A Novel CAE Method for Compression Molding Simulation of Carbon Fiber-Reinforced Thermoplastic Composite Sheet Materials

Journal of composites science

Article A Novel CAE Method for Compression Molding Simulation of Carbon Fiber-Reinforced Thermoplastic Composite Sheet Materials

Yuyang Song 1,*, Umesh Gandhi 1, Takeshi Sekito 2, Uday K. Vaidya 3, Jim Hsu 4, Anthony Yang 4 and Tim Osswald 5

1 Toyota Research Institute North America, Ann Arbor, MI 48105, USA; [email protected] 2 Material Creation & Analysis Department, Toyota Motor Corporation, Toyota 471-8572, Japan; [email protected] 3 Department of Mechanical, Aerospace and Biomedical Engineering, University of Tennessee, Knoxville, TN 37996, USA; [email protected] 4 Moldex3d Northern America Inc., Farmington Hills, MI 48331, USA; [email protected] (J.H.); [email protected] (A.Y.) 5 Polymer Engineering Center, University of Wisconsin-Madison, Madison, WI 53706, USA; [email protected] * Correspondence: [email protected]

Received: 20 April 2018; Accepted: 30 May 2018; Published: 1 June 2018

Abstract: Its high-specific strength and stiffness with lower cost make discontinuous fiber-reinforced thermoplastic (FRT) materials an ideal choice for lightweight applications in the automotive industry. Compression molding is one of the preferred manufacturing processes for such materials as it offers the opportunity to maintain a longer fiber length and higher volume production. In the past, we have demonstrated that compression molding of FRT in bulk form can be simulated by treating melt flow as a continuum using the conservation of mass and momentum equations. However, the compression molding of such materials in sheet form using a similar approach does not work well. The assumption of melt flow as a continuum does not hold for such deformation processes. To address this challenge, we have developed a novel simulation approach. First, the draping of the sheet was simulated as a structural deformation using the explicit finite element approach. Next, the draped shape was compressed using fluid mechanics equations. The proposed method was verified by building a physical part and comparing the predicted fiber orientation and warpage measurements performed on the physical parts. The developed method and tools are expected to help in expediting the development of FRT parts, which will help achieve lightweight targets in the automotive industry.

Keywords: compression molding; sheet material; computer-aided engineering (CAE); draping; recycled carbon ﬁbers

1. Introduction Fiber-reinforced composites offer exciting new possibilities for weight reduction due to their excellent mechanical properties, economical fuel consumption, and lower carbon footprint. For typical automotive components where large volume, low cost, and fast cycle time are desired, chopped reinforcing ﬁbers with thermoplastic resin are increasing in popularity [1–3]. Usually, such composite materials are made using injection molding, compression molding, or resin transfer molding processes [4,5]. Of these processes, compression molding offers the highest potential to maintain longer ﬁbers and hence better mechanical properties [6–8]; therefore, compression molding

J. Compos. Sci. 2018, 2, 33; doi:10.3390/jcs2020033 www.mdpi.com/journal/jcs J. Compos. Sci. 2018, 2, 33 2 of 16 is the preferredJ. Compos. Sci. method 2018, 2, x FOR by PEER which REVIEW to manufacture fiber-reinforced thermoplastic components2 of 16 in the J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 2 of 16 automotive industry. is the preferred method by which to manufacture fiber-reinforced thermoplastic components in the Thereautomotiveis the are preferred typically industry. method two by different which to manufacture initial formats fiber-reinforced for the fiber-reinforced thermoplastic components chargeused in the in the compressionautomotiveThere molding are industry. typically process, two i.e.,different (1) bulkinitial charge,formats for as the seen fiber-reinforced in Figure1, charge and (2) used sheet in the charge, as per FigurecompressionThere2. Typically, are molding typically the process, two bulk different i.e., charge (1) bulk initial is made charge, formats using as seen for a single/twinthein Figure fiber-reinforced 1, and screw (2) sheetcharge low charge, shear used as plasticator;in perthe compression molding process, i.e., (1) bulk charge, as seen in Figure 1, and (2) sheet charge, as per the desiredFigure size 2. Typically, of the bulk the comingbulk charge out is frommade theusing plasticator a single/twin is collectedscrew low andshear placed plasticator; in the the die for desiredFigure 2.size Typically, of the bulkthe bulk coming charge out isfrom made the using plasticator a single/twin is collected screw and low placedshear plasticator;in the die forthe compression.desired Sheet size of charge the bulk manufacturing coming out from is more the complex.plasticator Chopped is collected fibers and areplaced dispersed in the todie form for mats or spreadcompression. on the resin Sheet sheets charge and manufacturing are then partially is more consolidated. complex. Chopped The sheet fibers charge are dispersed is preferred to form because matscompression. or spread Sheet on the charge resin manufacturingsheets and are thenis more partially complex. consolidated. Chopped The fibers sheet are charge dispersed is preferred to form it startsbecause withmats or longer spreadit starts fibers onwith the and, longer resin since sheetsfibers it coversand,and aresince mostthen it co pa ofversrtially the most mold consolidated. of cavitythe mold surface The cavity sheet tosurface begincharge to with, isbegin preferred the with, material flow requiredthebecause material to it fillstarts flow the with moldrequired longer is shorter; fibersto fill and,the therefore, sincemold itis co shorter; therevers most is therefore, a of better the mold possibilitythere cavity is a surfacebetter of maintaining topossibility begin with, of a longer fiber length.maintainingthe material a flow longer required fiber length. to fill the mold is shorter; therefore, there is a better possibility of maintaining a longer fiber length.

Figure 1. Typical compression molding process using bulk charge. FigureFigure 1. Typical 1. Typical compression compression molding molding process using using bulk bulk charge. charge.

Figure 2. Typical compression molding process using sheet material. FigureFigure 2. Typical 2. Typical compression compression molding molding process using using sheet sheet material. material. As the use of fiber-reinforced polymer components in the industry is increasing, interest in using numericalAs the usesimulation—often of fiber-reinforced called polymer computer-a componentsided in theengineering industry is(CAE)—to increasing, interestaccelerate in using the Asdevelopmentnumerical the use of simulation—often fiber-reinforced process is also increasing. polymercalled computer-aThere components haveided been in numerousengineering the industry research (CAE)—to is increasing, activities accelerate on interest the CAEthe in using numericalsimulationdevelopment simulation—often of fiber-reinforcedprocess is calledalso increasing. polymeric computer-aided resinThere materials have engineering been in numerousthe past (CAE)—to 30 researchyears. Early accelerate activities on, Oswald on the the development [9,10] CAE processandsimulation is also Advani increasing. of [11,12] fiber-reinforced developed There polymeric have the basic been resinformulation numerous materials for researchin thethe compressionpast activities 30 years. molding Early on the on, simulation CAEOswald simulation [9,10] for of and Advani [11,12] developed the basic formulation for the compression molding simulation for fiber-reinforcedfiber-reinforced polymeric polymeric resin materialsmaterials. As in a the result past of 30 their years. work, Early commercial on, Oswald simulation [9,10 ]software and Advani called [11,12] fiber-reinforced polymeric materials. As a result of their work, commercial simulation software called developedCadpress the basic [13] formulationwas developed. for Cadpress the compression used simplified molding Barone–Caulk simulation [14] for flow fiber-reinforced geometry, which polymeric wasCadpress suitable [13] for was sheet developed. moulding Cadpress compound used (SMC) simplified (discontinuous Barone–Caulk fiber + [14] thermoset flow geometry, resin). Wang which et materials. As a result of their work, commercial simulation software called Cadpress [13] was al.was [15] suitable investigated for sheet compression moulding compound molding simulation(SMC) (discontinuous of a multilayer fiber composite+ thermoset with resin). continuous Wang et developed.fibersal. [15] Cadpressand investigated developed used compression a method simplified to moldingestimate Barone–Caulk simulationthe draped [of 14shape a] flowmultilayer and geometry,temperature composite which distribution with wascontinuous suitableof the for sheet mouldinglaminates.fibers and compound Kikuchideveloped [16] (SMC)a studiedmethod (discontinuous the to estimatecompression the fiber drmoldingaped + thermoset shape of woven-fabric and resin).temperature thermoplastic Wang distribution et al. [ 15composite] investigatedof the compressionlaminateslaminates. molding usingKikuchi finite simulation [16] elementstudied of theanalysis, a compression multilayer and de compositemoldingveloped ofan woven-fabric withalgorithm continuous to thermoplastic optimize fibers temperature andcomposite developed laminates using finite element analysis, and developed an algorithm to optimize temperature a methoddistribution to estimate to minimize the draped warpage. shape Three-dimens and temperatureional thermo-viscoplastic distribution of theanalysis laminates. was conducted Kikuchi [16] bydistribution Kim et al. to [17], minimize where thewarpage. rheological Three-dimens characteristicional ofthermo-viscoplastic the SMC sheet material analysis was was modeled conducted and studiedby the Kim compression et al. [17], where molding the rheological of woven-fabric characteristic thermoplastic of the SMC sheet composite material laminateswas modeled using and finite elementthe analysis, effects of and dwelling developed time, and an mold algorithm temperatur to optimizee on mold temperaturefilling and curing distribution were investigated to minimize usingthe effects finite ofelement dwelling analysis time, (FEA) and mold analysis. temperatur e on mold filling and curing were investigated warpage.using Three-dimensionalIn finite summary, element all analysis the previous thermo-viscoplastic (FEA) studies analysis. on the analysissimulation was of sheet conducted materials by for Kim the compression et al. [17], where the rheologicalmoldingIn summary, characteristicprocess focused all the of onprevious the SMC, SMC i.e.,studies sheetdiscontinuous on material the simu thermoset waslation modeled of materials,sheet materials and continuous the for effects the fiber compression of dwellinglaminate time, and moldstructures,molding temperature process or woven focused on fabric mold on with SMC, filling thermoset i.e., and discontinuous or curing thermoplastic were thermoset investigated resins. materials, The majority using continuous of finite research fiber element activitieslaminate analysis structures, or woven fabric with thermoset or thermoplastic resins. The majority of research activities (FEA) analysis.have concentrated only on the thermoforming simulation. Thermoset materials flow well due to low viscosityhave concentrated and therefore only were on the preferred. thermoforming However, simula thermosettion. Thermoset materials materialstake a longer flow time well to due cure to andlow In summary,viscosity and all therefore the previous were preferred. studies However, on the simulation thermoset materials of sheet take materials a longer time for the to cure compression and moldinghence process the cycle focused time onis longer. SMC, i.e.,Therefore, discontinuous thermoset thermosetmaterials are materials, preferred continuousfor low-volume fiber parts. laminate Sincehence wethe arecycle interested time is longer. in automotive Therefore, applications thermoset materialsthat require are apreferred higher volume, for low-volume thermoplastic parts. structures,materialsSince or we woven thatare dointerested fabric not require with in automotive thermoset curing and orapplicationscan thermoplastic be formed that in requirea resins. very shorta Thehigher cycle majority volume, time are of thermoplastic researchdesired. The activities have concentratedmaterials that only do not on require the thermoforming curing and can simulation.be formed in Thermoseta very short materialscycle time floware desired. well due The to low viscosity and therefore were preferred. However, thermoset materials take a longer time to cure and hence the cycle time is longer. Therefore, thermoset materials are preferred for low-volume parts. Since we are interested in automotive applications that require a higher volume, thermoplastic materials J. Compos. Sci. 2018, 2, 33 3 of 16 that do not require curing and can be formed in a very short cycle time are desired. The molten thermoplastic materials are highly viscous and therefore the flow geometry in filling the mold cavity is different when compared to thermoset materials. Consequently, the fiber orientation and fiber lengths are also different. A major challenge in CAE simulation of the discontinuous long fiber thermoplastic parts is the ability to simulate compression molding to predict the fiber orientation and warpage of the final part. For continuous fibers, the fiber orientation can be estimated by using the draping process simulation. Recently, CAE methods to simulate the compression molding process of fiber-reinforced thermoplastic composites using the bulk charge format have been developed and validated with physical tests by Song et al. [18]. However, the compression molding simulation approach developed for bulk materials does not work for the compression molding of sheet materials. This is because, when heated fiber-reinforced thermoplastic sheets are placed in the mold cavity, the sheets have very little stiffness and hence drape easily in the mold cavity under very little compressive force. This draping process involves significant rigid body motion of the sheets. The fluid mechanics-based approach used for compression molding of bulk material [19,20] is therefore not adequate to simulate the large rigid body motions that occur during the draping process. In this paper, a novel CAE method was proposed to address this challenge. The compression molding of sheet materials was simulated in two steps. First, the draping of the heated sheet was simulated using an explicit finite element approach. Such an explicit finite element approach, used for metal sheet stamping simulation, has been proven to be effective for large structural deformation as well as rigid body motions. Next, the draped shape of the fiber-reinforced thermoplastic sheet was used as a prepreg in the compression molding. This is similar to the compression molding of a bulk charge with a given shape, which is an already developed method in our previous work [18]. Furthermore, the proposed approach was demonstrated on a complex 3D shape part. The compression molding simulation was carried out using the proposed two-step approach and the fiber orientation and warpage of the finished part were estimated. To verify the proposed approach, actual parts were made and the measured warpage and fiber orientation results were compared with the predictions. Finally, the effect of the draping distance for the compression molding process was discussed and the ideal draping stroke for the simulation of the given part were estimated. Material properties required to support the simulation, i.e., high temperature stiffness and strength for the draping simulation and material flows and other temperature properties, were measured experimentally.

2. Challenges and Approaches

2.1. Sheet Material Compression Molding Issues Detailed steps for compression molding of sheet materials are presented in Figure3. There are four main steps. Step 1: Sheet material is heated above its melt temperature at 270 ◦C. Step 2: The heated sheet material is draped in the heated mold cavity, which includes a slight compression in the mold. Step 3: Compressing the sheet in the mold and holding at high pressure until solidification. Step 4: Solidified finished part is ejected from the mold and cooled. From the literature review provided in the first section, it was observed that most commercial software can undertake draping analysis for a continuous unidirectional fiber-reinforced thermoset composite, mainly for wrinkling and thickness analysis. For the compression of discontinuous thermoplastic composite materials, a method to simulate the mold filling process, which can estimate fiber orientation and fiber length, is not available. This is mainly because fiber-reinforced thermoplastic sheets need to be draped onto the mold cavity and then compressed; a single computational tool to address both in the same step is not available. This is because of elastic–plastic structural behavior during draping and the flow behavior of resin melt during compression; the integration of structural and fluid flow simulation for a continuum in one physical system are difficult. Therefore, we propose a two-step approach, as shown in Figure4. The entire compression molding of sheet material is divided into two steps. In the first step, the draping analysis is conducted using stamping capability J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 3 of 16

molten thermoplastic materials are highly viscous and therefore the flow geometry in filling the mold cavity is different when compared to thermoset materials. Consequently, the fiber orientation and fiber lengths are also different. A major challenge in CAE simulation of the discontinuous long fiber thermoplastic parts is the ability to simulate compression molding to predict the fiber orientation and warpage of the final part. For continuous fibers, the fiber orientation can be estimated by using the draping process simulation. Recently, CAE methods to simulate the compression molding process of fiber-reinforced thermoplastic composites using the bulk charge format have been developed and validated with physical tests by Song et al. [18]. However, the compression molding simulation approach developed for bulk materials does not work for the compression molding of sheet materials. This is because, when heated fiber-reinforced thermoplastic sheets are placed in the mold cavity, the sheets have very little stiffness and hence drape easily in the mold cavity under very little compressive force. This draping process involves significant rigid body motion of the sheets. The fluid mechanics-based approach used for compression molding of bulk material [19,20] is therefore not adequate to simulate the large rigid body motions that occur during the draping process. In this paper, a novel CAE method was proposed to address this challenge. The compression molding of sheet materials was simulated in two steps. First, the draping of the heated sheet was simulated using an explicit finite element approach. Such an explicit finite element approach, used for metal sheet stamping simulation, has been proven to be effective for large structural deformation as well as rigid body motions. Next, the draped shape of the fiber-reinforced thermoplastic sheet was used as a prepreg in the compression molding. This is similar to the compression molding of a bulk charge with a given shape, which is an already developed method in our previous work [18]. Furthermore, the proposed approach was demonstrated on a complex 3D shape part. The compression molding simulation was carried out using the proposed two-step approach and the fiber orientation and warpage of the finished part were estimated. To verify the proposed approach, actual parts were made and the measured warpage and fiber orientation results were compared with the predictions. Finally, the effect of the draping distance for the compression molding process was discussed and the ideal draping stroke for the simulation of the given part were estimated. Material properties required to support the simulation, i.e., high temperature stiffness and strength for the draping simulation and material flows and other temperature properties, were measured experimentally. J. Compos. Sci. 2018, 2, 33 4 of 16 2. Challenges and Approaches using an2.1. explicit Sheet Material ﬁnite Comp elementression analysis Molding (LS-DYNA Issues V971)Livermore Software Technology Corporation, Livermore,Detailed CA, USA). steps In for the compression second step, molding the drapedof sheet ma partterials is treated are presented as a prepreg in Figure for 3. theThere compression are moldingfour process main steps. using Step mold 1: Sheet ﬁll material analysis is heated (Moldex3D, above its R14) melt CoreTechtemperature System at 270 °C. Co., Step Ltd., 2: The Hsinchu J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 4 of 16 County,heated Taiwan). sheet material A translator is draped to convertin the heated the deformedmold cavity, shape which fromincludes LS-DYNA a slight compression to the initial in the shape for mold. Step 3: Compressing the sheet in the mold and holding at high pressure until solidification. compressionStepthermoplastic 4: Solidified molding composite finished was developed. materials,part is ejected a method from tothe simu moldlate and the cooled. mold filling process, which can estimate fiber orientation and fiber length, is not available. This is mainly because fiber-reinforced thermoplastic sheets need to be draped onto the mold cavity and then compressed; a single computational tool to address both in the same step is not available. This is because of elastic–plastic structural behavior during draping and the flow behavior of resin melt during compression; the integration of structural and fluid flow simulation for a continuum in one physical system are difficult. Therefore, we propose a two-step approach, as shown in Figure 4. The entire compression molding of sheet material is divided into two steps. In the first step, the draping analysis is conducted using stamping capability using an explicit finite element analysis (LS-DYNA V971)Livermore Software Technology Corporation, Livermore, CA, USA). In the second step, the draped part is treated as a prepreg for theFigure compression 3. Challenges molding and issues process when using compression mold fill molding analysis sheet (Moldex3D, materials. R14) CoreTech Figure 3. Challenges and issues when compression molding sheet materials. System Co., Ltd., Hsinchu County, Taiwan). A translator to convert the deformed shape from LS- DYNAFrom to thethe initialliterature shape review for compression provided in molding the first wassection, developed. it was observed that most commercial software can undertake draping analysis for a continuous unidirectional fiber-reinforced thermoset composite, mainly for wrinkling and thickness analysis. For the compression of discontinuous

Figure 4. Approaches used for compression molding of sheet materials. Figure 4. Approaches used for compression molding of sheet materials. 2.2. Measurement of Material Properties of the Sheet Material 2.2. Measurement of Material Properties of the Sheet Material 2.2.1. Sheet Material Properties Measured for Draping Analysis 2.2.1. SheetFor Material accurate Properties draping analysis, Measured temperature-dep for Draping Analysisendent material properties such as elastic Formodulus accurate and draping Poisson’s analysis, ratio are temperature-dependent needed for the sheets. The material sheet materials properties are made such from as elastic a water modulus slurry process, with 35% carbon fiber. The initial resin was made of polyamide 6 (PA6) fibers. Figure and Poisson’s ratio are needed for the sheets. The sheet materials are made from a water slurry process, 5a shows the procedures used to measure the mechanical properties of the sheet material. As shown, with 35%the carbontwo layers fiber. of fluffy The initialsheet materials resin was are madestacked of together polyamide under 6a (PA6)hot press fibers. and consolidated Figure5a shows to the proceduresform usedthe composite to measure plate. the The mechanical tensile bar propertiessamples are of cut the from sheet the material.consolidated As plate. shown, Tests the were two layers of fluffyperformed sheet materials following are the stacked ASTMD638-02 together standard under [21], a hot and press the samples and consolidated were tested toin a form temperature the composite plate. Thechamber tensile to obtain bar samples the mechanical are cut properties from the at consolidateddifferent temperatures. plate. Tests The modulus were performed of the material following the ASTMD638-02was obtained standardfrom the tensile [21], andtest. The the Poisson’s samples ratio were was tested also inmeasured a temperature as per the chamber ASTMD638 to obtain standard [21] using the optical imaging system. Table 1 shows the measured mechanical properties. the mechanical properties at different temperatures. The modulus of the material was obtained from The increase of the Poisson’s ratio in the end is due to the polymeric phase change at around 180 °C, the tensileand crystallization test. The Poisson’s happens ratioagain. wasThe modulus also measured and Poisson’s as per ratio the results ASTMD638 are the average standard of three [21] using the opticalrepeated imaging tests. system. Table1 shows the measured mechanical properties. The increase of the Poisson’s ratio in the end is due to the polymeric phase change at around 180 ◦C, and crystallization happens again. The modulus and Poisson’s ratio results are the average of three repeated tests.

Table 1. Temperature-dependent stress–strain curve for the sheet material.

Temperature (C) Modulus (Mpa) Poisson’s Ratio 23 14,091 0.339 80 10,444 0.239 130 10,118 0.237

180(a) 8728 0.179 (b) 200 7380 0.403 Figure 5. Temperature dependent young’s modulus and Poisson’s ratio measurement procedures. (a) Young’s modulus measurement; (b) Poisson’s ratio measurement. J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 4 of 16

thermoplastic composite materials, a method to simulate the mold filling process, which can estimate fiber orientation and fiber length, is not available. This is mainly because fiber-reinforced thermoplastic sheets need to be draped onto the mold cavity and then compressed; a single computational tool to address both in the same step is not available. This is because of elastic–plastic structural behavior during draping and the flow behavior of resin melt during compression; the integration of structural and fluid flow simulation for a continuum in one physical system are difficult. Therefore, we propose a two-step approach, as shown in Figure 4. The entire compression molding of sheet material is divided into two steps. In the first step, the draping analysis is conducted using stamping capability using an explicit finite element analysis (LS-DYNA V971)Livermore Software Technology Corporation, Livermore, CA, USA). In the second step, the draped part is treated as a prepreg for the compression molding process using mold fill analysis (Moldex3D, R14) CoreTech System Co., Ltd., Hsinchu County, Taiwan). A translator to convert the deformed shape from LS- DYNA to the initial shape for compression molding was developed.

Figure 4. Approaches used for compression molding of sheet materials.

2.2. Measurement of Material Properties of the Sheet Material

2.2.1. Sheet Material Properties Measured for Draping Analysis For accurate draping analysis, temperature-dependent material properties such as elastic modulus and Poisson’s ratio are needed for the sheets. The sheet materials are made from a water slurry process, with 35% carbon fiber. The initial resin was made of polyamide 6 (PA6) fibers. Figure 5a shows the procedures used to measure the mechanical properties of the sheet material. As shown, the two layers of fluffy sheet materials are stacked together under a hot press and consolidated to form the composite plate. The tensile bar samples are cut from the consolidated plate. Tests were performed following the ASTMD638-02 standard [21], and the samples were tested in a temperature chamber to obtain the mechanical properties at different temperatures. The modulus of the material was obtained from the tensile test. The Poisson’s ratio was also measured as per the ASTMD638 standard [21] using the optical imaging system. Table 1 shows the measured mechanical properties. J. Compos.The increase Sci. 2018 ,of2, 33the Poisson’s ratio in the end is due to the polymeric phase change at around 180 °C,5 of 16 and crystallization happens again. The modulus and Poisson’s ratio results are the average of three repeated tests.

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 5 of 16

Table 1. Temperature-dependent stress–strain curve for the sheet material.

Temperature (C) Modulus (Mpa) Poisson’s Ratio (a) (b) 23 14,091 0.339 Figure 5. Temperature dependent80 young’s modulus10,444 and Poisson’s ratio 0.239measurement procedures. Figure 5. Temperature dependent young’s modulus and Poisson’s ratio measurement procedures. (a) Young’s modulus measurement;130 (b) Poisson’s10,118 ratio measurement. 0.237 (a) Young’s modulus measurement; (b) Poisson’s ratio measurement. 180 8728 0.179 200 7380 0.403 2.2.2. Sheet Material Properties Measured for Compression Molding Analysis To2.2.2. accurately Sheet Material simulate Properties the compressionMeasured for Compression molding process, Molding actual Analysis material properties such as pressure–volume–temperatureTo accurately simulate the (PVT), compression viscosity, molding and thermalprocess, actual properties material are properties also needed such foras this processpressure–volume–temperature [22]. The sheet material properties(PVT), viscosity, were measuredand thermal by proper followingties are the also corresponding needed for this ASTM standardsprocess [23 [22].–26 The]. After sheet thematerial physical properties measurement, were measured all the by measurementfollowing the corresponding results were ASTM converted to thestandards moldex3D [23–26]. material After format.the physical These measurement, measured all composite the measurement properties results were were then converted imported to into the moldex3D material format. These measured composite properties were then imported into the the compression molding software Moldex3D’s material database and were applied to the process compression molding software Moldex3D’s material database and were applied to the process simulationsimulation for thefor the part part we we designed, designed, as as Figure Figure6 6 shows. shows.

(a) (b)

Figure 6. Measured charge material properties converted to the Moldex3D format. (a) Viscosity curve; Figure 6. Measured charge material properties converted to the Moldex3D format. (a) Viscosity curve; (b) PVT curve; (c) Heat capacity curve; (d) Viscoelasticity curve. (b) PVT curve; (c) Heat capacity curve; (d) Viscoelasticity curve.

J. Compos. Sci. 2018, 2, 33 6 of 16 J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 6 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 6 of 16 3.3. CAECAE SimulationSimulation 3. CAE Simulation 3.1.3.1. OverallOverall CAECAE SimulationSimulation ConceptConcept andand StepsSteps 3.1. Overall CAE Simulation Concept and Steps BasedBased onon thethe potentialpotential needsneeds forfor thethe light light weighting weighting of of automotive automotive components,components, aa three-cavitythree-cavity tool was Baseddesigned on the and potential the sample needs geom for theetry light is presentedweighting ofin automotive Figure 7. components, a three-cavity tool wastool designedwas designed and and the the sample sample geometry geometry is is presented presented in in Figure Figure 7.7 .

Figure 7. The shape and geometry of the designed part. Figure 7. The shape and geometry of the designed part. Figure 7. The shape and geometry of the designed part. After the initial design of the part, the question was how to use the CAE software to predict the Afterforming the process initial from design sheet of thematerials. part, theAnother question question was of how interest to use was the how CAE to predict software theto warpage predict the After the initial design of the part, the question was how to use the CAE software to predict the formingand process microstructural from sheet details materials. such as Anotherfiber orientation, question fiber of interest length, wasetc. of how the to part predict post-compression the warpage and forming process from sheet materials. Another question of interest was how to predict the warpage microstructuralmolding. As details proposed such in as Section fiber orientation, 2, the integrated fiber length, simulation etc. ofprocess the part combining post-compression LS-DYNA molding.and and microstructural details such as fiber orientation, fiber length, etc. of the part post-compression As proposedMoldex3D in was Section used2 ,for the this integrated part. simulation process combining LS-DYNA and Moldex3D was molding. As proposed in Section 2, the integrated simulation process combining LS-DYNA and used for thisFigure part. 8 shows the overall CAE simulation workflow for the part. The first step was the draping Moldex3Danalysis was in LS-DYNA. used for thisAfter part. the draping analysis, the draped part was transferred to the Moldex3D Figure8 shows the overall CAE simulation workflow for the part. The first step was the draping forFigure compression 8 shows molding the overall process CAE analysis. simulation The detailsworkflow of each for simulationthe part. The process first willstep be was discussed the draping analysis in LS-DYNA. After the draping analysis, the draped part was transferred to the Moldex3D analysisseparately in LS-DYNA. in the following After thesections. draping analysis, the draped part was transferred to the Moldex3D for compression molding process analysis. The details of each simulation process will be discussed for compression molding process analysis. The details of each simulation process will be discussed separately in the following sections. separately in the following sections.

Figure 8. CAE steps for the compression molding of sheet materials.

3.2. Draping Analysis with LS-DYNA For the firstFigureFigure step of 8.8. theCAE CAE simulation,steps stepsfor forthe theLS-DYNAcompression compression was used molding molding to predict of of sheet sheet the materials. materials. draping behavior of the sheet materials. A thermoforming module from LSTC was used in the simulation. The initial sheet 3.2.3.2. DrapingDrapingmaterial Analysis Analysiswas modeled withwith LS-DYNAusingLS-DYNA solid elements, as is shown in Figure 9. In the model, the total number of solid elements in the model was 15,000, the total number of shell elements in the model was 25,896, For the first step of the simulation, LS-DYNA was used to predict the draping behavior of the Forand thethe total first number step of theof nodes simulation, was 49,280. LS-DYNA was used to predict the draping behavior of the sheetsheet materials.materials. AA thermoformingthermoforming modulemodule fromfrom LSTCLSTC waswas usedused inin thethe simulation. simulation. TheThe initialinitial sheetsheet materialmaterial was was modeled modeled using using solid solid elements, elements, as as is is shown shown in in Figure Figure9. In9. theIn the model, model, the the total total number number of solidof solid elements elements in thein the model model was was 15,000, 15,000, the the total total number number of of shell shell elements elements in in the the model model was was 25,896,25,896, andand thethe total total number number of of nodes nodes was was 49,280. 49,280. The basic model information and boundary condition are shown in Figure9. During the thermoforming process, thermal properties of the sheet material and the tool were defined, and the mechanical properties of the sheets were also defined. Due to the draping/tearing of the sheet material at the corners, the adaptive remeshing technique was used in the simulation to reduce the instability J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 7 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 7 of 16 J. Compos. Sci. 2018, 2, 33 7 of 16 or any numerical convergence issues [27,28]. Figure 10 shows the displacement contour of the draped part.J. Compos. Then, Sci. the 2018 shape, 2, x FOR of the PEER draped REVIEW part was exported to Moldex3D for the next simulation step.7 of 16

Figure 9. Draping model information in LS-DYNA. Figure 9. Draping model information in LS-DYNA. The basic model information and boundary condition are shown in Figure 9. During the thermoforming process, thermal properties of the sheet material and the tool were defined, and the The basic model information and boundary condition are shown in Figure 9. During the mechanical properties of the sheets were also defined. Due to the draping/tearing of the sheet thermoforming process, thermal properties of the sheet material and the tool were defined, and the material at the corners, the adaptive remeshing technique was used in the simulation to reduce the mechanical properties of the sheets were also defined. Due to the draping/tearing of the sheet instability or any numerical convergence issues [27,28]. Figure 10 shows the displacement contour material at the corners, the adaptiveFigure 9. Drapingremeshing model technique information was in used LS-DYNA. in the simulation to reduce the of the draped part. Then, theFigure shape 9. Draping of the modeldraped information part was inexported LS-DYNA. to Moldex3D for the next instability or any numerical convergence issues [27,28]. Figure 10 shows the displacement contour simulation step. of theThe draped basic part. model Then, information the shape andof the boundary draped conditionpart was exportedare shown to inMoldex3D Figure 9. for During the next the simulationthermoforming step. process, thermal properties of the sheet material and the tool were defined, and the mechanical properties of the sheets were also defined. Due to the draping/tearing of the sheet material at the corners, the adaptive remeshing technique was used in the simulation to reduce the instability or any numerical convergence issues [27,28]. Figure 10 shows the displacement contour of the draped part. Then, the shape of the draped part was exported to Moldex3D for the next simulation step.

Figure 10. Displacement contours of the draped part (unit: m). Figure 10. Displacement contours of the draped part (unit: m). Figure 10. Displacement contours of the draped part (unit: m). 3.3. Compression Molding Analysis with Moldex3D 3.3. Compression Molding Analysis with Moldex3D 3.3.Based Compression on the Molding cavity designAnalysis from with Moldex3DFigure 6, a compression molding 3D model was built in Moldex3D,Based onas theshown cavity in Figure design 11. from The Figure element6, a compressiontype used in moldingthe model 3D was model tetra was and built hex inmesh; Moldex3D, the Based on the cavity design from Figure 6, a compression molding 3D model was built in astotal shown number in Figure of elements 11. The was element 5,944,826 type and used the in to thetal modelnumber was of tetranodes and was hex 2,942,618. mesh; the There total were number 11 of Moldex3D, as shown in Figure 11. The element type used in the model was tetra and hex mesh; the elementselements wasthrough 5,944,826 thickness. andFigure The the 10. totalgreen Displacement number surface ofon contours nodes the top wasof is the the 2,942,618. draped compression part There (unit: surface, were m). 11 the elements pink area through in total number of elements was 5,944,826 and the total number of nodes was 2,942,618. There were 11 thickness.the middle The is the green compression surface on zone, the topand isthe the gray compression area on the surface, bottom theis the pink cavity. area Figure in the middle12 shows is the elements through thickness. The green surface on the top is the compression surface, the pink area in compressionthe3.3. draped Compression part zone, as Moldingthe and charge the Analysis gray layover area with onthe Moldex3Dthe cavity. bottom The isred the area cavity. on the Figure top is12 assigned shows the as drapedthe charge; part as the middle is the compression zone, and the gray area on the bottom is the cavity. Figure 12 shows the yellow area on the bottom is assigned as the cavity. thethe charge drapedBased layover part on theas the the cavity cavity. charge design The layover red from area the Figure oncavity. the top6, The a is compressionred assigned area on as the themolding top charge; is assigned3D the model yellow as wasthe area charge;built on thein bottomtheMoldex3D, yellow is assigned area as shownon as the the bottomin cavity. Figure is 11.assigned The element as the cavity. type used in the model was tetra and hex mesh; the total number of elements was 5,944,826 and the total number of nodes was 2,942,618. There were 11 elements through thickness. The green surface on the top is the compression surface, the pink area in the middle is the compression zone, and the gray area on the bottom is the cavity. Figure 12 shows the draped part as the charge layover the cavity. The red area on the top is assigned as the charge; the yellow area on the bottom is assigned as the cavity.

Figure 11. Compression molding model built in Moldex3D. Figure 11. Compression molding model built in Moldex3D. Figure 11. Compression molding model built in Moldex3D.

Figure 11. Compression molding model built in Moldex3D.

J. Compos. Sci. 2018, 2, 33 8 of 16 J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 8 of 16

Figure 12. The draped part shown as charge layover the cavity. Figure 12. The draped part shown as charge layover the cavity.

The temperature information and deformed shape were taken into account in the compression moldingThe simulation temperature in Moldex3D. information The and governing deformed equations shape were of the taken fluid into mechanics account inthat the describe compression the transientmolding and simulation non-isothermal in Moldex3D. compression The governing flow motion equations are as follows: of the fluid mechanics that describe the transient and non-isothermal compression flow motion are as follows: ρ (1) ∂∙ρ0ρ + ∇·ρu = 0 (1) ∂t (2) ρ∂ ∙ ρ ρ (ρu) + ∇·(ρuu − σ) = ρg (2) ∂t (3) σσ = −pI +ηη∇u + ∇uT (3) ∂T . 22 (4) ρ ρCP ∙+ u·∇T ∙= ∇·(k∇T) + ηγγ (4) ∂t wherewhere ρ ρis isthe the density; density; u uis isthe the velocity velocity vector; vector; t ist isthe the time; time; σ σis isthe the total total stress stress tensor; tensor; g is g isthe the accelerationacceleration vector vector of ofgravity; gravity; p isp theis thepressure; pressure; η is ηtheis viscosity; the viscosity; Cp is Cthep is specific the specific heat; heat;T is theT is . temperature;the temperature; k is the kthermalis the thermalconductivity; conductivity; and γ is and the γshearis the rate. shear Tait’s rate. pVT Tait’smodelp isVT to modelexpress is a to thermodynamicexpress a thermodynamic state relationship, state relationship, where the where volume the volumeof the ofmaterial the material is a isfunction a function of ofthe the temperature/pressure.temperature/pressure. TheThe power power law law model model is isused used to todescribe describe comple complexx viscosity viscosity behaviors, behaviors, including including the the dramatic dramatic viscosityviscosity change change in ina lower a lower shear shear rate rate range. range. Details Details of ofthe the numerical numerical implementation implementation are are available available elsewhereelsewhere [29] [29 and] and so soare are not not repeated repeated here. here. After After the the part part is isejected ejected from from the the mold, mold, a free a free thermal thermal shrinkageshrinkage occurs occurs due due to thethe temperaturetemperature difference. difference. The The equilibrium equilibrium equation equation representing representing the stress the is: stress is: ∇σ = 0. (5) σ . (5) The relationship between stress and strain is: The relationship between stress and strain is:

1 T σσε= C(ε − αΔ∆T )withwith εε = ∇ U + ∇U (6) (6) CLTE 2 where σ is the stress; C is a fourth-order tensor and a function of the relaxation modulus E; αCLTE is where σ is the stress; C is a fourth-order tensor and a function of the relaxation modulus E; αCLTE is the coefficient of the linear thermal expansion tensor; ε is the strain tensor, which is transferred by the coefﬁcient of the linear thermal expansion tensor; ε is the strain tensor, which is transferred by the the volume shrinkage during molding; and U is the displacement vector, respectively. In order to volume shrinkage during molding; and U is the displacement vector, respectively. In order to consider consider the viscoelastic behavior after molding, a master function for the relaxation modulus is the viscoelastic behavior after molding, a master function for the relaxation modulus is described described using the generalized maxwell model, which is a function of time and temperature with using the generalized maxwell model, which is a function of time and temperature with temperature temperature shift factor aT described below: shift factor aT described below: −C (T−T ) 1 ref (7) ,/ with 10 C +T−T , E = E(Tref, time/aT) with aT = 10 2 ref , (7) where C1, C2 are constant; and Tref is the reference temperature. whereThe Ccompression1, C2 are constant; molding and processTref is the simulations reference temperature.were used in Moldex3D, version R14 [13]. The iARD–RPR (the improved anisotropic rotary diffusion model combined with the retarding principal rate model) was used in the simulation with Ci = 0.01. The other parameters used in the iARD–RPR model were Cm = 0.005, Ci = 0.1, and the fiber matrix interaction alpha factor was 0.7. The processing J. Compos. Sci. 2018, 2, 33 9 of 16

The compression molding process simulations were used in Moldex3D, version R14 [13]. J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 9 of 16 The iARD–RPR (the improved anisotropic rotary diffusion model combined with the retarding principalconditions rate in CAE model) are wasthe same used as in the the one simulation used to make with Cithe= physical 0.01. The part. other Theparameters material properties used in theused iARD–RPR in the simulation model wereare listedCm = in 0.005, TableCi 2= and 0.1, Figure and the 6. ﬁber matrix interaction alpha factor was 0.7. The processing conditions in CAE are the same as the one used to make the physical part. The material properties used in the simulationTable 2. are Material listed properties in Table2 andused Figurein the simulation.6.

Table 2.ItemsMaterial properties used inCarbon the simulation. Fiber PA6 Density (g/cc) 1.78 1.13 Fiber ItemsWeight Percentage Carbon Fiber 35% PA6 N/A Young’sDensity Modulus (g/cc) E1 (Mpa) 1.78 230,000 1.132400 FiberYoung’s Weight Modulus Percentage E2 (Mpa) 35% 23,000 N/A 2400 Young’s Modulus E1 (Mpa) 230,000 2400 Young’s ModulusPoisson’s E2 (Mpa) Ratio 23,000 0.26 2400 0.42 Poisson’sFiber Aspect Ratio Ratio 0.26 400 0.42N/A Fiber Aspect Ratio 400 N/A Fiber CLTE at fiber direction (1/K) 1 × 10−6 N/A Fiber CLTE at ﬁber direction (1/K) 1 × 10−6 N/A FiberFiber CLTE CLTE at transverse at transverse direction direction (1/K) (1/K) 1 × 10− 15 × 10−5 N/AN/A PolymerPolymer CLTE CLTE (1/K) (1/K) N/A N/A 8.3 8.3× ×10 10−5−5

4. Experimental Procedure

4.1. Carbon Fiber Sheet Material Manufacturing The first first step in in the the production production of of fiber-reinforc fiber-reinforceded thermoplastic thermoplastic composites composites is isthe the combining combining of ofthe the fiber fiber reinforcement reinforcement with with the thematrix matrix resin resin to fo torm form the sheet the sheet material. material. One Onetechnique technique that has that been has beendemonstrated demonstrated to produce to produce a high-qua a high-qualitylity self-supporting self-supporting preform preform is the wet-lay is the wet-lay paper making paper makingprocess process[30–32]. [In30 this–32]. process, In this process,the reinforcing the reinforcing fibers an fibersd thermoplastic and thermoplastic polymer polymerfibers are fibers dispersed are dispersed in water inalong water with along bonding with bondingagents. This agents. creates This a slurry creates th aat slurry is cast that onto is casta moving onto aforming moving wire forming and wirethen anddewatered, then dewatered, leaving a nonwoven leaving a nonwovenmat of thermoplasti mat of thermoplasticc and reinforcing and fibers. reinforcing The carbon fibers. fibers, The carbon which fibers,are made which from are recycled made from carbon recycled fiber carboncomposites, fiber composites,were supplied were by suppliedCarbon Conversion by Carbon Conversion(Lake City, (LakeSC, USA). City, The SC, carbon USA). fiber The carbonmats have fiber a random mats have fiber a orientation, random fiber and orientation, an average and fiber an length average of about fiber length12 mm, of and about diameter 12 mm, about and diameter 6–7 µm. aboutThe resin 6–7 µwasm. ThePA6 resin fibers was with PA6 an fibers average with length an average of about length 10 ofmm, about and 10 diameter mm, and of diameter 20 µm. Each of 20 sheetµm. Eachwith sheeta thickness with a of thickness about 0.2 of aboutmm was 0.2 stacked mm was to stacked form a to 2 formmm nominal a 2 mm nominalthickness thickness sheet. The sheet. mats The were mats fluffy were and fluffy had and in-plane had in-plane random random fiber fiberorientation. orientation. The Thesize sizeof the of mats the mats we used we used in this in thisstudy study were were 355 355mm mm × 254× mm.254 mm. Figure Figure 13 shows 13 shows the steps the steps used used in the in thewater water slurry slurry process process to form to form the thesheet sheet materials. materials.

Figure 13. Sheet material manufacturing process using water slurry method. Figure 13. Sheet material manufacturing process using water slurry method.

4.2. Actual Part Manufacturing 4.2. Actual Part Manufacturing Three cavity parts were built using carbon fiber (CF) (35%) + PA6 sheet material. The basic Three cavity parts were built using carbon ﬁber (CF) (35%) + PA6 sheet material. The basic procedure for the compression molding of the part is shown in Figure 14. After the sheet material procedure for the compression molding of the part is shown in Figure 14. After the sheet material was was prepared from the water slurry process, it was moved to the infrared oven, and heated to 343 °C prepared from the water slurry process, it was moved to the infrared oven, and heated to 343 ◦C before before it was moved to the press. it was moved to the press.

J. Compos. Sci. 2018, 2, 33 10 of 16 J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 10 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 10 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 10 of 16 FigureFigure 14. 14.Three-cavity Three-cavity tool tool and and the the partpart made using using sheet sheet material. material.

Table 3 shows the processing conditions of the manufacturing process. After the sample is Table3 shows the processing conditions of the manufacturing process. After the sample is demolded and cleaned,Figure it 14.is readyThree-cavity for warpage tool and and the partfiber made orientation using sheet measurement. material. Figure 15 shows demoldedthe warpage and cleaned, measurement it is ready method for warpage used for and the ﬁber three-cavity orientation tool. measurement. The FARO-Arm Figure (FARO 15 shows the warpageTechnologies,Table measurement 3 shows Inc., Lakethe method processing Mary, usedFL, conditionsUSA) for the [33,34], three-cavity of thewhich manufacturing is tool. a state-of-the-art The FARO-Armprocess. non-contact After (FARO the sample Technologies,scanning is Inc., Lakedemoldeddevice Mary, for andwarpage FL, cleaned, USA) measurement, [ 33it is,34 ready], which forwas warpage isused a state-of-the-art to collectand fiber the orientation point non-contact cloud measurement. of the scanning final part. Figure device 15 for shows warpage measurement,the warpage was measurement used to collect method the point used cloud for ofthe the three-cavity ﬁnal part. tool. The FARO-Arm (FARO

TheTechnologies, scanned pointInc., Lake cloud Mary, wasTable thenFL, 3. CompressionUSA) superimposed [33,34], molding which with parameters is the a state-of-the-art original used. computer-aided non-contact designscanning (CAD) device for warpage Figuremeasurement, 14. Three-cavity was used tool toand collect the part the made point using cloud sheet of thematerial. final part. data of the part, as shown. From theProcess superimposed Conditions contour, Actual the Values warpage information was obtained. Computed tomography (CT) scanning and volume graphics (VGSTUDIO MAX, Volume Graphics Table 3 shows the processingTableMelt 3. Compressionconditions Temperature ofmolding the manufacturing parameters270 °C used. process. After the sample is GmbH, Heidelberg, Germany) analysisMold [ 13Temperature,35] were used for70 measuring °C the ﬁber orientation. For ﬁber demolded and cleaned, it is readyProcess for warpage Conditions and fiber Actual orientation Values measurement. Figure 15 shows orientationthe warpage tensor analysis,measurement the partmethodCompression was cutused from for time thethe molded three-cavity60 part s attool. locations The FARO-Arm A, B, C, etc., (FARO as shown Melt Temperature 270 °C in FigureTechnologies, 16. The sample Inc., Lake size Mary, usedCompression FL, was USA) 10 mm [33,34], pressure× 10 which mm × is20002 a mm,state-of-the-art KN and the resolution non-contact used scanning for the CT Mold Temperature 70 °C device for warpageµ measurement, Chargewas used weight to collect the point150 gmcloud of the final part. scanning was 1 m. Compression time 60 s Compression pressure 2000 KN TableTable 3. 3.Compression Compressionmolding molding parameters used. used. Charge weight 150 gm Process Conditions Actual Values Process Conditions Actual Values Melt Temperature 270 °C Melt Temperature 270 ◦C Mold Temperature 70 °C Mold Temperature 70 ◦C CompressionCompression time time 60 s60 s CompressionCompression pressure pressure 20002000 KN KN ChargeCharge weight weight 150150 gm gm Figure 15. Warpage measurement for the final part.

The scanned point cloud was then superimposed with the original computer-aided design (CAD) data of the part, asFigure shown. 15. FromWarpage the measurement superimposed for thecontour, final part. the warpage information was obtained. Computed tomography (CT) scanning and volume graphics (VGSTUDIO MAX, Volume GraphicsThe scannedGmbH, pointHeidelberg, cloud wasGermany) then superimposed analysis [13,35] with were the originalused for computer-aided measuring the design fiber (CAD)orientation. data Forof the fiber part, orientation as shown. tensor From analysis, the superimposed the part was contour, cut from the the warpage molded information part at locations was obtained.A, B, C, etc., Computed as shown tomography in Figure 16.(CT) The scanning sample and size volume used was graphics 10 mm (VGSTUDIO × 10 mm × 2MAX, mm, Volumeand the

Graphicsresolution GmbH,used for Heidelberg,the CT scanning Germany) was 1 µm.analysis [13,35] were used for measuring the fiber orientation. For fiber orientationFigure tensor15. Warpage analysis, measurement the part wasfor the cut final from part. the molded part at locations A, B, C, etc., as shown inFigure Figure 15. 16.Warpage The sa measurementmple size used for was the 10 ﬁnal mm part. × 10 mm × 2 mm, and the resolutionThe scanned used for point the CT cloud scanning was wasthen 1 superimposedµm. with the original computer-aided design (CAD) data of the part, as shown. From the superimposed contour, the warpage information was obtained. Computed tomography (CT) scanning and volume graphics (VGSTUDIO MAX, Volume Graphics GmbH, Heidelberg, Germany) analysis [13,35] were used for measuring the fiber orientation. For fiber orientation tensor analysis, the part was cut from the molded part at locations A, B, C, etc., as shown in Figure 16. The sample size used was 10 mm × 10 mm × 2 mm, and the resolution used for the CT scanning was 1 µm.

Figure 16. Fiber orientation measurement using CT scan and volume graphic method.

Figure 16. Fiber orientation measurement using CT scan and volume graphic method. Figure 16. Fiber orientation measurement using CT scan and volume graphic method.

Figure 16. Fiber orientation measurement using CT scan and volume graphic method. J. Compos. Sci. 2018, 2, 33 11 of 16

5. ResultsJ. Compos. Sci. 2018, 2, x FOR PEER REVIEW 11 of 16

5.1. Warpage5.J. Compos.Results Sci. Comparison 2018, 2, x FOR PEER REVIEW 11 of 16 5.1.Figure5. Results Warpage 17 shows Comparison the warpage measurement location from both the measurement and CAE predication.Figure The 17 location shows the of thewarpage point measurement was the same location for both from cases both except the measurement that the notation and CAE used for 5.1. Warpage Comparison one waspredication. B and theThe otherlocation one of wasthe point A. After was the the same FARO-ARM for both cases laser except scanning, that the the notation scanned used data for were superimposedone wasFigure B and with 17 theshows the other original the one warpage was CAD A. designmeasurementAfter the and FA the RO-ARMlocation post-processing fromlaser bothscanning, the analysis measurementthe scanned was conducted data and wereCAE using Polyworksuperimposedpredication. viewer The (2016,with location the Innovmetric original of the CADpoint Software designwas the and Inc., same the Qu forpost-processinébec, both QC, cases Canada) exceptg analysis that [36 ]. wasthe At notationconducted location used B, using the for green line wasPolyworkone thewas warpedB viewer and the (2016, condition, other Innovmetric one whilewas A. Software the After gray the Inc., line FA QuRO-ARM wasébec, the QC, initiallaser Canada) scanning, geometry. [36]. Atthe location Similarly, scanned B, datathe at locationgreen were A, superimposed with the original CAD design and the post-processing analysis was conducted using the grayline was line the was warped the original condition, geometry, while the and gray the line colored was the contour initial wasgeometry. the warped Similarly, shape. at location The warpage A, thePolywork gray line viewer was the (2016, original Innovmetric geometry, Software and the Inc., colored Québec, contour QC, wasCanada) the warped [36]. At shape. location The B, warpagethe green comparison between the CAE prediction at location A and measurement at location B is shown in comparisonline was the between warped condition,the CAE prediction while the atgray location line was A andthe initialmeasurement geometry. at Similarly,location B at is location shown inA, Figure 17a. The effect of mold temperature to the warpage is shown in Figure 17b. It can be seen that Figurethe gray 17a. line The was effect the originalof mold geometry,temperature and to the the colored warpage contour is shown was in the Figure warped 17b. shape. It can The be seen warpage that withwithcomparison an increase an increase between in thein the moldthe mold CAE temperature, temperature, prediction at thethe location warpage A and increased increased measurement in inproportion. proportion. at location The B Theoverall is shown overall error in error betweenbetweenFigure the 17a. simulationthe The simulation effect and of and mold measurement measurem temperatureent waswas to the noted warpage to to be be aroundis around shown 8%. in 8%. Figure 17b. It can be seen that with an increase in the mold temperature, the warpage increased in proportion. The overall error between the simulation and measurement was noted to be around 8%.

(a) (b)

Figure 17. Effect of mold temperature to warpage. ( a) Warpage measurement locations; (b) Warpage Figure 17. Effect of mold(a) temperature to warpage. (a) Warpage(b measurement) locations; comparisons. (b) WarpageFigure 17. comparisons. Effect of mold temperature to warpage. (a) Warpage measurement locations; (b) Warpage 5.2. Fibercomparisons. Orientation Comparison 5.2. Fiber Orientation Comparison The key to successful simulation is the ability to simulate the physics of the event. For the fiber- 5.2. Fiber Orientation Comparison reinforcedThe key toanalysis, successful the fiber simulation orientation is plays the a ability key role to in simulate structuralthe properties physics of ofthe thepartevent. and hence For the fiber-reinforcedthe warpage.The key analysis, to successful the fibersimulation orientation is the ability plays to a simulate key role the in structuralphysics of the properties event. For of the the fiber- part and reinforced analysis, the fiber orientation plays a key role in structural properties of the part and hence hence theFigure warpage. 18 shows the fiber orientation measurement results for location A as an overview. It is the warpage. clearlyFigure shown 18 shows that all the the fiber tensors orientation added up measurement to 1. results for location A as an overview. It is Figure 18 shows the fiber orientation measurement results for location A as an overview. It is clearlyclearly shown shown that that all theall the tensors tensors added added up up to to 1.1.

Figure 18. Fiber orientation tensors in all three principal directions measured at location A.

Figure 18. Fiber orientation tensors in all three principal directions measured at location A. Figure 18. Fiber orientation tensors in all three principal directions measured at location A.

J. Compos. Sci. 2018, 2, 33 12 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 12 of 16 We also compared the fiber orientation at selected locations such as A, B, and C (see Figure 19). We also compared the fiber orientation at selected locations such as A, B, and C (see Figure 19). A11 is usually the direction of the fiber along the X direction, as shown in Figure 19. The comparison A11 is usually the direction of the fiber along the X direction, as shown in Figure 19. The comparison of theof the fiber fiber orientation orientation tensortensor A11 A11 between between the the pred predictioniction and measurement and measurement at these atlocations these is locations also is alsoshown. shown.

(a) (b)

Figure 19. Comparison of fiber orientation in locations A, B, and C for A11. (a) Sample size and part Figure 19. Comparison of fiber orientation in locations A, B, and C for A11. (a) Sample size and part information; (b) A11 at location A; (c) A11 at location B; (d) A11 at location C. information; (b) A11 at location A; (c) A11 at location B; (d) A11 at location C. According to the Department of Energy (DOE) findings, the discrepancy between the simulation andAccording measurement to the was Department within the of 15% Energy accuracy (DOE) criterion findings, [37]. theComparison discrepancy of the between fiber orientation the simulation andresults measurement at other waslocations within such the as 15% D, E, accuracy and F al criterionso showed [37 trends]. Comparison similar to ofA, theB, and fiber C. orientation The resultscomparison at other locationsresults are such not aspresented D, E, and here F alsofor brevity. showed As trends a reference, similar Table to A, B,4 shows and C. the The analytic comparison resultscomputed are not Young’s presented modulus here for E11 brevity. based Ason afiber reference, orientation Table from4 shows the thepredicted analytic vs. computed measured Young’sat locations A, B, and C. Such an analysis was based on the Mori–Tanaka Mean Field homogenization modulus E11 based on fiber orientation from the predicted vs. measured at locations A, B, and C. Such scheme [38]. It is clear that the E11 value showed less than a 10% difference between these two. an analysis was based on the Mori–Tanaka Mean Field homogenization scheme [38]. It is clear that the E11 value showedTable 4. Comparison less than a of 10% the analytic difference modulus between based on these two two.different fiber orientation tensors.

E11 (MPa) TableLocation/Modulus 4. Comparison of the analytic modulus based on two different fiber orientation tensors. Using Predicted Fiber Orientation Using Measured Fiber Orientation Agreement A 23,855 E11 (MPa) 21,972 8.6% Location/Modulus B Using Predicted23,252 Fiber Orientation Using Measured 25,524 Fiber Orientation8.9% Agreement C 22,791 24,353 6.4% A 23,855 21,972 8.6% B 23,252 25,524 8.9% 5.3. FiberC Length Comparison 22,791 24,353 6.4% Figure 20 shows the fiber length distribution comparison. The fiber lengths of the parts were 5.3.measured Fiber Length using Comparison the FASEP method [39]. In the compression molding process simulation, the fiber breakage calculation is used to obtain the fiber length distribution. The fiber breakage algorithm was Figure 20 shows the fiber length distribution comparison. The fiber lengths of the parts were measured using the FASEP method [39]. In the compression molding process simulation, the fiber breakage calculation is used to obtain the fiber length distribution. The fiber breakage algorithm was J. Compos. Sci. 2018, 2, 33 13 of 16

J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 13 of 16 developed from the Coretech company (Hsinchu County, Taiwan) and implemented in Moldex3D developedsoftware, so from the algorithmthe Coretech is available company anywhere (Hsinchu [ 40County,]. The comparisonTaiwan) and shows implemented that the ﬁberin Moldex3D length of software,the ﬁnal part so the is withinalgorithm the rangeis available of 3–4 a mm.nywhere [40]. The comparison shows that the fiber length of

(a)

(b)

Figure 20. Fiber length distribution comparison. (a) sample locations; (b) fiber length distribution at 6 locations. Figure 20. Fiber length distribution comparison. (a) sample locations; (b) ﬁber length distribution at 6 locations. 5.4. Stroke Distance Effect

5.4. StrokeThere Distanceis one issue/concern Effect in the compression molding of sheet materials using the proposed approach, i.e., when should the draping stop and how does a partially draped part affect the There is one issue/concern in the compression molding of sheet materials using the proposed warpage? In production, the compression molding process occurs continually at the established cycle approach, i.e., when should the draping stop and how does a partially draped part affect the warpage? times, typically 2–3 min depending on the part size. To address this issue, the LS-DYNA analysis was In production, the compression molding process occurs continually at the established cycle times, conducted considering three different boundary conditions. As shown in Figure 21, the draping typically 2–3 min depending on the part size. To address this issue, the LS-DYNA analysis was distance can be 100% of the gap between the top punch and the bottom die, or a fraction, i.e., 60%, conducted considering three different boundary conditions. As shown in Figure 21, the draping 80%, etc. The draped part is then transferred to moldex3D for compression molding simulation. distance can be 100% of the gap between the top punch and the bottom die, or a fraction, i.e., 60%, 80%, etc. The draped part is then transferred to moldex3D for compression molding simulation. It is shown that the draping distance has an effect on the prepreg shape; therefore, the ﬁnal warpage of the part is different for each case. From the comparison of warpage measurement results, it was seen that 80% of the stroke distance was best matched with the actual measurement results. When the mold temperature was 140 ◦C, similar CAE simulations were performed. It is shown that, for 80% of the draping distance, the warpage had matching results between CAE and measurement.

(a) (b)

Figure 21. Effect of the stroke to the warpage. (a) details about stroke; (b) warpage comparison. J. Compos. Sci. 2018, 2, x FOR PEER REVIEW 13 of 16 developed from the Coretech company (Hsinchu County, Taiwan) and implemented in Moldex3D software, so the algorithm is available anywhere [40]. The comparison shows that the fiber length of the final part is within the range of 3–4 mm.

(a)

(b)

Figure 20. Fiber length distribution comparison. (a) sample locations; (b) fiber length distribution at 6 locations.

5.4. Stroke Distance Effect There is one issue/concern in the compression molding of sheet materials using the proposed approach, i.e., when should the draping stop and how does a partially draped part affect the warpage? In production, the compression molding process occurs continually at the established cycle times, typically 2–3 min depending on the part size. To address this issue, the LS-DYNA analysis was conducted considering three different boundary conditions. As shown in Figure 21, the draping distanceJ. Compos. Sci.can2018 be, 2100%, 33 of the gap between the top punch and the bottom die, or a fraction, i.e.,14 60%, of 16 80%, etc. The draped part is then transferred to moldex3D for compression molding simulation.

(a) (b)

Figure 21. Effect of the stroke to the warpage. (a) details about stroke; (b) warpage comparison. Figure 21. Effect of the stroke to the warpage. (a) details about stroke; (b) warpage comparison.

6. Conclusions A novel two-step CAE method to simulate the compression molding of chopped carbon fiber-reinforced thermoplastic sheet material was developed. First, the draping simulation was carried out to model the draping of a heated carbon thermoplastic sheet on the mold cavity with very little pressure. Next, the draped part was transferred as a prepreg for a further compression molding process. The two-step approach helps with simulating the complex process whereby a solid elastic–plastic material is thermoformed and then compressed to flow like a highly viscous fluid. The effect of the draping distance to the final warpage was discussed and the ideal draping stroke for the given component was demonstrated. The predicted warpage, fiber length, and fiber orientation for the parts made from sheet material using the proposed approach were compared with the measurements from the physical parts. The comparison showed an acceptable match between the simulation and the actual measurements. In this study, it was assumed that the fibers in the sheet form were of in-plane random orientation, and, after draping, they remained as in-plane randomly oriented fibers. The research findings based on these models will benefit lightweight composite parts/components development for the automotive industry.

Author Contributions: U.G. conceived the idea, designed the project and helped with the manuscript. Y.S. conducted the experiment and simulation and prepared the data and manuscript. U.K.V. supported the test facility for the experiment. Tim Osswald gave academic suggestions. J.H. and A.Y. gave advice on the simulation and helped improving the software. T.S. helped with the experiment and provided insights into actual parts concerns. Conﬂicts of Interest: The authors declare no competing interests.

References

1. Isayev, A. Injection and Compression Molding Fundamentals; CRC Press: Boca Raton, FL, USA, 1987. 2. Beardmore, P.; Harwood, J.J.; Kinsman, K.R.; Robertson, R.E. Fiber-reinforced composites engineered structural materials. Science 1980, 208, 832–840. 3. Biron, M. Thermoplastics and Thermoplastic Composites; William Andrew: Norwich, NY, USA, 2012. 4. Long, A.C. (Ed.) Composites Forming Technologies; Woodhead Publishing: Cambridge, UK, 2014. 5. Baird, D.G.; Collias, D.I. Polymer Processing: Principles and Design; John Wiley & Sons: Hoboken, NJ, USA, 2014. 6. Dumont, P.; Orgéas, L.; Favier, D.; Pizette, P.; Venet, C. Compression moulding of SMC: In situ experiments, modelling and simulation. Compos. Part A Appl. Sci. Manuf. 2007, 38, 353–368. [CrossRef] 7. Miura, M.; Hayashi, K.; Yoshimoto, K.; Katahira, N. Development of Thermoplastic CFRP for Stack Frame. In SAE Technical Paper; SAE International: Detroit, MI, USA, 2016. J. Compos. Sci. 2018, 2, 33 15 of 16

8. Bruce, A.D.; Oswald, T.A. Compression Molding; Hanser Publications: Cincinnati, OH, USA, 2003. 9. Osswald, T.A.; Tucker, C.L. A boundary element simulation of compression mold filling. Polym. Eng. Sci. 1988, 28, 413–420. [CrossRef] 10. Oswald, T.A. Numerical Models for Compression Mold Filling Simulation. Ph.D. Thesis, University of Illinois, Champaign, IL, USA, 1987. 11. Advani, S.G.; Tucker, C.L. A numerical simulation of short fiber orientation in compression molding. Polym. Compos. 1990, 11, 164–173. [CrossRef] 12. Sozer, E.M.; Advani, S.G. Process Modeling in Composites Manufacturing; CRC Press: Boca Raton, FL, USA, 2010. 13. Rios, A.; Davis, B.; Gramann, P. Computer Aided Engineering in Compression Molding. In Proceedings of the Composites Fabricators Association, Tampa, FL, USA, 3–6 October 2001. 14. Barone, M.R.; Caulk, D.A. A model for the flow of a chopped fiber reinforced polymer compound in compression molding. J. Appl. Mech. 1986, 53, 361–371. [CrossRef] 15. Wang, P.; Hamila, N.; Boisse, P. Thermoforming simulation of multilayer composites with continuous fibres and thermoplastic matrix. Compos. Part B Eng. 2013, 52, 127–136. [CrossRef] 16. Hsiao, S.W.; Kikuchi, N. Numerical analysis and optimal design of composite thermoforming process. Comput. Methods Appl. Mech. Eng. 1999, 177, 1–34. [CrossRef] 17. Okine, R.K. Analysis of forming parts from advanced thermoplastic composite sheet materials. J. Thermoplast. Compos. Mater. 1989, 2, 50–76. [CrossRef] 18. Song, Y.; Gandhi, U.; Pérez, C.; Osswald, T.; Vallury, S.; Yang, A. Method to account for the fiber orientation of the initial charge on the fiber orientation of finished part in compression molding simulation. Compos. Part A Appl. Sci. Manuf. 2017, 100, 244–254. [CrossRef] 19. Manual of Moldex3D Software; CoreTech. Inc.: Taipei, Taiwan, 2014. 20. Manual of Moldflow Software; Autodesk. Inc.: New York, NY, USA, 2014. 21. ASTM D638-02a, Standard Test Method for Tensile Properties of Plastics, West Conshohocken, PA, USA. 2002. Available online: https://www.astm.org/DATABASE.CART/HISTORICAL/D638-02A.htm (accessed on 1 June 2018). 22. Eberle, A.P.; Baird, D.G.; Wapperom, P. Rheology of non-Newtonian fluids containing glass fibers: A review of experimental literature. Ind. Eng. Chem. Res. 2008, 47, 3470–3488. [CrossRef] 23. International Organization for Standardization. ISO 17 744:2004, Plastics—Determination of Specific Volume as a Function of Temperature and Pressure; International Organization for Standardization: Geneva, Switzerland, 2004. 24. ASTM D4440-15, Standard Test Method for Plastics: Dynamic Mechanical Properties Melt Rheology, West Conshohocken, PA, USA. 2015. Available online: https://www.astm.org/Standards/D4440.htm (accessed on 1 June 2018). 25. ASTM D5930-16, Standard Test Method for Thermal Conductivity of Plastics by Means of a Transient Line-Source Technique, West Conshohocken, PA, USA. 2016. Available online: https://www.astm.org/ DATABASE.CART/HISTORICAL/D5930-16.htm (accessed on 1 June 2018). 26. ASTM E1269-11, Standard Test Method for Determining Specific Heat Capacity Differential Scanning Calorimetry, West Conshohocken, PA, USA. 2011. Available online: https://www.astm.org/Standards/ E1269.htm (accessed on 1 June 2018). 27. Plewa, T.; Linde, T.; Weirs, V.G. Adaptive mesh refinement-theory and applications. Lect. Notes Comput. Sci. Eng. 2005, 41, 3–5. 28. Dong, Y.; Lin, R.J.T.; Bhattacharyya, D. Finite element simulation on thermoforming acrylic sheets using dynamic explicit method. Polym. Polym. Compos. 2006, 14, 307–328. 29. Lin, G.; Lai, D.L.; Huang, C.T.; Wang, C.C. Numerical Simulation for the Viscoelastic Effects on the Birefringence Variation for an Injected Optical Lens. SPE ANTEC Indianap. 2016, 2016, 1231–1235. 30. Huang, J.; Baird, D.G.; McGrath, J.E. Development of fuel cell bipolar plates from graphite filled wet-lay thermoplastic composite materials. J. Power Sources 2005, 150, 110–119. [CrossRef] 31. Loos, A.C. Processing of Composites; Hanser Publishers: Munich, Germany, 2000; p. 320. 32. Loos, A.C. Low-cost fabrication of advanced polymeric composites by resin infusion processes. Adv. Compos. Mater. 2001, 10, 99–106. [CrossRef] 33. Faro Arm Measuring System. Available online: http://www.faro.com/en-us/products/metrology/ measuring-arm-faroarm/overview (accessed on 1 June 2018). J. Compos. Sci. 2018, 2, 33 16 of 16

34. Verma, K.; Columbus, D.; Han, B. Development of real time/variable sensitivity warpage measurement technique and its application to plastic ball grid array package. IEEE Trans. Electron. Packag. Manuf. 1999, 22, 63–70. [CrossRef] 35. Manual of Volume Graphics Software; Volume Graphics GmbH: Heidelberg, Germany, 2014. 36. Ebewele, R.O. Polymer Science and Technology; CRC Press: Boca Raton, FL, USA, 2000. 37. Nguyen, B.N.; Fifield, L.S.; Gandhi, U.N.; Mori, S.; Wollan, E.J. Predictive Engineering Tools for Injection-Molded Long-Carbon-Thermoplastic Composites: Weight and Cost Analysis (No. PNNL-25646); Pacific Northwest National Lab. (PNNL): Richland, WA, USA, 2016. 38. Papathanasiou, T.D.; Guell, D.C. Flow-Induced Alignment in Composite Materials; Woodhead: Cambridge, UK, 1997. 39. FASEP Fiber Length Measurement System. Available online: http://www.fasep.biz/ (accessed on 1 June 2018). 40. Huang, C.T.; Tseng, H.C.; Vlcek, J.; Chang, R.Y. Fiber breakage phenomena in long fiber reinforced plastic preparation. Conf. Ser. Mater. Sci. Eng. 2015, 87, 012023. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).