Oelen und Emulsionen in einer Reibverschleisswaage, Mineral¨ol- Technik, 2(1).

[73] Karpat Yigit Ozel¨ Tugrul, 2008, Mechanics of high speed cutting with curvilinear edge tools, International Journal of Machine Tools and Manufacture, 48(2):195–208, doi:10.1016/j.ijmachtools.2007.08.015. 168 Bibliography

[74] Klocke F., Krieg T., Lugscheider E., Bobzin K., 2001, Testing and design of tool coatings with properties adapted to the use of biodegradable cutting ﬂuids, CIRP Annals - Manufacturing Technology, 50(1):57–60, doi:10.1016/S0007-8506(07)62070-8.

[75] Klocke Fritz, 2010, Manufacturing processes 1: Cutting. Springer-Verlag Berlin Hei- delberg, doi:10.1007/978-3-642-11979-8.

[76] Komanduri R. Von Turkovich B.F., jun 1981, New observations on the mechanism of chip formation when machining titanium alloys, Wear, 69(2):179–188, doi:10.1016/ 0043-1648(81)90242-8.

[77] Kuljanic E., Fioretti M, Beltrame L, Miani F., 1998, Milling Titanium Compressor Blades with PCD Cutter, CIRP Annals, 47(1):61–64, doi:10.1016/S0007-8506(07) 62785-1.

[78] Landels John Gray, 2000, Engineering in the Ancient World. Univ of California Press.

[79] Lawal S. A., Choudhury I. A., Nukman Y., 2013, Developments in the formulation and application of vegetable oil-based metalworking ﬂuids in turning process, International Journal of Advanced Manufacturing Technology, 67(5-8):1765–1776, doi:10.1007/s00170-012-4607-0.

[80] Lawal Sunday Albert, 2013, A review of application of vegetable oil-based cutting ﬂuids in machining non-ferrous metals, Indian Journal of Science and Technology, 6(1):3951–3956, doi:10.1016/j.ijmachtools.2011.09.003.

[81] Lawal Sunday Albert, Choudhury Imtiaz Ahmed, Nukman Yussof, 2014, Evaluation of vegetable and mineral oil-in-water emulsion cutting ﬂuids in turning AISI 4340 steel with coated carbide tools, Journal of Cleaner Production, 66:610–618, doi:10. 1016/j.jclepro.2013.11.066.

[82] Lee Kwangho, Hwang Yujin, Cheong Seongir, Choi Youngmin, Kwon Laeun, Lee Jaekeun, Kim Soo Hyung, 2009, Understanding the role of nanoparticles in nano- oil lubrication, Tribology Letters, 35(2):127–131, doi:10.1007/s11249-009-9441-7.

[83] Lei Shuting Liu Wenjie, 2002, High-speed machining of titanium alloys using the driven rotary tool, International Journal of Machine Tools & Manufacture, 42(42):653–661.

[84] Li Liang, Chang H., Wang M., Zuo Dun Wen, He Lin, 2004, Temperature Mea- surement in High Speed Milling Ti6Al4V, Key Engineering Materials, 259-260:804– 808, doi:10.4028/www.scientific.net/kem.259-260.804. Bibliography 169

[85] Libardi A., Schmid S. R., Sen M., Schneider W., 2013, Evaluation of ionic ﬂuids as lubricants in manufacturing, Journal of Manufacturing Processes, 15(4):414–418, doi:10.1016/j.jmapro.2013.06.005.

[86] Lorenz G, 1968, A rapid machinability test for small diameter bar stock, Ann. CIRP, 16:377.

[87] Lorenz G., 1985, Reliable Cutting Fluid Rating, CIRP Annals - Manufacturing Tech- nology, 34(1):95–99, doi:10.1016/S0007-8506(07)61732-6.

[88] Lutjering¨ Gerd, 2007, Titanium. Engineering Materials, Processes. Springer Berlin Heidelberg, Berlin, Heidelberg, 2nd ed. edition, doi:10.1007/978-3-540-73036-1.

[89] Machado A R Wallbank J, 1990, Machining of Titanium and its Alloys–a Review, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of En- gineering Manufacture, 204(1):53–60, doi:10.1243/PIME_PROC_1990_204_047_02.

[90] Madhavan V., Chandrasekar S., Farris T. N., 2002, Direct Observations of the Chip-Tool Interface in the Low Speed Cutting of Pure Metals, Journal of Tribol- ogy, 124(3):617, doi:10.1115/1.1398546.

[91] Malakooti Behnam, Wang Jun, Tandler Evan C., 1990, A sensor-based accelerated approach for multi-attribute machinability and tool life evaluation, International Journal of Production Research, 28(12):2373–2392, doi:10.1080/ 00207549008942872.

[92] Meier Linus, 2019, Methods to reduce variation in cutting tool life tests, Interna- tional Journal of Advanced Manufacturing Technology, pages 355–365, doi:10.1007/ s00170-019-03558-9.

[93] Merchant M. Eugene, 1945, Mechanics of the metal cutting process. I. Orthogonal cutting and a type 2 chip, Journal of Applied Physics, 16(5):267–275, doi:10.1063/ 1.1707586.

[94] Merchant M. Eugene, 1945, Mechanics of the metal cutting process. II. Plasticity conditions in orthogonal cutting, Journal of Applied Physics, 16(6):318–324, doi:10. 1063/1.1707596.

[95] Merchant M Eugene, Ernst Hans, Krabacher E J, 1953, Radioactive cutting tools for rapid tool-life testing, TRANS. ASME, 75:549.

[96] Merchant M. Eugene Krabacher E. J., 1951, Radioactive tracers for rapid measurement of cutting tool life, Journal of Applied Physics, 22(12):1507–1508, doi:10. 1063/1.1699907. 170 Bibliography

[97] Merchant M Eugene Shaw M. C., 1950, Machining theory and practice, Am Soc Metals, 9:5–44.

[98] Meyer Daniel, Redetzky Marvin, Brinksmeier Ekkard, 2017, Microbial-based metalworking ﬂuids in milling operations, CIRP Annals - Manufacturing Technology, 66(1):129–132, doi:10.1016/j.cirp.2017.04.019.

[99] Minfray Clotilde, Fromentin Guillaume, Bierla Aleksandra, Martin Jean Michel, Le Mogne Thierry, 2014, The eﬀect of an organic pentasulﬁde EP additive in turning and milling operations, Wear, 317(1-2):129–140, doi:10.1016/j.wear.2014.05.009.

[100] Moufki A., Dudzinski D., Molinari A., Rausch M., 2000, Thermoviscoplastic modelling of oblique cutting: Forces and chip ﬂow predictions, International Journal of Mechanical Sciences, 42(6):1205–1232, doi:10.1016/S0020-7403(99)00036-3.

[101] Mu Liwen, Wu Jian, Matsakas Leonidas, Chen Minjiao, Vahidi Alireza, Grahn Mat- tias, Rova Ulrika, Christakopoulos Paul, Zhu Jiahua, Shi Yijun, 2018, Lignin from hardwood and softwood biomass as a lubricating additive to ethylene glycol, Molecules, 23(3):1–10, doi:10.3390/molecules23030537.

[102] Naerheim Y Kendig M, 1987, Evaluation of Cutting Fluid Eﬀectiveness in Machining using Electrochamical Techniques, Wear, 114:51–57.

[103] Nageswara Rao D. Vamsi Krishna P., 2008, The inﬂuence of solid lubricant particle size on machining parameters in turning, International Journal of Machine Tools and Manufacture, 48(1):107–111, doi:10.1016/j.ijmachtools.2007.07.007.

[104] Nakagawa H., Watanabe S., Fujita T., Sakamoto M., 1998, Characteristic properties of cutting ﬂuid additives made from the derivatives of some polymeric nonionic surface-active agents, Journal of Surfactants and Detergents, 1(2):207–211, doi:10. 1007/s11743-998-0021-7.

[105] Norgard Gregory Bremer Peer-Timo, jul 2013, Ridge-Valley graphs: Combinatorial ridge detection using Jacobi sets, Computer Aided Geometric Design, 30(6):597–608, doi:10.1016/j.cagd.2012.03.015.

[106] Odelros S., Kaplan B., Kritikos M., Johansson M., Norgren S., 2017, Experimental and theoretical study of the microscopic crater wear mechanism in titanium machining, Wear, 376-377:115–124, doi:10.1016/j.wear.2017.01.104.

[107] Olsson Mikael, Söderberg Staffan, Jacobson Staffan, Hogmark Sture, 1989, Sim- ulation of cutting tool wear by a modified pin-on-disc test, International Journal Bibliography 171

of Machine Tools and Manufacture, 29(3):377–390, doi:10.1016/0890-6955(89) 90007-2.

[108] Opitz Herwart Hake Oswald, 1956, Wear Analysis of Hard Metal Turning Tools by Means of Radioisotope, Microtecnic, 10(1):5–9.

[109] Orowan E., 1943, The Calculation of Roll Pressure in Hot and Cold Flat Rolling, Pro- ceedings of the Institution of Mechanical Engineers, 150(1):140–167, doi:10.1243/ pime_proc_1943_150_025_02.

[110] Orra Kashfull Choudhury Sounak K., 2018, Mechanistic modelling for predicting cutting forces in machining considering eﬀect of tool nose radius on chip formation and tool wear land, International Journal of Mechanical Sciences, 142-143(December 2017):255–268, doi:10.1016/j.ijmecsci.2018.05.004.

[111] Osama Mohamed, Singh Amarpreet, Walvekar Rashmi, Khalid Mohammad, Chan- dra Thummalapalli, Manikyam Sekhara, Wai Wong, 2017, Tribology International Recent developments and performance review of metal working ﬂuids, Tribology In- ternational, 114(May):389–401, doi:10.1016/j.triboint.2017.04.050.

[112] Paulo Davim J, 2014, Machining of Titanium Alloys. Springer Berlin Heidelberg.

[113] Peikert Ronald Sadlo Filip, mar 2008, Height Ridge Computation and Filtering for Visualization, 2008 IEEE Paciﬁc Visualization Symposium, pages 119–126, doi:10. 1109/PACIFICVIS.2008.4475467.

[114] Pervaiz Salman, Rashid Amir, Deiab Ibrahim, Nicolescu Cornel Mihai, 2016, An experimental investigation on eﬀect of minimum quantity cooling lubrication (MQCL) in machining titanium alloy (Ti6Al4V), International Journal of Advanced Manu- facturing Technology, 87(5-8):1371–1386, doi:10.1007/s00170-016-8969-6.

[115] Pham Minh Quan, Yoon Hae Sung, Khare Varsha, Ahn Sung Hoon, 2014, Evaluation of ionic liquids as lubricants in micro milling - Process capability and sustainability, Journal of Cleaner Production, 76:167–173, doi:10.1016/j.jclepro.2014.04.055.

[116] Popov Alexey Krasnikova Iuliia, 2018, Reducing Labor Intensity in the Develop- ment of New Universal Cutting Fluids for Machining, Manufacturing Technology, 18(1):99–105.

[117] Postnikov S. N., 1967, Penetrating ability of cutting ﬂuids, Wear, 10(2):142–150, doi:10.1016/0043-1648(67)90085-3. 172 Bibliography

[118] Prasad Kanika Chakraborty Shankar, 2016, A quality function deployment-based model for cutting ﬂuid selection, Advances in Tribology, 2016:10, doi:10.1155/2016/ 3978102.

[119] Puls H., Klocke F., Lung D., 2014, Experimental investigation on friction under metal cutting conditions, Wear, 310(1-2):63–71, doi:10.1016/j.wear.2013.12.020.

[120] Puls Hendrik, Klocke Fritz, Lung Dieter, 2012, A new experimental methodology to analyse the friction behaviour at the tool-chip interface in metal cutting, Production Engineering, 6(4-5):349–354, doi:10.1007/s11740-012-0386-6.

[121] Rao R. Venkata Gandhi O. P., 2001, Digraph and matrix method for the selection, identiﬁcation and comparison of metal-cutting ﬂuids, Proceedings of the Institution of Mechanical Engineers, Part J: Journal of Engineering Tribology, 215(1):25–33, doi:10.1243/1350650011541710.

[122] Rech J., Arrazola P. J., Claudin C., Courbon C., Pusavec F., Kopac J., 2013, Characterisation of friction and heat partition coeﬃcients at the tool-work material interface in cutting, CIRP Annals - Manufacturing Technology, 62(1):79–82, doi:10.1016/j.cirp.2013.03.099.

[123] Redetzky M., Rabenstein A., Seidel B., Brinksmeier E., Wilhelm H., 2015, The inﬂuence of cell counts, cell size, EPS and microbial inclusions on the lubrication properties of microorganisms, Production Engineering, 9(2):149–159, doi:10.1007/ s11740-014-0592-5.

[124] Redetzky Marvin, Rabenstein Andreas, Palmowski B., Brinksmeier Ekkard, 2014, Microorganisms as a Replacement for Metal Working Fluids, Advanced Materials Research, 966-967:357–364, doi:10.4028/www.scientific.net/AMR.966-967.357.

[125] Rehbinder P., 1947, New Physico-Chemical Phenomena in the Deformation and Mechanical Treatment of Solids, Nature, 159:866–867, doi:10.1038/160650a0.

[126] Schaal N., Kuster F., Wegener K., jan 2015, Springback in Metal Cutting with High Cutting Speeds, Procedia CIRP, 31:24–28, doi:10.1016/J.PROCIR.2015.03.065.

[127] Schulz Joachim, Brinksmeier Ekkard, Meyer Daniel, 2013, On the Interactions of Additives in Metalworking Fluids with Metal Surfaces, Lubricants, 1(4):75–94, doi:10.3390/lubricants1040075.

[128] Sharif S. Rahim E. A., 2007, Performance of coated- and uncoated-carbide tools when drilling titanium alloy-Ti-6Al4V, Journal of Materials Processing Technology, 185(1-3):72–76, doi:10.1016/j.jmatprotec.2006.03.142. Bibliography 173

[129] Sharma Kumar A., Tiwari Arun Kumar, Dixit Amit Rai, 2015, Mechanism of Nanoparticles functioning and Eﬀects in Machining Processes: A review, Materi- als Today: Proceedings, 2(4-5):3539–3544, doi:10.1016/j.matpr.2015.07.331.

[130] Shashidhara Y. M. Jayaram S. R., 2010, Vegetable oils as a potential cutting ﬂuid-An evolution, Tribology International, 43(5-6):1073–1081, doi:10.1016/j.triboint. 2009.12.065.

[131] Shaw M. C., 1958, Action of Metal Fluids at Low Speeds, Wear, 2:217–227.

[132] Shaw Milton Clayton Cookson J O, 2005, Metal cutting principles, volume 2. Oxford university press New York.

[133] Shi Hanmin, 1999, Chip-ejection interference in cutting processes of modern cutting tools, Science in China, Series E: Technological Sciences, 42(3):275–281, doi:10. 1007/BF02916773.

[134] Shi Hanmin, 2018, Metal Cutting Theory. Springer Series in Advanced Manufacturing. Springer International Publishing, Cham, doi:10.1007/ 978-3-319-73561-0.

[135] Singh Rabesh Kumar, Dixit Amit Rai, Mandal Amitava, Sharma Anuj Kumar, 2017, Emerging application of nanoparticle-enriched cutting ﬂuid in metal removal processes: a review, Journal of the Brazilian Society of Mechanical Sciences and En- gineering, 39(11):4677–4717, doi:10.1007/s40430-017-0839-0.

[136] Smith T., Naerheim Y., Lan M. S., 1988, Theoretical analysis of cutting ﬂuid interaction in machining, Tribology International, 21(5):239–247, doi:10.1016/ 0301-679X(88)90001-1.

[137] Smolenicki D., Boos J., Kuster F., Roelofs H., Wyen C. F., 2014, In-process measurement of friction coeﬃcient in orthogonal cutting, CIRP Annals - Manufacturing Technology, 63(1):97–100, doi:10.1016/j.cirp.2014.03.083.

[138] Sørby K. Tønnessen K., 2006, High-pressure cooling of face-grooving operations in Ti6Al4V, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 220(10):1621–1627, doi:10.1243/09544054JEM474.

[139] Soshko V A Shkarapata Ya. E, sep 1987, Certain relationships of the inﬂuence of polymer-containing cutting ﬂuids on the machining of metals, Soviet materials science : a transl. of Fiziko-khimicheskaya mekhanika materialov / Academy of Sci- ences of the Ukrainian SSR, 22(5):539–540, doi:10.1007/BF00723516. 174 Bibliography

[140] Srikant R. R., Rao D. N., Subrahmanyam M. S., Krishna P. Vamsi, 2009, Applica- bility of cutting ﬂuids with nanoparticle inclusion as coolants in machining, Pro- ceedings of the Institution of Mechanical Engineers, Part J: Journal of Engineering Tribology, 223(2):221–225, doi:10.1243/13506501JET463.

[141] Stabler G. V., 1951, The Fundamental Geometry of Cutting Tools, Proceedings of the Institution of Mechanical Engineers, 165(1):14–26, doi:10.1243/pime_proc_1951_ 165_008_02.

[142] Suda S., Yokota H., Inasaki I., Wakabayashi T., 2002, A synthetic ester as an optimal cutting ﬂuid for minimal quantity lubrication machining, CIRP Annals - Manu- facturing Technology, 51(1):95–98, doi:10.1016/S0007-8506(07)61474-7.

[143] Suresh Kumar Reddy N. Venkateswara Rao P., 2006, Experimental investigation to study the eﬀect of solid lubricants on cutting forces and surface quality in end milling, International Journal of Machine Tools and Manufacture, 46(2):189–198, doi:10.1016/j.ijmachtools.2005.04.008.

[144] Sutter G. List G., 2013, Very high speed cutting of Ti-6Al-4V titanium alloy - Change in morphology and mechanism of chip formation, International Journal of Machine Tools and Manufacture, 66:37–43, doi:10.1016/j.ijmachtools.2012.11.004.

[145] Sutthiruangwong Sutha, Mori Gregor, K¨osters Rolf, mar 2005, Passivity and pseu- dopassivity of cemented carbides, International Journal of Refractory Metals and Hard Materials, 23(2):129–136, doi:10.1016/J.IJRMHM.2004.11.006.

[146] Takeyama H. Murata R., 1963, Basic Investigation of Tool Wear, Journal of En- gineering for Industry, 85(1):33–37, doi:10.1115/1.3667575.

[147] Tobler Val´erie, 2009, Konstruktion eines Tribometers zur Bestimmung von Rei- bungskennwerten auf Basis einer Drehmaschine, Master thesis, ETH Zurich.¨

[148] Tomoda Hideyuki, Sugimoto Yasuhiro, Tani Yumiko, Watanabe Shoji, 1998, Char- acteristic properties of cutting ﬂuid additives derived from the reaction products of hydroxyl fatty acids with some acid anhydrides, Journal of Surfactants and Deter- gents, 1(4):533–537, doi:10.1007/s11743-998-0055-x.

[149] Torr P.H.S. Zisserman A., apr 2000, MLESAC: A New Robust Estimator with Appli- cation to Estimating Image Geometry, Computer Vision and Image Understanding, 78(1):138–156, doi:10.1006/cviu.1999.0832.

[150] Uehara K, 1973, New attempts for short time tool-life testing, Annals of the CIRP, 22(1):23–24. Bibliography 175

[151] Unknown, 2016. Metalworking Fluids Market for Removal, Forming, Protect- ing and Treating Applications - Global Industry Analysis, Size, Share, Growth, Trends and Forecast 2016 - 2024. https://www.transparencymarketresearch. com/metalworking-fluids-market.html (Accessed: 2019-04-01).

[152] Unknown, 2018, The TTT Tapping-Torque-Testsystem as a Multiple Evaluation System for Eﬃciency Ratings of Lubricants, Tapping Tools and Coatings, Technical report, microtap GmbH.

[153] Unknown, 2018. Titanium Metal - Global Industry, Markets & Outlook. https: //roskill.com/market-report/titanium-metal/ (Accessed: 2019-04-02).

[154] Upton D. P., 1996, An assessment of cutting ﬂuid performance and improvements in surface ﬁnish, Transactions of the Institute of Metal Finishing, 74(3):103–105, doi:10.1080/00202967.1996.11871106.

[155] Usui E., Shirakashi T., Kitagawa T., 1984, Analytical prediction of cutting tool wear, Wear, 100(1-3):129–151, doi:10.1016/0043-1648(84)90010-3.

[156] Vasu V. Reddy G. Pradeep Kumar, 2012, Eﬀect of minimum quantity lubrication with Al2O3nanoparticles on surface roughness, tool wear and temperature dissipation in machining Inconel 600 alloy, Proceedings of the Institution of Mechani- cal Engineers, Part N: Journal of Nanoengineering and Nanosystems, 225(1):3–15, doi:10.1177/1740349911427520.

[157] Vasyshak M O Soshko A I, 1986, Influence of polymer additions to cutting fluids in their effectiveness, Materials Science, 22(3):339–340.

[158] Vaughn Robert L., 1966, Modern metals machining technology, Journal of Manu- facturing Science and Engineering, Transactions of the ASME, 88(1):65–71, doi:10. 1115/1.3670894.

[159] Wakabayashi T, Williams J A, Hutchings I M, 1995, The kinetics of gas-phase lubrication in the orthogonal machining of an aluminium alloy, Proceedings of the institution of Mechanical Engineers. J. Eng. Tribol., 209:131–136.

[160] Wang Shibo Clarens Andres F., 2013, Analytical model of metalworking ﬂuid penetration into the ﬂank contact zone in orthogonal cutting, Journal of Manufacturing Processes, 15(1):41–50, doi:10.1016/j.jmapro.2012.09.015.

[161] Wang Z.G., Rahman M., Wong Y.S., feb 2005, Tool wear characteristics of binderless CBN tools used in high-speed milling of titanium alloys, Wear, 258(5-6):752–758, doi:10.1016/j.wear.2004.09.066. 176 Bibliography

[162] Wang Z.G., Wong Y.S., Rahman M., jan 2005, High-speed milling of titanium alloys using binderless CBN tools, International Journal of Machine Tools and Manu- facture, 45(1):105–114, doi:10.1016/j.ijmachtools.2004.06.021.

[163] Warren Anna, Nylund Anders, Olefjord Ingemar, 1996, Oxidation of tungsten and tungsten carbide in dry and humid atmospheres, International Journal of Refrac- tory Metals and Hard Materials, 14(5-6):345–353, doi:10.1016/s0263-4368(96) 00027-3.

[164] Watanabe S., Fujita T., Fukuda S., Hirano K., Sakamoto M., 1986, Characteristic properties as cutting ﬂuid additives of the products from the reaction of unsaturated fatty acids with maleic anhydride, Materials Chemistry and Physics, 15(1):89–96, doi:10.1016/0254-0584(86)90091-X.

[165] Watanabe S., Fujita T., Sakamoto M., Endo J., Kawahara H., 1992, Para- alkoxybenzoic acids as additives in water-based cutting ﬂuids, Journal of Materials Science Letters, 11(8):498–500, doi:10.1007/BF00731117.

[166] Watanabe S., Fujita T., Sakamoto M., Kamaru H., Kawahara H., 1991, Char- acteristic properties of cutting ﬂuid additives derived from fatty alcohols, Journal of the American Oil Chemists Society, 68(1):44–46, doi:10.1007/BF02660308.

[167] Watanabe S., Fujita T., Sakamoto M., Kuramochi T., Nakagawa H., 1993, New water-soluble cutting ﬂuid additives derived from aromatic hydroxycarboxylic acids, Journal of Materials Science, 28(5):1341–1344, doi:10.1007/BF01191975.

[168] Watanabe S, Fujita T, Sakamoto M, Nakagawa H, 1987, New-type water-soluble cutting ﬂuid additives from polymeric surface active agents, Materials chemistry and physics, 16(5-6):575–581.

[169] Watanabe S, Fujita T, Sakamoto M, Nakagawa H, Ohmori Y, 1996, Characteristic properties of cutting ﬂuid additives derived from the products of boric acid and various aminoalcohols, Journal of materials science letters, 15(13):1111–1114.

[170] Watanabe S., Fujita T., Sakamoto M., Nakagawa H., Ohmori Y., 1996, Char- acteristic properties of cutting ﬂuid additives derived from the products of boric acid and various aminoalcohols, Journal of Materials Science Letters, 15(13):1111– 1114, doi:10.1007/BF00539952.

[171] Watanabe S, Fujita T, Suga K, Sugahara K, 1983, Products from the reaction of fatty acids with ketones and their characteristic properties as cutting ﬂuids additives, Journal of the American Oil Chemists’ Society, 60(1):116–119. Bibliography 177

[172] Watanabe S, Fujita T, Yoneshima T, 1985, New additives derived from fatty acids for water-based cutting ﬂuids, Journal of the American Oil Chemists’ Society, 62(1):125–127.

[173] Watanabe S, Fujita T, Yoneshima T, Nakagawa H, 1985, Half esters of dibasic acids as additives in water-based cutting ﬂuids, Journal of the American Oil Chemists’ Society, 62(11):1607–1608.

[174] Watanabe S., Nakagawa H., Ohmori Y., Fujita T., Sakamoto M., 1994, Characteristic properties of cutting ﬂuid additives derived from the derivatives of ricinoleic acid polymers, Journal of the American Oil Chemists’ Society, 71(9):1003–1006, doi:10. 1007/BF02542269.

[175] Watanabe S., Nakagawa H., Ohmori Y., Fujita T., Sakamoto M., Haga T., 1996, Characteristic properties of cutting ﬂuid additives derived from half esters of dimer acids, JAOCS, Journal of the American Oil Chemists’ Society, 73(6):807–809, doi:10.1007/BF02517959.

[176] Watanabe Shoji, Fujita Tsutomu, Suga Kyoichi, Inaba Akio, 1983, New additives for water-based cutting ﬂuids, Materials Chemistry and Physics, 8(6):573–577.

[177] Wildner Klas Jonsson, 2014, Chemical Eﬀects on Formation and Propagation of Comb Cracks in Coated Cemented Carbide Milling Inserts, Degree project, KTH.

[178] Williams J. A. Tabor D., 1977, The role of lubricants in machining, Wear, 43(3):275– 292, doi:10.1016/0043-1648(77)90125-9.

[179] Winter M., Bock R., Herrmann C., 2012, Investigation of a new ecologically benign metalworking ﬂuid in abrasive machining processes to substitute mineral oil based ﬂuids, Procedia CIRP, 1(1):393–398, doi:10.1016/j.procir.2012.04.070.

[180] Winter Marius, Bock Ralf, Herrmann Christoph, 2013, Investigation of a new polymer-water based cutting ﬂuid to substitute mineral oil based ﬂuids in grinding processes, CIRP Journal of Manufacturing Science and Technology, 6(4):254–262, doi:10.1016/j.cirpj.2013.07.003.

[181] Wyen C. F. Wegener K., 2010, Inﬂuence of cutting edge radius on cutting forces in machining titanium, CIRP Annals - Manufacturing Technology, 59(1):93–96, doi:10. 1016/j.cirp.2010.03.056.

[182] Wyen Carl-Frederik, Knapp Wolfgang, Wegener Konrad, apr 2012, A new method for the characterisation of rounded cutting edges, The International Journal of Advanced Manufacturing Technology, 59(9):899–914, doi:10.1007/s00170-011-3555-4. 178 Bibliography

[183] Xavior M. Anthony Adithan M., 2009, Determining the inﬂuence of cutting ﬂuids on tool wear and surface roughness during turning of AISI 304 austenitic stainless steel, Journal of Materials Processing Technology, 209(2):900–909, doi:10.1016/j. jmatprotec.2008.02.068.

[184] Yan Jincan, Zeng Xiangqiong, Ren Tianhui, Van Der Heide Emile, 2014, Bound- ary lubrication of stainless steel and CoCrMo alloy based on phosphorous and boron compounds in oil-in-water emulsion, Applied Surface Science, 315(1):415–424, doi:10.1016/j.apsusc.2014.07.160.

[185] Zakharova E. S., Markova I. Yu, Maslov A. L., Polushin N. I., Laptev A. I., 2017, Morphology of powders of tungsten carbide used in wear-resistant coatings and deposition on the PDC drill bits, Journal of Physics: Conference Series, 857(1), doi:10.1088/1742-6596/857/1/012058.

[186] Zemzemi F, Rech J, Salem W Ben, Dogui A, Kapsa P, 2007, Development of a friction model for the tool-chip-workpiece interfaces during dry machining of AISI4142 steel with TiN coated carbide cutting tools, International Journal of Machining and Machinability of Materials, 2(3/4):361.

[187] Zhang Yanqiao Jun Martin B G, 2015, Feasibility of lignin as additive in metalworking ﬂuids for micro-milling, Journal of Manufacturing Processes, 16(4):503–510, doi:10.1016/j.jmapro.2014.07.001.

[188] Zimmerman Julie B., Takahashi Satsuki, Hayes Kim F., Skerlos Steven J., 2003, Experimental and statistical design considerations for economical evaluation of metalworking ﬂuids using the tapping torque test, Journal of the Society of Tribologists and Lubrication Engineers, 59(3):17. Curriculum Vitae 179

Curriculum Vitae

Personal Data Name: LinusMeier Date of birth: 17 April 1992 Place of birth: Muttenz, Switzerland

School education 2003-2009 State college of higher educations (Kantonsschule Zurcher¨ Oberland)

Academic studies 2009-2012 BSc Mechanical Engineering, ETH Zurich 2012-2014 MSc Mechanical Engineering with specialisation in manufacturing science, ETH Zurich 2014-2019 Doctorate at IWF, ETH Zurich

Occupation 2011-2014: Teaching and research assistant, ETH Zurich 2014-2019: Research associate IWF, ETH Zurich. 180 List of Publications

List of Publications

Conferences

• L. Meier, N. Schaal, K. Wegener, 2017, In-process Measurement of the Coeﬃcient of Friction on Titanium, Procedia CIRP, 58:163-168

• L. Meier, 2018, In process tribology during turning of TiAl6V4, Swiss Tribology Technical Meeting

• L. Meier, K. Wegener, 2019, Constant Power In-Process Tribometry, MTTRF An- nual Meeting

• L. Meier, L. Seeholzer, K. Wegener, 2019, A Generalized Force and Chip Flow Model for Oblique Cutting and Varying Uncut Chip Crosssections, MM Science Journal, 2019(04):3027-2034

Journals

• L. Meier, 2019, Methods to reduce variation in cutting tool life tests, International Journal of Advanced Manufacturing Technology, 103:355-356

• L. Meier, M. Eglin, 2019, Modeling Tool Wear in Titanium Cutting with an In- Process Tribometer, Industrial Lubrication and Tribology, in press