Structural Health Monitoring of Walking Dragline Excavator Using Acoustic Emission

applied sciences

Article Structural Health Monitoring of Walking Dragline Excavator Using Acoustic Emission

Vera Barat 1,2, Artem Marchenkov 1,* , Dmitry Kritskiy 3, Vladimir Bardakov 1,2, Marina Karpova 1, Mikhail Kuznetsov 1, Anastasia Zaprudnova 1, Sergey Ushanov 1,2 and Sergey Elizarov 2

1 Moscow Power Engineering Institute, 14, Krasnokazarmennaya Str., 111250 Moscow, Russia; [email protected] (V.B.); [email protected] (V.B.); [email protected] (M.K.); [email protected] (M.K.); [email protected] (A.Z.); [email protected] (S.U.) 2 LLC “Interunis-IT”, 20b, Entuziastov Sh., 111024 Moscow, Russia; [email protected] 3 JSC “SUEK”: 20b, Lenin Str., 660049 Krasnoyarsk, Russia; [email protected] * Correspondence: [email protected]

Abstract: The article is devoted to the organization of the structural health monitoring of a walking dragline excavator using the acoustic emission (AE) method. Since the dragline excavator under study is a large and noisy industrial facility, preliminary prospecting researches were carried out to conduct effective control by the AE method, including the study of AE sources, AE waveguide, and noise parameters analysis. In addition, AE filtering methods were improved. It is shown that application of the developed filtering algorithms allows to detect AE impulses from cracks and defects against a background noise exceeding the useful signal in amplitude and intensity. Using the proposed solutions in the monitoring of a real dragline excavator during its operation made it possible to identify a crack in one of its elements (weld joint in a dragline back leg). Citation: Barat, V.; Marchenkov, A.; Kritskiy, D.; Bardakov, V.; Karpova, M.; Kuznetsov, M.; Zaprudnova, A.; Keywords: acoustic emission; structure health monitoring; walking excavator; AE impulse detection Ushanov, S.; Elizarov, S. Structural Health Monitoring of Walking Dragline Excavator Using Acoustic Emission. Appl. Sci. 2021, 11, 3420. 1. Introduction https://doi.org/10.3390/app11083420 The dragline excavator is a widely known type of heavy equipment used in civil engineering and surface mining industry. Commonly, a system of a dragline consists of Academic Editor: a large bucket that is suspended from a boom with wire ropes (Figure1). The hoist rope Giuseppe Lacidogna is usually driven by diesel or electric motors. Its main function is to support the dragline bucket and hoist-coupler assembly from the boom. The dragrope is generally applied to Received: 10 March 2021 move the bucket horizontally. Unlike conventional types of excavators, dragline digs by Accepted: 7 April 2021 dragging the bucket only in one direction—towards the excavator [1]. Published: 11 April 2021 Currently, most dragline systems are operated in the open pit mining industry to process blasted rocks located both above and below the horizon level of the dragline Publisher’s Note: MDPI stays neutral excavator. Moreover, draglines have large dimensions—the boom length may reach with regard to jurisdictional claims in 90 m, the total height can constitute up to 100 m, and the bucket capacity may achieve published maps and institutional affil- 20 tons. Nowadays, the design of draglines is favorable due to a short working cycle and iations. high performance. Currently, more than 500 walking dragline excavators are in operation in the world, of which more than 200 are in Russia. Leading dragline manufacturers are Caterpillar, Liebherr, Uralmash, Bucyrus, and others. Draglines operate all year-round in detrimental weather conditions: significant tem- Copyright: © 2021 by the authors. perature fluctuations, great levels of humidity, and even regions with increased seismic Licensee MDPI, Basel, Switzerland. activity. Both the influence of external weather conditions and constant cyclic loadings This article is an open access article can lead to the formation of fatigue fracture-initiating defects in the metal elements of distributed under the terms and the dragline structure. Such defects can reduce the operation life of the equipment due conditions of the Creative Commons Attribution (CC BY) license (https:// to premature destruction. The breakdown basically causes considerable financial costs creativecommons.org/licenses/by/ associated with equipment downtime, repair costs, and terrible accidents [2]. The most 4.0/). common type of dragline excavator damage is cracks in the back legs; for the period from

Appl. Sci. 2021, 11, 3420. https://doi.org/10.3390/app11083420 https://www.mdpi.com/journal/applsci Appl. Sci. 2021, 11, x FOR PEER REVIEW 2 of 15

premature destruction. The breakdown basically causes considerable financial costs associated with equipment downtime, repair costs, and terrible accidents [2]. The most common type of dragline excavator damage is cracks in the back legs; for the period from 2015 to 2020, 15 accidents caused by breakage of back legs and five accidents caused by breakage of the dragline boom were registered in Russia. The most widely spread non-destructive testing (NDT) of dragline structural components is ultrasonic testing, which is carried out in the welded joints of an excavator twice a year. However, this method does not allow detect abrupt changes in the stress–strain Appl. Sci. 2021, 11, 3420 2 of 15 state. Furthermore, it is extremely time-consuming and leads to long-lasting downtime. Draglines are commonly equipped with a structural health monitoring system with video supervision systems and an electrical equipment diagnostics system, but operational 2015 to 2020, 15 accidents caused by breakage of back legs and ﬁve accidents caused by monitoring of thebreakage metal structures of the dragline conditions boom were is usually registered not in Russia.performed.

FigureFigure 1. Basic components1. Basic components of a dragline of ESH-20.90: a dragline 1—back ESH-20.90: legs, 2—boom, 1—back 3—hoist legs, rope,2—boom, 4—mast, 3—hoist 5—bucket, rope, 6—dragrope. 4— mast, 5—bucket, 6—dragrope. The most widely spread non-destructive testing (NDT) of dragline structural components is ultrasonic testing, which is carried out in the welded joints of an excavator twice In this paper,a year. the However,possibility this of method using doesthe acoustic not allow emission detect abrupt (AE) changes method in theto stress–straindetect defects in the draglinestate. Furthermore, metal structures it is extremely is discussed. time-consuming and leads to long-lasting downtime. AE is a non-destructiveDraglines are commonlytesting (NDT) equipped method with based a structural on the health physical monitoring phenomenon system with of video elastic waves emissionsupervision when systems a material and an electrical undergoes equipment irreversible diagnostics changes system, in but its operational internal moni- structure, for example,toring of because the metal of structures crack formation. conditions The is usually AE method not performed. has a high sensitivity In this paper, the possibility of using the acoustic emission (AE) method to detect to crack detectiondefects and is in used the dragline to detect metal cracks structures at early is discussed.stages of their growth. An important advantage of the AE AEmethod is a non-destructive is the ability to testing detect (NDT) defects method located based at ona great the physical distance phenomenon from the control point;of cracks elastic and waves other emission defects when emit a material AE waves, undergoes which irreversible propagate changes over the in itstest- internal ing structure andstructure, can be detected for example, by the because AE sensors of crack formation.at about 10 The m AEfrom method the defect. has a high The sensitivity AE method is passive,to crackit does detection not require and is probing used to detect action, cracks which at early allows stages it ofto theirbe effectively growth. An used important advantage of the AE method is the ability to detect defects located at a great distance from in structural health monitoring systems for monitoring loaded structures in the operating the control point; cracks and other defects emit AE waves, which propagate over the testing mode without decommissioning.structure and can be detected by the AE sensors at about 10 m from the defect. The AE AE testing providesmethod is passive,low noise it does immunity, not require th probingerefore, action, such whichexperiments allows it toare be tradition- effectively used ally carried out forin structural equipment health when monitoring the operat systemsion forof equipment monitoring loaded is suspended. structures Recently, in the operating the technology ofmode AE monitoring without decommissioning. which requires the use of specific methods for detecting AE signals against a AEnoise testing background provides low has noise beco immunity,me more therefore, popular such [3–5]. experiments AE monitoring are traditionally is carried out for equipment when the operation of equipment is suspended. Recently, the performed for both statically and dynamically loaded equipment. Most widespread ap- technology of AE monitoring which requires the use of speciﬁc methods for detecting plication of AE AEtesting signals in againstthe operating a noise background mode is testing has become of bearings, more popular compressors, [3–5]. AE monitoring and pumps [6–8]. Theis various performed defects for both have statically a specif andic dynamically AE signature loaded with equipment. a certain periodicity Most widespread connected with aapplication rotation frequency. of AE testing The in theperiod operatingicity analysis mode is testing allows of to bearings, recognize compressors, the AE and activity against thepumps noise [6– process.8]. The various Some defects of the havemost a speciﬁccomplicated AE signature structures with for a certain AE struc- periodicity tural health monitoringconnected are with railways a rotation and frequency. roads bridges The periodicity [9,10]. analysisThe complicity allows to consists recognize of the AE activity against the noise process. Some of the most complicated structures for AE structural the fact that trainshealth and monitoring cars are oftentimes are railways anda noise roads source bridges and [9,10 ].a Theload complicity excitation, consists and ofthe the fact that trains and cars are oftentimes a noise source and a load excitation, and the moment of passage of vehicles is simultaneously the moment of the most probable occurrence of AE activity. Solving of this problem is achieved by reducing the distance between AE sensors, Appl. Sci. 2021, 11, x FOR PEER REVIEW 3 of 15

Appl. Sci. 2021, 11, 3420 3 of 15 moment of passage of vehicles is simultaneously the moment of the most probable occurrence of AE activity. Solving of this problem is achieved by reducing the distance between AE sensors, high-frequency filtration, and using additional strain gauges or vibration measurements.high-frequency Another filtration, way to and detect using the additional AE activity strain of a gauges defect during or vibration the testing measurements. of the noisyAnother structure way is toto detectdesign the a scaled AE activity model ofof athe defect testing during structure. the testing This method of the noisy is success- structure fullyis used to design for the a AE scaled testing model of the of various the testing chemical structure. reactors, This absorbers, method isand successfully crystallizers used [11,12].for theInvestigation AE testing of of the the scaled various model chemical of the reactors, test structure absorbers, allows and obtaining crystallizers a pattern [11, 12]. of acousticInvestigation noise, of in the this scaled case modelthe appearance of the test of structure a defect allows can be obtaining detected a patternas a change of acoustic of regularitynoise, inin thisthe caseparameters the appearance of the reference of a defect noise can signal. be detected If it is as impossible a change ofto regularitydesign a in model,the parametersthe study of of noise the reference is carried noise out signal.on the operating If it is impossible equipment. to design There a model,are some the pa- study persof devoted noise is to carried the AE out testing on the of operatinghandling eq equipment.uipment: cranes, There arelifts, some and excavators papers devoted [13– to 17].the However, AE testing as a of rule, handling the testing equipment: of hand cranes,ling equipment lifts, and involves excavators a special [13–17]. low-speed However, as loadinga rule, program the testing which of handlingprovides equipmenta low noisyinvolves level. AE a testing special of low-speed a walking loading dragline program ex- cavatorwhich is a provides rather difficult a low noisy problem, level. since AE testingit involves of a the walking testing dragline of a large-sized excavator structure is a rather in difficultdifficult weather problem, conditions since it involves and at thean extremely testing of ahigh large-sized noise level. structure in difficult weather conditionsThis paper and describes at an extremely a method high of design noise of level. a walking dragline excavator monitoring system basedThis paper on the describes application a method of the of AE design method. of a walkingA simplified dragline block excavator diagram monitoring of the monitoringsystem basedsystem on is theshown application in Figure of 2. the It co AEnsists method. of hardware A simplified part and block software diagram parts, of the the monitoringsoftware parts system in turn is shownare divided in Figure into2 ac. Itquisition consists system of hardware and da partta analysis and software system. parts, the software parts in turn are divided into acquisition system and data analysis system.

Figure 2. Block diagram of the structure health monitoring system. Figure 2. Block diagram of the structure health monitoring system. The hardware equipment is represented by a digital measuring system A-Line 32D DDMThe hardware (produced equipment by Interunis-IT is represented company). by a Its digital main measuring feature consists system of A-Line subsequential 32D DDMconnection (produced of by the Interunis-IT measuring channels.company). A Its data main acquisition feature consists system of provides subsequential the traditional con- nectionAE impulseof the measuring threshold channels. detection, A calculation data acquisition of AE parameters,system provides and location the traditional of AE sources AE based on the determination of the difference in the time of arrival of AE impulses to impulse threshold detection, calculation of AE parameters, and location of AE sources neighboring sensors. The data analysis algorithm provides intelligent processing of AE based on the determination of the difference in the time of arrival of AE impulses to neigh- data for detection of the defects against the background of high-intensity technological boring sensors. The data analysis algorithm provides intelligent processing of AE data for noise. detection of the defects against the background of high-intensity technological noise. In a preliminary analysis, the possibility of carrying out AE monitoring to assess the In a preliminary analysis, the possibility of carrying out AE monitoring to assess the strain–stress state of the dragline components was studied by identification of favorable strain–stress state of the dragline components was studied by identification of favorable and adverse factors. The most unfavorable factor is a high level of noise in the dragline and adverse factors. The most unfavorable factor is a high level of noise in the dragline operating cycle. In most cases it is caused by the movement of the dragrope and hoisting operating cycle. In most cases it is caused by the movement of the dragrope and hoisting ropes. Therefore, the high-intensity acoustic noise can reach 75 dB in the 100–400 kHz ropes. Therefore, the high-intensity acoustic noise can reach 75 dB in the 100–400 kHz range. Despite this critical level, monitoring still can be performed because the sources of range. Despite this critical level, monitoring still can be performed because the sources of noise, as a rule, are not localized at certain points in the structure and are not identified noise,on as the a diagram.rule, are not The localized favorable at certain factor ispoints the waveguide in the structure form. and The are dragline not identified structural on thecomponents diagram. The usually favorable obtain factor the form is the of waveguide hollow pipelines form. The and dragline could bestructural interpreted com- as a ponentsone-layered usually cylindric obtain the waveguide. form of hollow In the absencepipelines of and multiple could reflections be interpreted and scattering, as a one- the layeredAE impulses cylindric havewaveguide. a relatively In the short absence rise timeof multiple and duration, reflections while and providing scattering, a significantthe AE difference in the waveform of the useful and noise data. Appl. Sci. 2021, 11, x FOR PEER REVIEW 4 of 15

impulses have a relatively short rise time and duration, while providing a significant difference in the waveform of the useful and noise data.

2. Materials and Methods Appl. Sci. 2021, 11, 3420 4 of 15 This section represents methods for analyzing various factors which impact on the AE structural health monitoring technology and collecting information about the AE source, waveguide properties, and the features of the structure loading in the operating 2. Materials and Methods cycle, as well as the parameters of noise. This section represents methods for analyzing various factors which impact on the AE 2.1.structural AE Source health Characterization monitoring technology and collecting information about the AE source, waveguide properties, and the features of the structure loading in the operating cycle, as wellMechanical as the parameters tests were of noise. carried out on specimens made of the same steel as the dragline back legs to determine the AE parameters of a fatigue crack. Specimens were pro- duced2.1. AEfrom Source 16G Characterization low-alloyed steel (marked in accordance with Russian state standard) sheets usingMechanical laser testscutting. were The carried chemical out on specimenscomposition made of of the the studied same steel steels as the is draglinepresented in Tableback 1. legs to determine the AE parameters of a fatigue crack. Specimens were produced from 16G low-alloyed steel (marked in accordance with Russian state standard) sheets Tableusing 1. The laser chemical cutting. composition The chemical of composition the 16G low-alloyed of the studied steel. steels is presented in Table1.

Table 1. The chemical compositionThe Content of the of 16G Chemical low-alloyed Elements steel. (% wt.)

C Mn TheSi Content ofCr Chemical Ni Elements (%Al wt.) Mo S P 0.14–0.18C 1.0–1.2 Mn 0.15–0.3 Si ≤ Cr0.2 Ni≤0.3 0.02–0.05 Al Mo≤0.8 S≤0.02 P ≤0.025 0.14–0.18 1.0–1.2 0.15–0.3 ≤0.2 ≤0.3 0.02–0.05 ≤0.8 ≤0.02 ≤0.025 The shape and overall dimensions of the specimens (Figure 3a) were selected on the basis of Russian state standard GOST 25.506-85. Thus, a series of rectangular specimens The shape and overall dimensions of the specimens (Figure3a) were selected on the with a thickness of 5 mm and with an edge notch were prepared. The width is equal to b basis of Russian state standard GOST 25.506-85. Thus, a series of rectangular specimens = 50with mm, a thicknessand the length of 5 mm l = and 350 with mm an was edge chosen notch due were to prepared. the condition The width of l ≥ is 2 equalb and to taking intob =account 50 mm, the and distance the length betweenl = 350 the mm grips. was chosen The notch due towidth the condition e also should of l ≥ meet2b and the re- quirementtaking into (e account≤ 0.06b) the and distance was selected between as the e grips.= 4 mm, The the notch opening width eanglealso should Θ corresponded meet the to (30°requirement ≤ Θ ≤ 60°) (ande ≤ 0.06wasb )Θ and = 45°, was and selected the notch as e = 4depth mm, the h = opening 10 mm. angle Θ corresponded to (30◦ ≤ Θ ≤ 60◦) and was Θ = 45◦, and the notch depth h = 10 mm.

(a) (b) FigureFigure 3. The 3. The design design of ofthe the test test specimen: specimen: ( (aa)) scheme scheme ofof the the specimen; specimen; (b )(b) diagram diagram of the of the loading loading cycle. cycle.

A Afatigue fatigue crack crack was was grown grown on on each each specimen specimen by by cyclic cyclic tensile tensile loading usingusing thethe pul- pulsation cycle (cycle asymmetry coefﬁcient Rc = 0) with a frequency of 5 Hz and a sation cycle (cycle asymmetry≈ coefficient· Rc = 0) with a frequency of 5 Hz and a maximum maximum cycle load ofσσmax 0.6σσy, where σy is a yield stress. The number of loading cyclecycles load was of chosenσmax ≈ so0.6· thaty, thewhere fatigue y crackis a yield developing stress.under The numb cyclicer loading of loading in the lateralcycles was chosennotch so region that the would fatigue reach crack a certain developing length. Nominal under crackcyclic lengths loading ranged in the from lateral 3 to notch 15 mm. region wouldThus, reach the total a certain length length. of the notch Nominal and crackcrack was lengths approximately ranged from 15–28 3 to mm 15 depending mm. Thus, the totalon length the specimen. of the notch After theand crack crack growth, was approximately the specimens were15–28 tested mm underdepending static tensionon the spec- using a loading scheme adopted during industrial AE testing (Figure3b) [ 18]. A series of experiments was done for 10 specimens with identical parameters. Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 15

Appl. Sci. 2021, 11, 3420 imen. After the crack growth, the specimens were tested under static tension using a 5load- of 15 ing scheme adopted during industrial AE testing (Figure 3b) [18]. A series of experiments was done for 10 specimens with identical parameters. The A-Line 32D AE system (LLC “Interunis-IT”) was chosen during the experiment. It wasThe equipped A-Line 32Dwith AEfour system measuring (LLC channels “Interunis-IT”) including was the chosen PAEF during-014 preamplifier the experiment. and GT200It was equipped resonant withsensor four (LLC measuring “GlobalTest”, channels resonant including frequency the PAEF-014 180 kHz). preampliﬁer The intrinsic and noiseGT200 of resonant the equipm sensorent, (LLC preamplifier “GlobalTest”,, and AE resonant sensor frequency was 26 dB. 180 The kHz). discrimination The intrinsic thresh- noise oldof the was equipment, selected as preampliﬁer, 45 dB (6 dB higher and AE than sensor the noise was 26 level dB. of The the discrimination testing machine). threshold was selectedFigure 4 aspresents 45 dB (6the dB dependence higher than of the the noise cumulative level of AE the hits testing depending machine). on the current Figure4 presents the dependence of the cumulative AE hits depending on the current stress and time. Cumulative AE hits have a power-law dependence of the σ/σy relation whichstress andis typical time. for Cumulative the crack presence. AE hits have The anumber power-law of AE dependence hits registered of the duringσ/σy therelation load- whiching of the istypical specimens for thevaried crack in the presence. range from The number118 to 240 of with AE hitsan average registered value during about the of loading of the specimens varied in the range from 118 to 240 with an average value about of N∑ = 185. The number of AE hits depends on the loading stress. For more objective evalu- ationN∑ = of 185. the Thematerial number emissivity of AE the hits Palmer depends–Heald on the model loading was implemented. stress. For more objective evaluation of the material emissivity the Palmer–Heald model was implemented.

350 300 250 200 150 100

Cummulative Cummulative AE hits 50 0 0 0.2 0.4 0.6 0.8

σ/σy experiment approximation

(a) (b)

Figure 4.4. CumulativeCumulative acousticacoustic emissionemission (AE)(AE) hitshits NNΣΣ relations:relations: ((a)) cumulativecumulative AEAE hitshits NNΣΣ vs.vs. current stressstress σσ/σ/σyy and its Palmer–Heald approximation; (b) cumulative AE hits NΣ and loading force P vs. time t. Palmer–Heald approximation; (b) cumulative AE hits NΣ and loading force P vs. time t.

According to the Palmer–HealdPalmer–Heald model, 휋 휎 푁Σ = 퐷 ∙ 푎 ∙ (푠푒푐 ( π )σ− 1) (1) NΣ = D·a· sec 2 휎푦 − 1 (1) 2 σy where  is the actual stress; y—yield stress; a is half the length of the crack; D—dimen- wheresional σcoefficientis the actual of stress;the modelσy—yield (1/mm), stress; dependinga is half the on lengththe characteristics of the crack; D of—dimensional the material, temperaturecoefficient of, theand model type of (1/mm), stress–strain depending state. on the characteristics of the material, tempera- ture,It and is presented type of stress–strain in [19] that state. the multiplicative coefficient of the Palmer–Heald model D canIt be is presentedconsidered in as [19 parameters] that the multiplicative that may be applied coefficient to describe of the Palmer–Heald the material AE model emis-D sivitycan be independent considered as of parameters loading conditions, that may be crack applied length, to describe and stress the material intensity AE factor. emissivity De- pendenceindependent of ofcumulative loading conditions, AE hits vs. crack time length, was approximated and stress intensity using factor. the Palmer Dependence–Heald equationof cumulative for the AE data hits obtained vs. time after was testing approximated of each 10 using specimens the Palmer–Heald in the series. equationAn example for ofthe such data an obtained approximation after testing is given of each in Figure 10 specimens 4b. For ineach the dependence, series. An example the values of such of the an parameterapproximation D, the is givenerror δ inD Figure, and the4b. coefficient For each dependence, of determination the values R2 were of the determined parameter andD, 2 presentedthe error δ inD, the and Table the coefficient 2. The parameter of determination D was calculatedR were based determined on the andvalues presented presented in inthe Table Table 22 .and The constituted parameter 1D51.1was ± 8.4 calculated (1/mm). based on the values presented in Table2 and constituted 151.1 ± 8.4 (1/mm). Appl. Sci. 2021, 11, x FOR PEER REVIEW 6 of 15

Appl. Sci. 2021, 11, 3420 6 of 15 Table 2. AE emissivity estimation—the Palmer–Heald model parameters.

D, 1/mm δD, % R2 Table 2. AE emissivity137.5 estimation—the Palmer–Heald0.93 model parameters. 0.97 149.6 0.87 0.96 D, 1/mm δD, % R2 144.9 1.4 0.92 137.5152.3 0.930.82 0.97 0.98 149.6 0.87 0.96 144.9151.4 1.40.93 0.92 0.96 152.3155.3 0.821.4 0.98 0.99 151.4165.1 0.933.1 0.96 0.95 155.3160.5 1.41.5 0.99 0.94 165.1 3.1 0.95 153.2 1.3 0.98 160.5 1.5 0.94 153.2141.3 1.32.1 0.98 0.97 141.3144.9 2.11.4 0.97 0.92 Since144.9 the maximum load in the operation 1.4 cycle of the dragline is approximately 0.92 0.6σy, the approximately number of AE hits related to crack length is N∑/a = D (sec(0.3π) − 1) ≈ 106Since . the maximum load in the operation cycle of the dragline is approximately 0.6σy, the approximatelyFigure number5 shows of the AE dependence hits related to of crack the lengthamplitudes is N∑ /ofa =AED (sec(0.3impulsesπ) − emitted1) ≈ 106. by the crackFigure depending5 shows the on dependencetime (Figure of 5a) the and amplitudes the AE impulse of AE impulses amplitudes emitted probability by the crackdistribu- dependingtion in double on time logarithmic (Figure5a) andaxes the (Figure AE impulse 5b). The amplitudes amplitudes probability of the AE distribution impulses emitted in doubleby the logarithmic crack do axesnot exceed (Figure 572b). dB, The and amplitudes the amplitude of the AEdistribution impulses corresponds emitted by the to the crackWeibull do not law exceed with 72 the dB, parameters and the amplitude k = 1.17, λ distribution = 6.13. corresponds to the Weibull law with the parameters k = 1.17, λ = 6.13.

(a) (b)

FigureFigure 5. Amplitudes 5. Amplitudes of of AE AE impulses: impulses: (a ()a) AE AE impulse impulse amplitudesamplitudes A and loading loading force force PP vs.vs. time time t; (tb;() bAE) AE impulse impulse ampli- amplitudestudes distribution. distribution.

AssessmentAssessment of of the the AE AE emissivity emissivity and and determination determination ofof thethe range of of AE AE impulse impulse am- amplitudesplitudes emittedemitted byby the the crack crack are are used used to to determine determine the the parameters parameters of theof the AE AE structural structural healthhealth monitoring monitoring system system and and estimate estimate the the reliability reliability of AEof AE impulse impulse detection detection against against backgroundbackground noise. noise.

2.2.2.2. AE AE Waveguide Waveguide Characterization Characterization In theIn the previous previous section, section, the the source source of of AE AE is is characterized. characterized. However,However, to detect detect AE AE sig- signalnal impulsesimpulses effectivelyeffectively against against a a background background of of noise, noise, it isit necessaryis necessary to assessto assess how how AE AE parametersparameters vary vary during during propagation propagation along along the waveguide. the waveguide. Traditionally, Traditionally, the investigation the investiga- of thetion waveguide of the waveguide of the testingof the testing objects objects is carried is carried out mainly out mainly empirically empirically by measuring by measuring thethe Hsu-Nielsen Hsu-Nielsen lead lead break break response response at various at various points points of theof the structure. structure. In thisIn this study study an an analytical approach is implemented due to limited access to the dragline constructive analytical approach is implemented due to limited access to the dragline constructive el- elements. Load-bearing metal structures of dragline such as back legs and booms represent ements. Load-bearing metal structures of dragline such as back legs and booms represent hollow tubes with a diameter of 20 inches which could be considered as a single-layer cylindrical waveguide. Ordinary form of the waveguide allows to calculate AE signal propagation analytically using modal analysis [20]. Appl. Sci. 2021, 11, 3420 7 of 15

Modal analysis is a common method for the waveguide calculating based on the properties of orthogonality and completeness of normal waves. It allows to calculate in Fourier space the displacement of the testing structure at the point z caused by the action of a point source with the waveform u(0,t). The result signal is determined using the expression u(z, f ) = u(0, f )·ap(z, f ) (2) where u(0,t) is the signal emitted by the AE source located in the point z = 0, and u(0,f ) = F[u(0,t)] is its Fourier transform. It was shown in [21] that when the external force is a point source, the coefﬁcients ap(z) can be determined using the simpliﬁed formula:

jkpz ap(z) = e /4 (3)

where kp = (2πf )/cph(f ) is the wave number; cph(f ) is the phase velocity of the normal wave, f —frequency. In the case of a cylindrical waveguide with a large radius of curvature, Lamb waves can be considered as the main type of normal waves. One of the main aspects related to the influence of the waveguide is attenuation. The attenuation effect is a difficult problem from the point of view of interpreting the AE data, since different wave modes and different frequency components are characterized by different attenuation coefficients, which has a significant effect on the waveform of AE impulse. The attenuation coefficients of the Lamb waves modes γS0 and γA0 are a linear combination of the attenuation coefficients of the longitudinal wave α and transverse wave β: γ = A α + B β S0 S0 S0 (4) γA0 = AA0α + BA0β

where AS0, BS0, AA0, BA0—coefficients that are determined by the type of Lamb wave, relative plate thickness, and Poisson’s ratio. Analytical expressions of the coefficients are given in [22], their values depend on frequency f and thickness of the testing structure h. The analytical method based on the modal analysis, supplemented by analytical consideration of the attenuation coefficient Appl. Sci. 2021, 11, x FOR PEER REVIEWand the impulse response of AE sensor, allows simulating AE signals of a realistic form [23]. 8 of 15

Figure6 represents the calculated signal (Figure6a) and experimental signal obtained using the Hsu-Nielsen simulator (Figure6b) for wall thickness 20 mm at 4 m from the AE sensor.

(a) (b)

FigureFigure 6. Comparison 6. Comparison of of the the modelled modelled and and experimental AE AE signals signalsU Ufor for a waveguidea waveguide length length of 4 of m 4 and m and a thickness a thickness of of 20 mm:20 ( mm:a) the (a simulated) the simulated signal, signal, (b) (theb) the measured measured signal. signal.

Based on the analytical method, a parametric model of the AE impulse can be designed. It presents the dependence of the AE parameters (rise time (RT), rise angle (RA), and duration) on propagation distance. The RA parameter represents the value inverse to slow rate and can be calculated with the equation RA = RT/A, (5) where A is the amplitude of the AE impulse. In the case of one-layered metallic waveguide, the AE impulse characterizes, as a rule, with some specific parameters, values such as a small RA and a short RT. The determination of the expected priory parameters of the AE impulse in advance makes it possible to identify them effectively against the background of noises, whose impulse components correspond to a larger time scale. Figure 7 shows RT and RA versus propagation distance L, which determine the expected values of AE parameters for a given distance between the AE source and sensor. These dependences can be interpreted as a parametric model of the AE impulse, based on which the identification of AE impulses against the background of noise will be carried out.

300 250 200

, ms 150 RT 100 50 0 0 5 10 15 L, m

(a) (b)

Figure 7. Dependence of AE parameters versus propagation distance L: (a) rise angle (RA) vs. distance L, (b) rise time (RT) vs. distance L. Appl. Sci. 2021, 11, x FOR PEER REVIEW 8 of 15

Appl. Sci. 2021, 11, 3420 8 of 15 (a) (b) Figure 6. Comparison of the modelled and experimental AE signals U for a waveguide length of 4 m and a thickness of 20 mm: (a) the simulated signal, (b) the measured signal. Based on the analytical method, a parametric model of the AE impulse can be designed. It presentsBased theon the dependence analytical method, of the AE a parametric parameters model (rise of time the (RTAE), impulse rise angle can (beRA de-), and duration)signed. It onpresents propagation the dependence distance. of The theRA AEparameter parameters represents (rise time the (RT value), rise inverseangle (RA to), slow rateand andduration) can be on calculated propagation with distance. the equation The RA parameter represents the value inverse to slow rate and can be calculated with the equation RA = RT/A, (5) RA = RT/A, (5) wherewhereA Ais is thethe amplitudeamplitude of of the the AE AE impulse. impulse. InIn thethe casecase of of one-layered one-layered metallic metallic waveguide, waveguide, the the AE AE impulse impulse characterizes, characterizes, as as a a rule, withrule, some with speciﬁcsome specific parameters, parameters, values values such assuch a small as a smallRA and RAa and short a shortRT. The RT. determination The deter- ofmination the expected of the priory expected parameters priory parameters of the AE impulseof the AE in impulse advance in makes advance it possible makes it to possi- identify themble to effectively identify them against effectively the background against th ofe background noises, whose of noises, impulse whose components impulse correspondcompo- tonents a larger correspond time scale. to a larger Figure time7 shows scale.RT Figureand 7RA showsversus RTpropagation and RA versus distance propagationL, which determinedistance L, thewhich expected determine values the ofexpected AE parameters values of forAE aparameters given distance for a given between distance the AE sourcebetween and the sensor. AE source These and dependences sensor. These can dependences be interpreted can asbe ainterpreted parametric as model a parametric of the AE impulse,model of based the AE on impulse, which the based identiﬁcation on which ofthe AE identification impulses against of AE theimpulses background against of the noise willbackground be carried of out.noise will be carried out.

300 250 200

, ms 150 RT 100 50 0 0 5 10 15 L, m

(a) (b)

FigureFigure 7. Dependence 7. Dependence of AEof AE parameters parameters versus versus propagation propagation distance distanceL :(L:a ()a) rise rise angle angle ( RA)(RA)vs. vs. distancedistance L,,( (b)) rise rise time time (RT) vs. distance L. (RT) vs. distance L.

RA varies exponentially with the distance, while RT increases linearly. Both parameters have rather low values; at 15 m, the RA does not exceed 50 ms/V, when the RT of the leading edge of the AE signal pulse for the same distance turns out to be less than 300 µs.

2.3. Noise Parameters The main source of acoustic noise measured during the AE testing of walking dragline excavator is friction and vibration generated by the movement of the hoist ropes and dragropes that control the bucket position. The noise has a stochastic non-stationary nature, its intensity turned out to be different for AE sensors installed on various dragline constructive parts. The lower the noise intensity, the greater is the distance of the structural component from the moving ropes. Figure8 shows the empirical probability function of the noise long-time realization measured with a different sensor located on the structural components above and below the boom, and on the side frame—back legs and a mast. The highest noise level was observed along the measuring channel installed in the lower part of the mast near the dragropes, while the lowest level was detected along the channels installed on the lower part of the boom. It can be explained by the fact that they are placed far away from the trajectory of the ropes. Appl. Sci. 2021, 11, x FOR PEER REVIEW 9 of 15

RA varies exponentially with the distance, while RT increases linearly. Both parameters have rather low values; at 15 m, the RA does not exceed 50 ms/V, when the RT of the leading edge of the AE signal pulse for the same distance turns out to be less than 300 μs.

2.3. Noise Parameters The main source of acoustic noise measured during the AE testing of walking dragline excavator is friction and vibration generated by the movement of the hoist ropes and dragropes that control the bucket position. The noise has a stochastic non-stationary nature, its intensity turned out to be different for AE sensors installed on various dragline constructive parts. The lower the noise intensity, the greater is the distance of the structural component from the moving ropes. Figure 8 shows the empirical probability function of the noise long-time realization measured with a different sensor located on the structural components above and below the boom, and on the side frame—back legs and a mast. The highest noise level was observed along the measuring channel installed in the lower part of the mast near the dragropes, while the lowest level was detected along the

Appl. Sci. 2021,channels11, 3420 installed on the lower part of the boom. It can be explained by the fact9 of that 15 they are placed far away from the trajectory of the ropes.

Figure 8. Empirical probability distribution of acoustic noise for a various dragline component. Figure 8. Empirical probability distribution of acoustic noise for a various dragline component. It can be inferred from Figure8 that there is a high level of noise with a standard 8 It can be deviationinferred of 40–45from dB. Figure The maximum 8 that valuesthere typical is a forhigh a set level of 10 samplesof noise (that with correspond a standard to 100 s with a sampling frequency equal to 1 MHz) are 50–65 dB. The noise is non- 8 deviation of 40–45stationary dB. because The maximu it is causedm by values the uneven typical movement for a ofset the of ropes 10 in samples different phases(that correspond to 100 sof with the dragline a sampling operation frequency cycle. For instance,equal to when 1 MHz) the bucket are 50–65 is loaded, dB. the The dragrope noise is is nonstationary becausepulled. it When is caused there is aby stage the of unloading,uneven movement both the dragrope of the and hoistropes rope in are different weakened phases and moved. When the platform turns, no movement of the ropes is visible, and the intensity of the draglineof operation noise is dramatically cycle. reduced.For instance, Figure9a when presents the the longtimebucket noiseis loaded, waveform; the its dragrope peak is pulled. When correspondsthere is a tostage 68 dB, of and unloading, in the low-intensity both phasethe dragrope ~ 40–45 dB. Aand favorable hoist factor rope is are that weakened and moved.the impulse When noise the componentsplatform differturns, from no the movement AE signal impulses of the by ropes greater is values visible, of the and the RT and RA parameters. Figure9b shows the empirical distribution of the RA parameter intensity of noisecorresponding is dramatically to noise impulse reduced. components. Figure The9a distributionpresents the mode longtime equaled the noise value wave- Appl. Sci. 2021, 11, x FOR PEER REVIEW 10 of 15 form; its peak RAcorresponds= 900 ms/V, whichto 68 isdB, signiﬁcantly and in the higher low-intensity than the values phase of the RA~ 40–45parameter dB. forA favorable AE factor is that theimpulses impulse emitted noise by a components defect. differ from the AE signal impulses by greater values of the RT and RA parameters. Figure 9b shows the empirical distribution of the RA parameter corresponding to noise impulse components. The distribution mode equaled the value RA = 900 ms/V, which is significantly higher than the values of the RA parameter for AE impulses emitted by a defect.

(a) (b)

Figure 9. Parameters of the noise process: (a) noise waveform longtime realization, (b) distribution of RA parameter for the Figure 9. Parameters of the noise process: (a) noise waveform longtime realization, (b) distribution of RA parameter for thenoise noise impulse impulse components. components.

2.4. Filtering Method Conventionally, the detection of AE impulses is carried out by the threshold method, in accordance with AE, impulses are detected during the threshold crossing. This method is highly effective and convenient in a case when a noise has a stationary character and can be rejected with the nonlinear threshold filtering. However, in the case of nonstationary impulse noise, the use of the threshold method is not acceptable. According to the AE non-destructive testing method, the threshold is set in such a way that no more than one false hit detection is recorded in 100 s. According to the probability distribution of noise samples (Figure8), the amplitude discrimination threshold should be set between 55 and 65 dB for different locations. Such an increase in the amplitude discrimination threshold leads to a significant increase in the probability of missing a defect. A preliminary analysis of the AE source shows that at a value of the amplitude discrimination threshold of 45 dB, the average emissivity of the material is about 105 AE hits per 1 mm of crack growth. However, it can be extrapolated that with an increase in the discrimination threshold from 45 to 60 dB, the number of AE hits drastically decreases by about 94% due to the exponential distribution of the AE pulse amplitudes (Figure4b). The emissivity in this case will be on the order of 5–7 AE hits per 1 mm of crack growth. A special method was developed to increase the accuracy of crack detection in the dragline load-bearing metal structures. In accordance with this method, the amplitude discrimination level is set below the noise level to allow recording both the AE process and the noise technological process during the dragline operation. In this case, a type II error associated with the defect missing is minimized by increasing the probability of a type I error connected with false detection of a defect due to erroneous noise impulse location. Since the noise intensity is significantly higher than the AE hits count rate, special filtering is developed to increase the reliability of the testing. Filtering is effective and specific since it is based on the parametric model of AE impulse, representing a set of characteristic values of the AE parameters corresponding to a certain distance between the AE sensor and the defect. A spatial parametric filter is proposed for detecting AE impulses from nonstationary impulse noise. This filter excludes from the location schemes those AE events whose parameters are more consistent with the noise process and are not typical for AE impulses caused by a defect. Since the tested elements of the dragline are long pipelines (the length significantly exceeds the diameter), a linear circuit with an antenna formed by two sensors is used to determine the location of the defect. The filtering algorithm assumes that the waveform of AE impulse is determined primarily by the parameters of the acoustic waveguide. The acoustic emission impulses emitted by the AE source are stochastic and characterized by a short duration, on the order of nanoseconds. Since the propagation of the AE signals has a dispersive character, when propagating from the emission point to the AE sensor, the signal duration increases and reaches 100–1000 µs at a distance about 10 meters, and the signal waveform is also determined mainly by the phases of different frequency components of the Lamb modes. In this case, the AE impulse, which is initially stochastic in nature, becomes principally deterministic, and its waveform and spectrum are mainly determined by the parameters of the waveguide. Based on the above, the following filtering method is applicable: for all located AE events, two RT parameters (RT1 for sensor 1 and RT2 for sensor 2) and two RA parameters (RA1 for sensor 1 and RA2 for sensor 2) are evaluated. Then, the correspondence between the values of the RA, RT parameters and a defect distance are determined. If the AE event is located at the distance x from the 1st sensor, the parameters RT1 and RA1 would correspond to the distance x, and RT2 and RA2 would correspond to distance (L–x), where L is the distance between the sensors and x is the estimated distance between the AE source and the sensor. An AE event is considered as corresponding to the propagation of a defect if, in addition to the standard location criteria associated with the discrepancy of the location Appl. Sci. 2021, 11, x FOR PEER REVIEW 11 of 15

The filtering algorithm assumes that the waveform of AE impulse is determined primarily by the parameters of the acoustic waveguide. The acoustic emission impulses emitted by the AE source are stochastic and characterized by a short duration, on the order of nanoseconds. Since the propagation of the AE signals has a dispersive character, when propagating from the emission point to the AE sensor, the signal duration increases and at about meters reaches 100–1000 μs at a distance about 10 meters, and the signal waveform is also determined mainly by the phases of different frequency components of the Lamb modes. In this case, the AE impulse, which is initially stochastic in nature, becomes principally deterministic, and its waveform and spectrum are mainly determined by the parameters of the waveguide. Based on the above, the following filtering method is applicable: for all located AE events, two RT parameters (RT1 for sensor 1 and RT2 for sensor 2) and two RA parameters (RA1 for sensor 1 and RA2 for sensor 2) are evaluated. Then, the correspondence between the values of the RA, RT parameters and a defect distance are determined. If the AE event is located at the distance x from the 1st sensor, the parameters RT1 and RA1 would correspond to the distance x, and RT2 and RA2 would correspond to distance (L–x), where L is Appl. Sci. 2021, 11, 3420 11 of 15 the distance between the sensors and x is the estimated distance between the AE source and the sensor. An AE event is considered as corresponding to the propagation of a defect if, in addition to the standard location criteria associated with the discrepancy of the loca- tionamplitude, amplitude,RA andRA andRT parametersRT parameters estimated estimated for AEfor AE impulses impulses measured measured with with both both AE AEchannels channels correspond correspond to theto the reference reference value value with with an erroran error less less than than 30%. 30%. RARA andand RTRT parametersparameters reference reference values values vs. vs. distance distance for for both both measuring measuring channels channels are are shownshown in in Figure Figure 10.10.

(a) (b) Figure 10. Parametrical model of AE impulse: (a) RA depending on the coordinate of AE source, (b) RT depending on the Figure 10. Parametrical model of AE impulse: (a) RA depending on the coordinate of AE source, (b) RT depending on the coordinate of AE source. coordinate of AE source.

3.3. Results Results TheThe approaches approaches proposed proposed in in this this work work were were tested tested during during structural structural health health monitor- moni- ingtoring of an of ESH an ESH 20.90 20.90 walking walking dragline dragline excavator. excavator. The reason The reason for the for AE the testing AE testing was a was pre- a viousprevious damage damage to the to left the dragline left dragline back back legs legs(a crack (a crack in the in weld). the weld). For data For dataacquisition, acquisition, the INTERUNIS-ITthe INTERUNIS-IT A-Line A-Line 32D 32D DDM DDM system system wa wass used, used, equipped equipped with with 30 30 measuring measuring chan- chan- nels,nels, each each of of which which was was connected connected to to resonant resonant sensors sensors (AE (AE GT200) GT200) with with a a bandwidth bandwidth of of Appl. Sci. 2021, 11, x FOR PEER REVIEW140–200 kHz and a resonance frequency of 180 kHz. The measurement was carried out 12 of 15 140–200 kHz and a resonance frequency of 180 kHz. The measurement was carried out in thein thefrequency frequency range range from from 100 to 100 400 to kHz, 400 kHz, whic whichh ensures ensures a low a intensity low intensity of the ofnoise the envi- noise ronment.environment. AE sensors AE sensors were were installed installed on onthe the main main metal metal structural structural dragline dragline components: components: three sensors for the left and three for the right back legs, eight for the upper boom, four for three sensors for the left and three for the right back legs, eight for the upper boom, four thefor lowerthe lower boom, boom, four for four the lowerfor the back lower legs andback three legs for and the three mast. Thefor the distance mast. between The distance be- thetween AE sensorsthe AE varied sensors between varied 5 and between 11 m. Figure 5 and 11 11shows m. theFigure location 11 shows of AE sensors the location on of AE asensors dragline on boom. a dragline boom.

FigureFigure 11. 11.Location Location of AEof AE sensors sensor on as draglineon a dragline boom. boom.

Intense acoustic noise was measured during dragline operation cycle. The amplitudes of the noise impulses covered almost the entire dynamic range of the AE system from the threshold level of 50 dB to 90 dB. The noise activity reached about 55 AE hits per second on average. Figure 12 shows the values of AE impulse amplitude and cumulative AE hits obtained within an hour. More than 200,000 impulses were registered during an hour of observation. Since the defect emissivity estimated in Section 2.1 was approximately 100 hits per loading cycle, it was obvious that most AE data were caused by noise and this case can hide the presence of a dangerous defect.

100 100 80 80 60 60 , dB

A 40 40 20 0 20 AE hits count rate 0 1000 2000 3000 0 t, s 0 1000 2000 3000 t, s

(a) (b)

Figure 12. AE data measured during structural health monitoring of walking dragline excavator: (a) amplitudes of AE impulses in a period of an hour, (b) AE hits count rate in a period of an hour.

The most radical noise filtering result was reached with the help of the location procedure. The main source of noise was the ropes movement, and it did not correspond to a specific spatial coordinate displayed on the “location”. Figure 13 presents the linear location diagram of AE sources on the right back leg. The location diagram was formed by two antennas, the first one included sensor 1 and sensors 2, the second one—sensor 2 and sensor 3. AE sensors were located at 1 m, 9 m, and 15 m from the upper edge of the bac Appl. Sci. 2021, 11, x FOR PEER REVIEW 12 of 15

for the lower boom, four for the lower back legs and three for the mast. The distance between the AE sensors varied between 5 and 11 m. Figure 11 shows the location of AE sensors on a dragline boom.

Appl. Sci. 2021, 11, 3420 12 of 15

Figure 11. Location of AE sensors on a dragline boom.

IntenseIntense acousticacoustic noisenoise waswas measured measured during during dragline dragline operation operation cycle. cycle. The The amplitudes ampli- tudesof the of noise the impulsesnoise impulses covered covered almost almost the entire the dynamicentire dynamic range ofrange the AEof the system AE fromsystem the fromthreshold the threshold level of 50level dB of to 50 90 dB dB. to The 90 dB. noise The activity noise activity reached reached about 55about AE 55 hits AE per hits second per secondon average. on average. Figure Figure 12 shows 12 shows the values the values of AE of impulseAE impulse amplitude amplitude and and cumulative cumulative AE AEhits hits obtained obtained within within an hour.an hour. More More than than 200,000 200,000 impulses impulses were were registered registered during during an houran hourof observation. of observation. Since Since the defectthe defect emissivity emissivity estimated estimated in Section in Section 2.1 was 2.1 was approximately approxi- 100mately hits 100 per hits loading per loading cycle, it cycle, was obviousit was obvious that most that AE most data AE were data caused were caused by noise by and noise this andcase this can case hide can the hide presence the presence of a dangerous of a dangerous defect. defect.

100 100 80 80 60 60 , dB

A 40 40 20 0 20 AE hits count rate 0 1000 2000 3000 0 t, s 0 1000 2000 3000 t, s

(a) (b)

Figure 12. AE data measured during structural health monitoring of walking dragline excavator: (a) amplitudes of AE Figure 12. AE data measured during structural health monitoring of walking dragline excavator: (a) amplitudes of AE impulsesimpulses inin aa periodperiod ofof anan hour,hour, ((bb)) AEAE hitshits countcount rate in a period of of an an hour. hour.

The most most radical radical noise noise filtering filtering result result was was reached reached with with the help the helpof the of location the location procedure.procedure. The Themain main source source of noise of noise was the was ropes the ropes movement, movement, and it and did itnot did correspond not correspond to a specificto a specific spatial spatial coordinate coordinate displayed displayed on the on “location”. the “location”. Figure Figure 13 presents 13 presents the linear the loca- linear tionlocation diagram diagram of AE of sources AE sources on the on right the right back back leg. leg.The Thelocation location diagram diagram was wasformed formed by twoby two antennas, antennas, the first the firstone included one included sensor sensor 1 and 1sensors and sensors 2, the second 2, the secondone—sensor one—sensor 2 and sensor2 and sensor3. AE sensors 3. AE sensorswere located were locatedat 1 m, 9 at m, 1 m,and 9 15 m, m and from 15 the m fromupper the edge upper of the edge bac of the bac leg, respectively. Figure 13 reveals the result of location of AE data measured during monitoring of the dragline in the period of a day. Figure 13a shows a location result without preliminary data filtering; out of 4,000,000 AE hits measured on average by each measuring channel, about 1.5%, of the order of 16,000 pulses, were located. Even though most of the noisy data was discarded, the location diagram displayed much noise due to the uniform filling which was not typical for real AE sources. Figure 13b depicts the location using the standard criterion for the location amplitude discrepancy. The AE signal pulse amplitude was recalculated taking into account the attenuation coefficient α from the registration point to the location point where the potential defect was located. Moreover, this amplitude did not exceed the permissible range of the AE sensors sensitivity (6 dB). Location amplitude filtration made it possible to exclude about 60% of false locations from the location diagram. The number of localized AE events decreased to 5125, while a localized cluster was formed in the center of the back leg in the welded joint (x = 8500–8800 mm). The increase in the number of indications along the location diagram edges had a noisy nature and it could be explained by friction between the fixing elements of the back leg. Appl. Sci. 2021, 11, x FOR PEER REVIEW 13 of 15

leg, respectively. Figure 13 reveals the result of location of AE data measured during monitoring of the dragline in the period of a day. Figure 13a shows a location result without preliminary data filtering; out of 4,000,000 AE hits measured on average by each measuring channel, about 1.5%, of the order of 16,000 pulses, were located. Even though most of the noisy data was discarded, the location diagram displayed much noise due to the uni- Appl. Sci. 2021, 11, 3420 13 of 15 form filling which was not typical for real AE sources.

(a) (b) (c)

FigureFigure 13. 13. ResultsResults of of AE AE event event location location on on a a right right back back leg leg of of a a dragline: dragline: (a (a) )without without filtering, filtering, (b (b) )result result of of the the amplitude amplitude filtering,filtering, (c (c) )result result of of the the spatial spatial parametric parametric filtering. filtering.

FigureParametric 13b depicts filtering the turned location out using to be the the standard most considerable criterion for approach the location to noise amplitude disposal. discrepancy.The result of The its application AE signal ispulse represented amplitude in Figurewas recalculated 13c. The number taking of into localized account AE the events at- tenuationdecreased coefficient to 1517, while α from both the uniform registration background point to the noise location and false point clusters where at the the potential location defectdiagram was edges located. were Moreover, removed this from amplitude the diagram. did not However, exceed the permissible number of events range thatof the in AEthe sensors location sensitivity cluster in the(6 dB). area Location of x ≈ 8800 amplitude mm did filtration not change. made it possible to exclude about To60% verify of false the locations reliability from of the the obtainedlocation diagram. results, ultrasonic The number testing of localized was carried AE events out in decreasedthe area of to the 5125, identified while a location localized (i.e., cluster in the was welded formed joint in ofthe the center right of back the leg).back Asleg a in result the weldedof the control,joint (x a= crack 8500–8800 in the weldedmm). The joint increase with a in length the number of 27 mm of was indications revealed. along It formed the locationdue to thediagram lack of edges root penetration.had a noisy nature and it could be explained by friction between the fixing elements of the back leg. 4. Discussion Parametric filtering turned out to be the most considerable approach to noise disposal.Within The result the frameworkof its application of this study,is repres theented possibility in Figure of the 13c. AE The structural number health of localized monitor- AEing events of a dragline decreased excavator to 1517, was while confirmed, both unif andorm a methodbackground for interpreting noise and false diagnostic clusters data at thewas location developed. diagram This methodedges were allows removed to identify from AE the signal diagram. pulses However, which are the responsible number forof eventsa defect that presence in the location despite cluster the background in the area of of intense x ≈ 8800 noise. mm did not change. ToThe verify distinctive the reliability feature of of the this obtained study, compared results, ultrasonic with other testing approaches was carried to AE out moni- in thetoring area ofof objectsthe identified with a location high noise (i.e., level, in the is welded that a priori joint of information the right back was leg). widely As a used result in ofthe the construction control, a crack of data in the processing welded joint algorithms. with a length At the of preliminary 27 mm was stagerevealed. of the It formed work, a duedetailed to the analysislack of root of all penetration. AE monitoring technology components was performed: the AE source parameters, the acoustic path characteristics, and the noise features. The cumulative 4.analysis Discussion of a priori information showed that the average noise level was several times higher than the average level of the AE pulse amplitudes. In addition, the intensity of the Within the framework of this study, the possibility of the AE structural health moni- noise exceeds the AE activity by more than an order of magnitude. This is a significant toring of a dragline excavator was confirmed, and a method for interpreting diagnostic obstacle for monitoring; however, the analysis of longtime noisy impact showed that its datamaximum was developed. level is about This method 80 dB. It allows also has to i andentify impulsive AE signal nature pulses with which quiet are periods responsi- up to ble40–45 for a dB, defect which presence is at least despite 50% the of thebackground time signal of realization.intense noise. Since the phase of the low noiseThe level distinctive corresponds feature to theof this phase study, of loading compared the dragline with other bucket, approaches it is possible to AE to moni- detect toringAE signal of objects pulses with against a high the noise background level, is of that noise. a priori information was widely used in the constructionThe AE diagnostics of data processing is based both algorith on settingms. At the thresholdpreliminary value stage of of the the amplitude work, a detaileddiscrimination analysis below of all theAE noisemonitoring level andtechnology on identification components of was the AEperformed: signal pulses the AE by classification of the signal pulses and noise impulse components. The classification is done with the use of an acoustic waveguide parametric model. Classification criteria are specific and allow to remove up to 97% of noise data that differ in parameters from the AE signal pulses. The main advantage of the proposed approach is that the whole extended structure can be controlled due to the low discrimination threshold. At this time, AE monitoring systems are now commonly used only for the monitoring of the regions with the existing Appl. Sci. 2021, 11, 3420 14 of 15

defects. The reliability of the obtained results is confirmed by the fact that a defect in the weld of the dragline back leg was determined based on the AE monitoring data. The results obtained in the study have an applied character and a certain practical significance. They are useful for application in the control or monitoring of the dragline metal components during operation. Within the framework of the research, four dragline excavators “E-Sh-20.90”, produced at the “Uralmash” plant from 1985 to 1994, were examined. For each of them, similar parameters of the noise process and close values of the acoustic waveguide parameters were observed. Despite the fact that defects were detected only in one of the monitored objects, the noise filtering procedure was effective in each case under consideration.

5. Conclusions In this paper, a new technique for detecting defects in metal structures of a dragline walking excavator in operating mode is proposed. The technique was implemented using the structural health monitoring system based on the AE method application. The main advantage of the proposed approach is the intelligent processing of diagnostic data based on the results of a preliminary study of defect parameters, acoustic waveguide characteristics, and noise process parameters. Taking into account this information made it possible to design a technique capable of detecting AE pulses corresponding to defects against a background of high-intensity noise. The effectiveness of the proposed approach is conﬁrmed by the detection of a defect in the back leg weld of an “ESH-20.90” dragline excavator.

Author Contributions: Conceptualization, V.B. (Vera Barat) and S.E.; methodology, V.B. (Vera Barat) and V.B. (Vladimir Bardakov); software, V.B. (Vladimir Bardakov); validation, V.B. (Vera Barat) and D.K.; formal analysis, S.U.; investigation, V.B. (Vladimir Bardakov) and D.K.; resources, D.K. and S.E.; data curation, V.B. (Vera Barat), M.K. (Marina Karpova), M.K. (Mikhail Kuznetsov), and A.Z.; writing—original draft preparation, V.B. (Vera Barat); writing—review and editing, A.M.; visualization, S.U.; supervision, S.E.; project administration, D.K. and S.E.; funding acquisition, V.B. (Vera Barat). All authors have read and agreed to the published version of the manuscript. Funding: The research was carried out within the framework of the project “Diagnostics of dissimilar welded joints of pearlitic and austenitic steels by the acoustic emission” with the support of a grant from NRU “MPEI” for implementation of scientific research programs “Energy”, “Electronics, Radio Engineering and IT”, and “Industry 4.0, Technologies for Industry and Robotics” in 2020–2022 (project No. 20/22-0000028/44). Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data presented in this study are available on request from the corresponding author. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Guericke, B.L.; Abramov, I.L.; Guericke, P.B. The strategy of preventive maintenance of mining machines based on the recognition of their actual technical condition. Min. J. 2008, 7, 70–80. 2. Dorosheev, Y.S.; Kirichuk, A.S. Problems of increasing the technological reliability of mining equipment. Min. Inf. Anal. Bull. 2010, 3, 176–182. 3. Chiementin, X.; Mba, D.; Charnley, B.; Lignon, S.; Dron, J.P. Effect of the denoising on acoustic emission signals. J. Vib. Acoust. 2010, 132, 310091–310099. [CrossRef] 4. Rastegaev, A.; Linderov, M.L.; Merson, D.L.; Afanasyev, M.A.; Danyuk, A.V. Monitoring of fracture of welded joints in hazardous facilities by acoustic emission under static and cyclic loadings. Indian J. Sci. Technol. 2015, 8, 1–7. [CrossRef] 5. Elizarov, S.V.; Barat, V.A.; Terentyev, D.A.; Kostenko, P.P.; Bardakov, V.V.; Alyakritsky, A.L.; Koltsov, V.G.; Troﬁmov, P.N. Acoustic Emission Monitoring of Industrial Facilities under Static and Cyclic Loading. Appl. Sci. 2018, 8, 1228. [CrossRef] 6. Cockerill, A.; Clarke, A.; Pullin, R.; Bradshaw, T.; Cole, P.; Holford, K.M. Determination of rolling element bearing condition via acoustic emission. J. Eng. Tribol. 2016, 230, 1377–1388. [CrossRef] Appl. Sci. 2021, 11, 3420 15 of 15

7. Beale, C.; Niezrecki, C.; Inalpolat, M. An adaptive wavelet packet denoising algorithm for enhanced active acoustic damage detection from wind turbine blades. Mech. Syst. Signal Process. 2020, 142, 106754. [CrossRef] 8. Bejger, A.; Drzewieniecki, J.B. The use of acoustic emission to diagnosis of fuel injection pumps of marine diesel engines. Energies 2019, 12, 4661. [CrossRef] 9. Johnson, M.; Ozevin, D.; Washer, G.; Ono, K.; Gostautas, R.; Tamutus, T. Acoustic emission method for real-time detection of steel fatigue crack in eyebar. Transp. Res. Rec. J. Transp. Res. Board 2012, 2313, 72–79. [CrossRef] 10. Nowak, M.; Lyasota, I.; Kisała, D. Testing the Node of a Railway Steel Bridge Using an Acoustic Emission Method. Advances in Acoustic Emission Technology. In Advances in Acoustic Emission Technology; Springer Proceedings in Physics; Springer: Cham, Switzerland, 2017; pp. 265–275. 11. Wang, H.; Huang, Z.; Liao, Z.; Jiang, B.; Wang, J.; Yang, Y.; Yang, B. From noise to information: A new technology of oleﬁn polymerization ﬂuidized bed reactor based on acoustic emission. J. Zhejiang Univ. Sci. A 2017, 18, 245–255. [CrossRef] 12. Anastasopoulos, A.; Kourousis, D.; Botten, S.; Wang, G. Acoustic emission monitoring for detecting structural defects in vessels and offshore structures. Ships Offshore Struct. 2009, 4, 363–372. [CrossRef] 13. Niccolini, G.; Lacidogna, G.; Carpinteri, A. Structural stability assessment of double girder crane by acoustic emission technique. In Proceedings of the ICF 2017—14th International Conference on Fracture, Rhodes, Greece, 18–23 June 2017; pp. 503–504. 14. Wu, Z.; Shen, G.; Zhang, J.; Tao, X.; Ni, D. Investigation of acoustic emission characteristics on harbor portal crane. In Advances in Acoustic Emission Technology; Springer Proceedings in Physics; Springer: Cham, Switzerland, 2017; pp. 243–250. 15. Xu, B.; Wu, Q. Stress fatigue crack propagation analysis of crane structure based on acoustic emission. Eng. Fail. Anal. 2020, 109, 104206. [CrossRef] 16. Nasonov, M.Y.; Lykov, Y.V.; Trong, D.D. The study of the resource and durability of metal structures of excavators after the expiration of the service life. Ugol 2020, 1127, 13–17. [CrossRef] 17. Muravin, G.; Muravin, B.; Lezvinsky, L. Acoustic emission non-destructive inspection method for cranes. Weld. World 2008, 52, 211–215. 18. PB 03-593-03. Regulations on Organization and Execution for Acoustic-Emission Inspection of Vessels, Boilers, Apparatus and Process Piping; Regulatory Document; PIO OBT Publishing: Moscow, Russia, 2003. (In Russian) 19. Barat, V.; Marchenkov, A.; Elizarov, S. Estimation of Fatigue Crack AE Emissivity Based on the Palmer–Heald Model. Appl. Sci. 2019, 9, 4851. [CrossRef] 20. Li, J.; Rose, J.L. Excitation and propagation of non-axisymmetric guided waves in a hollow cylinder. J. Acoust. Soc. Am. 2001, 109, 457–464. [CrossRef][PubMed] 21. Seco, F.; Jiménez, A.R. Modelling the generation and propagation of ultrasonic signals in cylindrical waveguides. In Ultrasonic Waves; State University of Campinas: Campinas, Brazil, 2012; pp. 1–28. 22. Viktorov, I. Rayleigh and Lamb Waves. Physical Theory and Applications, 1st ed.; Springer: Hawthorne, CA, USA, 1967; p. 154. 23. Barat, V.; Terentyev, D.; Bardakov, V.; Elizarov, S. Analytical Modeling of Acoustic Emission Signals in Thin-Walled Objects. Appl. Sci. 2020, 10, 279. [CrossRef]