sensors

Article Article TrainTrain Hunting Hunting Related Related FastFast DegradationDegradation ofof a Railway Crossing—Condition Monitoring and Crossing—Condition Monitoring and Numerical Numerical Verification Verification Xiangming Liu * and Valéri L. Markine * Xiangming Liu * and Valéri L. Markine * Department of Engineering Structures, Delft University of Technology, 2628 CN Delft, The Netherlands * DepartmentCorrespondence: of Engineering [email protected] Structures, Delft Univer (X.L.);sity [email protected] of Technology, Delft 2628 (V.L.M.) CN, The Netherlands * Correspondence: [email protected] (X.L.); [email protected] (V.L.M)


Received: 4 March 2020; Accepted: 15 April 2020; Published: 17 April 2020
 Received: 4 March 2020; Accepted: 15 April 2020; Published: date Abstract: This paper presents the investigation of the root causes of the fast degradation of a railway Abstract: This paper presents the investigation of the root causes of the fast degradation of a railway crossing. The dynamic performance of the crossing was assessed using the sensor-based crossing crossing. The dynamic performance of the crossing was assessed using the sensor-based crossing instrumentation, and the measurement results were verified using the multi-body system (MBS) instrumentation, and the measurement results were verified using the multi-body system (MBS) vehicle-crossing model. Together with the field inspections, the measurement and simulation results vehicle-crossing model. Together with the field inspections, the measurement and simulation indicate that the fast crossing degradation was caused by the high wheel-rail impact forces related results indicate that the fast crossing degradation was caused by the high wheel-rail impact forces to the hunting motion of the passing trains. Additionally, it was shown that the train hunting was related to the hunting motion of the passing trains. Additionally, it was shown that the train hunting activated by the track geometry misalignment in front of the crossing. The obtained results have was activated by the track geometry misalignment in front of the crossing. The obtained results have notnot only only explained explained thethe extremeextreme valuesvalues inin thethe measuredmeasured responses, but but also also shown shown that that crossing crossing degradationdegradation is is notnot alwaysalways caused by by the the problems problems in in the the crossing crossing itself, itself, but but can canalso also be caused be caused by byproblems problems in inthe the adjacent adjacent track track structures. structures. The Thefindings findings of this of thisstudy study were were implemented implemented in the in thecondition condition monitoring monitoring system system for for railway railway crossi crossings,ngs, using using which which timely timely and and correctly correctly aimed aimed maintenancemaintenance actions actions can can be be performed. performed.

Keywords:Keywords: railwayrailway crossing; crossing; wheel-rail wheel-rail impact; impact; train train hunting; hunting; numerical numerical verification; verification; railway railway track trackmaintenance maintenance

1.1. Introduction Introduction InIn the the railway railway track track system, system, turnouts turnouts (switches (switches and and crossings) crossings) are essentialare essential components components that allowthat trainsallow to trains pass to from pass one from track one to track another. to another. A standard A standard railway railway turnout turnout is composed is composed of three of three main main parts: switchparts: panel, switch closure panel, panel,closure and panel, crossing and crossing panel, as panel, shown as inshown Figure in1 .Figure In a railway 1. In a railway turnout, turnout, the crossing the panelcrossing is featured panel is to featured provide to the provide flexibility the flexib for trainsility for to passtrains in to di passfferent in different routes. routes.

FigureFigure 1. 1.Standard Standard left-handleft-hand railwayrailway turnout and the definition definition of of the the passing passing routes. routes.

ForFor rigid rigid crossings crossings that that are are commonlycommonly usedused inin conventionalconventional railway railway lines, lines, the the gap gap between between the the wingwing rail rail and and the the nose nose rail rail usually usually results results in in high high wheel-rail wheel-rail impacts impacts in in the the transition transition region region where where the wheelthe wheel load transitsload transits from thefrom wing the rail wing to therail noseto th raile nose (vice rail versa, (vice Figure versa,2 ),Figure which 2), makes which the makes crossing the a vulnerablecrossing a spot vulnerable in the railway spot in track.the railway In the track. case of In crossings the case thatof crossings are mainly that used are formainly the throughused for routethe trathroughffic (e.g., route crossings traffic in (e.g., the crossover),crossings in there the crossove is no specificr), there speed is no limit specific [1] and speed trains limit can [1] pass and through trains Sensors 2020, 20, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors

Sensors 2020, 20, 2278; doi:10.3390/s20082278 www.mdpi.com/journal/sensors Sensors 2020, 20, 2278 2 of 19 Sensors 2020, 20, x FOR PEER REVIEW 2 of 18 thecan crossings pass through with the a high crossings velocity with of a up high to velocity 140 km/ ofh. up The to high140 km/h. train The velocity high makestrain velocity the wheel-rail makes impactthe wheel-rail more serious. impact In more the Dutchserious. railway In the system,Dutch railway around system, 100 crossings around are100 urgently crossings replaced are urgently every yearreplaced [2] due every to unexpected year [2] due fatal to defects, unexpected which notfatal only defects, result which in substantial not only maintenance result in substantial efforts, but alsomaintenance lead to tra efforts,ffic disruption but also and lead can to traffic even adisruptionffect traffic and safety. can even affect traffic safety.

FigureFigure 2. 2.Wheel-rail Wheel-rail interactioninteraction in the railway crossing crossing for for through through route route traffic. traffic.

InIn contrast contrast to to a a switch switch panel, panel, whereinwherein sensorssensors are instrumented for for condition condition monitoring monitoring [3,4] [3,4 ] andand remaining remaining useful useful life life prediction prediction [[5],5], monitoringmonitoring in a crossing panel panel is is usually usually absent. absent. As As a aresult, result, thethe real-time real-time information information on on the the conditioncondition ofof railwayrailway crossingscrossings is limited. The The present present maintenance maintenance activitiesactivities are are mainly mainly reactive reactive and and based based on theon the experience experience of the of contractors.the contractors. In this In case,this case, the root thecauses root ofcauses the crossing of the degradationcrossing degradation are not alwaysare not resolved always resolved by the maintenance by the maintenance actions, andactions, the crossingsand the arecrossings likely to are be likely operated to be in operated a degraded in condition.a degraded To condition. improve To this improve situation, this necessary situation, guidance necessary for maintenanceguidance for actions maintenance is highly actions required. is highly required. ProperProper crossing crossing maintenance maintenance usuallyusually reliesrelies on condition assessment assessment and and degradation degradation detection, detection, whichwhich can can be be realized realized through through field fiel monitoring.d monitoring. In recentIn recent years, years, condition condition monitoring monitoring techniques techniques have beenhave frequently been frequently applied inapplied the railway in the industry. railway Aside industry. from theAside above-mentioned from the above-mentioned instrumentation oninstrumentation the turnout switches, on the turnout vehicle-based switches, monitoring vehicle-based systems monitoring have beensystems applied have inbeen track applied stiffness in measurementtrack stiffness [ 6measurement] and estimation [6] and [7], estimation track alignment [7], track estimation alignment [8 ],estimati hangingon sleepers[8], hanging detection sleepers [9 ], anddetection track fault [9], and detection track fault [10], detection etc. Compared [10], etc. with Compared the normal with track,the normal the current track, the studies current on studies railway crossingson railway are crossings mainly based are mainly on numerical based on simulation. numerical Typical simulation. contributions Typical contributions include wheel-rail include interaction wheel- analysisrail interaction [11–21], damageanalysis analysis[11–21], damage [16,17,22 analysis,23], and [16,17,22,23], prediction [18 and,24 ,prediction25] as well [18,24,25] as crossing as geometry well as andcrossing track stigeometryffness optimization and track stiffness for better optimizati dynamicon for performance better dynamic [16,26 performance]. Field measurements [16,26]. Field are mainlymeasurements used for theare validationmainly used of numericalfor the validation models. of The numerical monitoring models. of railway The monitoring crossings for of conditionrailway assessmentcrossings for and condition degraded assessment component and detection degraded is component still limited. detection is still limited. In the previous study, key indicators for the crossing condition assessment based on the field In the previous study, key indicators for the crossing condition assessment based on the field measurement were proposed [27,28]. Additionally, a numerical vehicle-crossing model was measurement were proposed [27,28]. Additionally, a numerical vehicle-crossing model was developed developed using a multi-body system (MBS) method to provide the fundamental basis for the using a multi-body system (MBS) method to provide the fundamental basis for the condition condition indicators [29]. In this study, the condition indicators, as well as the MBS model, were indicators [29]. In this study, the condition indicators, as well as the MBS model, were applied applied in the condition monitoring of a fast degraded railway crossing. The main goals of this study in the condition monitoring of a fast degraded railway crossing. The main goals of this study were were to investigate the root causes of the crossing degradation as well as to assess the effectiveness to investigate the root causes of the crossing degradation as well as to assess the effectiveness of the of the current maintenance actions. current maintenance actions. Based on the objectives, this paper is presented in the following order. The experimental and numericalBased ontools, the including objectives, the this crossing paper condition is presented indicators, in the followingare briefly order.introduced The experimentalin Section 2. The and numericalmeasurement tools, results including and the the crossing crossing degradation condition anal indicators,ysis as well are as briefly the effectiveness introduced of inthe Section current2 . Themaintenance measurement actions results are presented and the crossing in Sections degradation 3 and 4. Based analysis on the as wellmeasurement as the eff resultsectiveness and offield the currentinspections, maintenance the root actions causes arefor the presented fast crossing in Sections degradation3 and4. were Based investigated on the measurement with the assistance results and of fieldthe inspections,MBS model, the as rootpresented causes in for Section the fast 5. crossing In Sectio degradationn 6, the verification were investigated of the effectiveness with the assistance of the ofmaintenance the MBS model, actions as is presented given. Finally, in Section in Section5. In Section7, major6 ,conclusions the verification are provided. of the e ffectiveness of the maintenance actions is given. Finally, in Section7, major conclusions are provided. 2. Methodology 2. Methodology In this section, the experimental tools for the crossing condition monitoring, as well as the indicatorsIn this for section, the crossing the experimental condition assessment, tools for the are crossing briefly introduced. condition monitoring,The MBS vehicle-crossing as well as the indicatorsmodel for for the the verification crossing of condition the experi assessment,mental findings are briefly is also introduced.presented. The MBS vehicle-crossing model for the verification of the experimental findings is also presented.

Sensors 2020, 20, 2278 3 of 19 Sensors 2020, 20, x FOR PEER REVIEW 3 of 18 Sensors 2020, 20, x FOR PEER REVIEW 3 of 18 2.1. Experimental Tools 2.1. Experimental Tools The experimental tools mainly consisted of the in-sitein-site instrumentationinstrumentation system modifiedmodified from ESAH-MThe (Elektronische experimental toolsSystem mainly Analyse consisted Herzstückb of theereich-Mobil) in-site instrumentation and the video system gauge modified system from(VGS) ESAH-MESAH-M (Elektronische (Elektronische System System AnalyseAnalyse Herzstückbereich-Mobil)Herzstückbereich-Mobil) and and the the video video gauge gauge system system (VGS) (VGS) for wayside monitoring, as briefly described below. Both tools have already been introduced and forfor wayside wayside monitoring, monitoring, asas brieflybriefly describeddescribed below. Both Both tools tools have have already already been been introduced introduced and and actively applied in previous studies. Detailed information regarding the installation and data activelyactively applied applied in previousin previous studies. studies. Detailed Detailed information information regarding regarding the installation the installation and data and processing data processing can be found in [27,30]. canprocessing be found can in [be27 ,found30]. in [27,30]. 2.1.1. Crossing Instrumentation 2.1.1.2.1.1. Crossing Crossing Instrumentation Instrumentation The main components of the crossing instrumentation are an accelerometer attached to the TheThe main main components components of theof the crossing crossing instrumentation instrumentation are anare accelerometer an accelerometer attached attached to the crossingto the crossing nose rail for 3-D acceleration measurement, a pair of inductive sensors attached in the nosecrossing rail for nose 3-D rail acceleration for 3-D acceleration measurement, measurement, a pair of inductive a pairsensors of inductive attached sensors in the attached closure panel in the for trainclosureclosure detection panel panel for asfor well traintrain as detectiondetection train velocity asas wellwell calculation, as train andvelocityvelocity the main calculation,calculation, unit for and dataand the collection.the main main unit unit An for overview for data data ofcollection.collection. the instrumented An An overview overview crossing of of thethe is instrumentedinstrumented shown in Figure crossing3. isis shownshown in in Figure Figure 3. 3.

FigureFigure 3.3. CrossingCrossing instrumentation based based on on ESAH-M. ESAH-M.

TheThe mainmain main outputsoutputs outputs of ofof the thethe crossing crossingcrossing instrumentation instrumentatio werenn werewere the thethe dynamic dynamicdynamic responses responses responses of theof of the crossing the crossing crossing nose, includingnose,nose, including including the wheel-rail thethe wheel-railwheel-rail impact accelerationsimpactimpact acceleration and locations,ss andand locations,locations, etc. All these etc.etc. All responsesAll thesethese responses wereresponses calculated were were withincalculatedcalculated the within transitionwithin the the transition region,transition which region,region, can which be obtained can be obtainedobtained through through through field inspection field field inspection inspection [29]. Based[29]. [29]. Based Based on these on on measuredthesethese measured measured responses responsesresponses and the andand correlation thethe correlationcorrelation analysis analanal betweenysisysis betweenbetween the responses thethe responsesresponses [28], two [28], critical[28], two two condition critical critical indicatorsconditioncondition indicators related indicators to therelated related wheel toto impactthethe wheelwheel and impact fatigue and area, fatiguefatigue respectively, area,area, respectively, respectively, were proposed. were were proposed. proposed. TheThe wheel wheelwheel impact impactimpact is is reflectedis reflectedreflected by theby verticalthe vertical accelerations, accelerations,accelerations, which whichwhich were obtained werewere obtained obtained from the from from crossing the the andcrossingcrossing processed and and processed processed through through statisticalthrough statisticalstatistical analysis. analysis. This indicator ThisThis indicatorindicator is mainly is is mainly mainly based based onbased the on on magnitude the the magnitude magnitude of the impactsofof the the impacts impacts due to eachdue due to passingto each each passingpassing wheel (Figure wheelwheel4 (Figurea), and the4a),), changes andand thethe inchangeschanges time indicate in in time time theindicate indicate di fferent the the different conditiondifferent stagesconditioncondition of the stages stages crossing of of the the (Figure crossing crossing4b). (Figure(Figure 4b).

Figure 4. Indicator for the wheel impact. (a) Procedure for the obtainment of wheel impacts. (b) FigureExample 4.4.Indicator Indicatorof the variation for for the the wheelof wheelthe impact.wheel impact. impacts (a) Procedure(a) inProcedure different for the conditionfor obtainment the obtainment stages. of wheel of impacts. wheel impacts. (b) Example (b) ofExample the variation of the variation of the wheel of the impacts wheel in impacts different in conditiondifferent condition stages. stages. The fatigue area is defined as the region where the majority of wheel impacts are located on the crossing,The fatigue and where area area is isultimately defined defined as asthe the the crack region region initiates where where (Figure the the majority majority 5a). In of practice, ofwheel wheel impacts the impacts fatigue are arelocated area located can on be the on thecrossing,simplified crossing, and as and thewhere confidence where ultimately ultimately interval the theofcrack [a crack – σinitiates, a initiates+ σ], where (Figure (Figure a is 5a). the5 a). meanIn Inpractice, value practice, of the the the fatigue wheel-rail fatigue area area impact can can be simplifiedlocations, asand the σ confidenceis the standard interval deviation. of [a – Theσ, a +location σ], where and a issize the of mean the fatigue value ofarea the are wheel-rail critical values impact locations, and σ is the standard deviation. The location and size of the fatigue area are critical values

Sensors 2020, 20, 2278 4 of 19

beSensors simplified 2020, 20, x as FOR the PEER confidence REVIEW interval of [a σ, a + σ], where a is the mean value of the wheel-rail4 of 18 Sensors 2020, 20, x FOR PEER REVIEW − 4 of 18 impact locations, and σ is the standard deviation. The location and size of the fatigue area are critical valuesfor the assessment for the assessment of crossing ofcrossing wear and wear plastic and deformation. plastic deformation. A wide fatigue A wide area fatigue usually area represents usually well-maintainedrepresents well-maintained rail geometry. rail geometry. As demonstrated As demonstrated in Figure in 5b, Figure when5b, whenthe crossing the crossing condition condition was wasdegraded degraded from from “Worn” “Worn” to “Damaged”, to “Damaged”, the the fatigue fatigue area area was was dramatically dramatically narrowed narrowed and and shifted further fromfrom thethe theoreticaltheoretical point point (TP) (TP) of of the the crossing. crossing. More More information information about about the the fatigue fatigue area area can can be befound found in thein the previous previous study study [27 ].[27].

Figure 5. DemonstrationDemonstration of of the the crossing crossing fatigue fatigue area area detection. detection. (a) (Definitiona) Definition of the of thefatigue fatigue area. area. (b) (Exampleb) Example of the of thefatigue fatigue area area changes changes in different in different crossing crossing condition condition stages. stages.

2.1.2. Wayside Monitoring System The VGS for wayside monitoring is a remote measurement device based on digital image The VGS for wayside monitoring is a remote measurement device based on digital image correlation (DIC).(DIC). It It uses uses high-speed high-speed digital digital cameras cameras to measure to measur the dynamice the dynamic movements movements of the selected of the selectedtargets in targets the track. in the The track. system, The set system, up together set up withtogether the targets with the installed targets on installed the crossing on the rail crossing next to rail the nextinstrumented to the instrumented accelerometer, accelerometer, is shown in Figureis shown6a, andin Figure the demo 6a, and of the the displacement demo of themeasurement displacement measurementis shown in Figure is shown6b. The in Figure main outputs 6b. The aremain the outp verticaluts are displacements the vertical displacements of the tracked targetsof the tracked with a targetsstable sampling with a stable frequency sampling of up frequency to 200 Hz. of up to 200 Hz.

Figure 6. Wayside monitoring. ( a) System setup. ( (bb)) The The screen screen of of displacement displacement measurements.

Due to the limitation of the experimental conditions, the wayside monitoring system is usually set Due to the limitation of the experimental conditions, the wayside monitoring system is usually up close by the side of the track, which will introduce extra noise in the measured displacement results. set up close by the side of the track, which will introduce extra noise in the measured displacement To improve the accuracy of the measurement, the noise part needs to be eliminated. The noise mainly results. To improve the accuracy of the measurement, the noise part needs to be eliminated. The noise comes from the ground-activated camera vibration, which can be manually activated by hammering mainly comes from the ground-activated camera vibration, which can be manually activated by the ground near the camera. The measured camera vibrations in both the time and frequency domains hammering the ground near the camera. The measured camera vibrations in both the time and are given in Figure7. frequency domains are given in Figure 7.

Sensors 2020,, 20,, 2278x FOR PEER REVIEW 5 of 1918 Sensors 2020, 20, x FOR PEER REVIEW 5 of 18

Figure 7. Ground activated camera vibration. (a) Time domain signal. (b) Frequency domain Figureresponses. 7.7.Ground Ground activated activated camera camera vibration. vibration. (a) Time (a) domainTime domain signal. (signal.b) Frequency (b) Frequency domain responses. domain responses. Despite the didifferencesfferences in the displacement responsesresponses in the two monitored crossings, the main resonanceDespite of the the differences cameracamera vibrationvibration in the displacement waswas aroundaround resp 15–4515–45onses Hz.Hz. in theIn In the two previous monitored study crossings, [30], [30], thethe main resonancecomponents of inthe thethe camera displacementdisplacement vibration signalsignal was werearoundwere elaborated.el aborated.15–45 Hz. TheIn thetrain-track previous components study [30], related the main to componentsdisplacement in responses the displacement are mainly signal distributed were belo belowelaborated.w 10 Hz, The which train-track do not overlapcomponents with therelated camera to vibrationdisplacement introduced introduced responses noise. noise. are The mainly The noise noise distributed part part due due to belo camera tow camera 10 vibrationHz, vibration which can do canthen not then beoverlap reduced be reduced with through the through camera low- vibrationlow-passpass filtering, filtering,introduced as shown as shownnoise. in Figure The in Figure noise 8. 8part. due to camera vibration can then be reduced through low- pass filtering, as shown in Figure 8.

Figure 8. Examples of the measured rail vertical displacement. Figure 8. Examples of the measured rail vertical displacement. The magnitude of the dynamic vertical displacement of the rail directly reflectsreflects the intensity of the track The movementmagnitude dueof the to dynamic the passing vertical trains. displacement By By comparing comparing of thethe measured rail directly rail reflects displacement the intensity with the of thereference track movement level,level, which which due can can to be the obtainedbe passing obtained from trains. from numerical By numerical comparing simulation simulation the measured using theusing parametersrail the displacement parameters in the designed with in the referencecondition,designed condition,level, the ballast which the settlement canballast be settlementobtained level of the fromlevel monitored ofnumerical the monitored location simulation can location be estimated.using can bethe estimated. Theparameters MBS modelThe in MBS the for designedthemodel crossing for condition, the performance crossing the performance ballast analysis settlement is analysis described level is laterdescribedof the in monitored this later section. in locationthis section. can be estimated. The MBS model for the crossing performance analysis is described later in this section. 2.2. Multi-Body System (MBS) Vehicle-Crossing Model 2.2. Multi-Body System (MBS) Vehicle-Crossing Model The numericalnumerical model model for for the the crossing crossing performance performance analysis analysis was developedwas developed using theusing MBS the method MBS VI-RailmethodThe (FigureVI-Rail numerical9 a).(Figure The model rail9a). pads,forThe the rail clips, crossing pads, and clips, ballast performance and were ballast simulated analysis were simulated as was spring developed andas spring damping using and elementsdampingthe MBS (railmethodelements busing VI-Rail (rail and busing (Figure base and busing, 9a). base The Figure bu railsing, pads,9b). Figure In clips, the 9b). vehicleand In ballast the model, vehicle were the model,simulated car body, the ascar bogie spring body, frames andbogie damping and frames the wheelsetselementsand the wheelsets (rail were busing modeled were and modeled as base rigid bu bodies sing,as rigid Figure with bodies both 9b). with the In primarythe both vehicle the suspension primary model, thesuspension and car secondary body, and bogie suspensionsecondary frames andtakensuspension the into wheelsets account taken into (Figurewere account modeled9b). (Figure The as track rigid 9b). model Thebodies track was with amodel straight both was the line a primarystraight with the linesuspension crossing with the panel and crossing (Figuresecondary panel9c) suspensionsituated(Figure 9c) in thesituatedtaken middle into in accountthe of the middle track. (Figure of The the 9b). railtrack. The element Thetrack rail model for element the was acceleration fora straight the acceleration and line displacement with theand crossing displacement extraction panel was(Figureextraction the 9c) lumped was situated the rail lumped in mass the middle located rail mass of 0.3 the located m track. from 0.3 theThe m TPrail from of element the the crossing TP for of thethe (Figure accelerationcrossing9d), (Figure which and displacement is9d), consistent which is withextractionconsistent the setup withwas of thethe the lumpedsetup field of measurements rail the massfield measurementslocated (Figures 0.3 m3 andfr (Figuresom6a). the TP3 and of the6a). crossing (Figure 9d), which is consistent with the setup of the field measurements (Figures 3 and 6a).

Sensors 2020, 20, x FOR PEER REVIEW 6 of 18 Sensors 2020, 20, 2278 6 of 19 Sensors 2020, 20, x FOR PEER REVIEW 6 of 18

Figure 9. Multi-body system (MBS) model. ( a)) Vehicle-track model. ( (b)) Flexible connections in the model.Figure ( c9.)) Crossing CrossingMulti-body profiles. profiles. system ( (d (MBS))) Rail Rail element elementmodel. (for a) Vehicle-trackacceleration extraction. model. (b) Flexible connections in the model. (c) Crossing profiles. (d) Rail element for acceleration extraction. The detailed model development, experimental validation, and numerical verification verification can be found The in the detailed previous model study development, [[29].29]. Corresponding experimental to the validation, condition and indicators, numerical the mainverification outputs can of be the MBSfound modelmodel in the areare previous thethe wheel wheel study impact impact [29]. acceleration, acceleration,Corresponding transition transi to thetion regioncondition region and and indicators, wheel-rail wheel-rail the contact contactmain forces. outputs forces. Using of Using the the MBStheMBS MBS model, model model, theare thethe condition conditionwheel impact of of the the acceleration, monitored monitored crossing, transi crossitionng, as regionas well well and as as the thewheel-rail detecteddetected contact tracktrack forces.degradations, Using canthe be MBS verified.verified. model, the condition of the monitored crossing, as well as the detected track degradations, can be verified. 3. Field Measurements and Analysis 3. Field Measurements and Analysis The monitoredmonitored crossing crossing was was a casta cast manganese manganese crossing crossing with anwith angle an ofangle 1:9. Asof part1:9. ofAs a crossover,part of a The monitored crossing was a cast manganese crossing with an angle of 1:9. As part of a trainscrossover, mainly trains pass mainly the crossing pass the incrossing the facing in the through facing routethrough (Figure route2 )(Figure with a 2) velocity with a ofvelocity around of crossover, trains mainly pass the crossing in the facing through route (Figure 2) with a velocity of 140around km/ h.140 The km/h. on-site The view on-site of the view crossing of the is crossi shownng in is Figure shown 10 ina. AccordingFigure 10a. to According the maintenance to the around 140 km/h. The on-site view of the crossing is shown in Figure 10a. According to the record,maintenance this crossing record, wasthis crossing suffering was from suffering fast degradation from fast withdegradation the service with life the of service only around life of threeonly maintenance record, this crossing was suffering from fast degradation with the service life of only yearsaround (18 three years years on average (18 years [2]). on At average the beginning [2]). At of the the conditionbeginning monitoring,of the condition the damaged monitoring, crossing the around three years (18 years on average [2]). At the beginning of the condition monitoring, the was completely renovated. damageddamaged crossing crossing was was completely completely renovated.renovated.

FigureFigure 10. 10. Overview Overview of of the the monitoredmonitored crossing. ( (aa))) On-siteOn-site On-site view.view. view. ( (b (b)) )Sketch Sketch Sketch view. view.

FigureFigure 1010b 10bb gives gives a a sketch sketch view view of of thethe crossing, crossing, including the the setupsetup of of the the monitoring monitoring devices devices and and thethe layoutlayout layout ofof of the the the adjacent adjacent adjacent structures, structur structures,es, especially especiallyespecially the the small small bridge bridge in frontinin frontfront of theof of the crossing.the crossing. crossing. Considering Considering Considering that thethatthat bridge the the bridge bridge is located is is located located quite quite quite close close close to the to to monitoredthethe monitomonitored crossing, crossing, the thethe performance performancperformanc ofee of theof the the crossing crossing crossing might might might be abeffbe ectedaffected affected by theby by the bridge,the bridge, bridge, which wh which willich will will be discussed bebe discusseddiscussed later. later. TheThe measurementmeasurement measurement results resultsresults from fromfrom the the crossingthe crossingcrossing instrumentation instrumentation were basedwerewere based onbased multiple onon multiplemultiple train passages train train inpassagespassages one monitoring in in one one monitoring monitoring day. For the day. day. wayside ForFor thethe monitoring, waysidewayside mo onenitoring, sufficient oneone train sufficientsufficient passage train train is passage enoughpassage is to is enough estimateenough thetoto estimate ballastestimate condition. the the ballast ballast To condition. condition. maximally ToTo reduce maximallymaximally the influence redureduce the of influence the vehicle-related ofof thethe vehicle-related vehicle-related variables, the variables, variables, selected resultsthethe selected selected were results restricted results were were to therestricted restricted commonly toto thethe operated commonlycommonly VIRM operated trains VIRMVIRM with velocities trainstrains with with of velocities aroundvelocities 140 of of around km around/h. 140140 km/h. km/h. 3.1. Wheel Impacts 3.1. Wheel Impacts 3.1. WheelBased Impacts on the estimated transition regions, the wheel impact accelerations were calculated. Based on the estimated transition regions, the wheel impact accelerations were calculated. The The distributionBased on the ofestimated the wheel transition impacts regions, due to the multiple wheel impact wheel passagesaccelerations is shown were calculated. in Figure The11a. distribution of the wheel impacts due to multiple wheel passages is shown in Figure 11a. The sample distribution of the wheel impacts due to multiple wheel passages is shown in Figure 11a. The sample

Sensors 2020, 20, 2278 7 of 19 SensorsSensors 20202020,, 2020,, xx FORFOR PEERPEER REVIEWREVIEW 77 ofof 1818 size,Thesize, sample inin thisthis case, size,case, inwaswas this 7878 case, passingpassing was wheels.wheels. 78 passing ItIt cancan wheels. bebe seenseen It can thatthat be thethe seen wheelwheel that impactsimpacts the wheel presentedpresented impacts aa presented bimodalbimodal distribution.adistribution. bimodal distribution. AroundAround 80%80% Around ofof thethe wheel 80%wheel of impactsimpacts the wheel werewere impacts belowbelow were 5050 g,g, below whilewhile 50thethe g, remainingremaining while the 20%20% remaining ofof thethe wheel20%wheel of impactsimpacts the wheel werewere impacts extremelyextremely were highhigh extremely withwith aa high meanmean with valuvalu aee mean ofof aroundaround value 350350 of g. aroundg. SuchSuch aa 350 polarizedpolarized g. Such distributiondistribution a polarized ofdistributionof impactsimpacts indicatesindicates of impacts thethe indicateshighlyhighly unstableunstable the highly wheel-railwheel-rail unstable interactioninteraction wheel-rail inin interactionthisthis crossing.crossing. in It thisIt waswas crossing. demonstrateddemonstrated It was indemonstratedin aa previousprevious studystudy in a previous [29][29] thatthat study forfor thisthis [29 typetype] that ofof for rarailwayilway this type crossing,crossing, of railway thethe averageaverage crossing, levellevel the ofof average thethe wheelwheel level impactimpact of the iswheelis aroundaround impact 5050 g,g, is meaningmeaning around 50 thatthat g, the meaningthe 20%20% ofof that highhigh the impactsimpacts 20% of ofof high thetheimpacts monitoredmonitored of the crossingcrossing monitored areare alreadyalready crossing moremore are thanalreadythan sevenseven more timestimes than higherhigher seven thanthan times thethe higher averageaverage than impactimpact the average level.level. impactItIt cancan bebe level. imaginedimagined It can thatthat be imagined suchsuch highhigh that impactsimpacts such high willwill dramaticallyimpactsdramatically will dramaticallyaccelerateaccelerate thethe accelerate degradationdegradation the procedure degradationprocedure ofof procedure thethe crossing.crossing. of the crossing.

FigureFigure 11.11.11. Vertical VerticalVertical acceleration accelerationacceleration responses responsesresponses of the ofof monitoredthethe monitoredmonitored crossings. crossings.crossings. (a) Distribution ((aa)) DistributionDistribution based on basedbased multiple onon multipletrainmultiple passages traintrain passagespassages in one day. inin one (oneb) Example day.day. ((bb)) Example ofExample impacts ofof due impactsimpacts to one duedue bogie. toto oneone bogie.bogie.

AnAn example example of of the the impact impact acceleration acceleration response response inin the the time-domain time-domain duedue toto thethe first firstfirst bogie bogie of of a a VIRMVIRM traintraintrain is isis shown shownshown in inin Figure FigureFigure 11 11b.b.11b. It canItIt cancan be be seenbe sseeneen that thatthat for theforfor twothethe two passingtwo passingpassing wheels wheelswheels from from thefrom same thethe bogie,samesame bogie,thebogie, impacts thethe impactsimpacts can be can quitecan bebe di quitequitefferent. different.different. The impact TheThe impaimpa due toctct theduedue front toto thethe wheel frontfront was wheelwheel up was towas 350 upup g, toto while 350350 g,g, the whilewhile rear thewheelthe rearrear activated wheelwheel activatedactivated vertical acceleration verticalvertical accelerationacceleration was only waswas 20g. onlyonly It has 2020 tog.g. beItIt hashas noted toto bebe that notednoted thehigh thatthat impactsthethe highhighwere impactsimpacts not werealwayswere notnot introduced alwaysalways introducedintroduced by the front byby wheel, thethe frontfront but appeared wheel,wheel, butbut to haveappearedappeared random toto havehave occurrences. rarandomndom Such occurrences.occurrences. results further SuchSuch resultsconfirmedresults furtherfurther the instabilityconfirmedconfirmed of thethe wheel-rail instabilityinstability interaction ofof wheel-railwheel-rail at this interactioninteraction crossing. atat thisthis crossing.crossing.

3.2.3.2. FatigueFatigue AreaArea TheThe measuredmeasured fatiguefatiguefatigue area areaarea of ofof the thethe monitored monitoredmonitored crossing crossingcrossing is presentedisis presentedpresented in Figureinin FigureFigure 12. 12. It12. can ItIt can becan seen bebe seenseen that thatthethat wheel thethe wheelwheel impacts impactsimpacts were werewere widely widelywidely distributed distributeddistributed at 0.22–0.38 atat 0.22–0.380.22–0.38 m from mm fromfrom the thth TPee TPTP with withwith the thethe fatigue fatiguefatigue area areaarea size sizesize of of0.16of 0.160.16 m. m. Accordingm. AccordingAccording to the toto previousthethe previousprevious study studystudy [28], the[28],[28], transition thethe transitiontransition region regionregion (Figure (Figure(Figure2) for this 2)2) typeforfor thisthis of crossing typetype ofof crossingiscrossing around isis 0.15–0.4 aroundaround m.0.15–0.40.15–0.4 The fatiguem.m. TheThe areafatiguefatigue widely areaarea coveredwidelywidely coveredcovered 64% of the64%64% transition ofof thethe transitiontransition region, region,region, which canwhichwhich be canconsideredcan bebe consideredconsidered to be in toto line bebe with inin lineline the withwith expectation thethe expectationexpectation of a new crossingofof aa newnew profile. crossingcrossing Such profile.profile. results SuchSuch further resultsresults confirmed furtherfurther confirmedthatconfirmed the crossing thatthat thethe rail crossingcrossing was not rara wornilil waswas or notnot deformed. wornworn oror deformed.deformed.

FigureFigure 12. 12. MeasuredMeasured fatiguefatigue areaarea of of the the monitored monitored crossing. crossing.

ItIt has has to to be be noted noted that that the the fatigue fatigue area area does does not not conform conform to to the the normal normal distribution distribution (referring (referring to to thethe “Worn” “Worn” stagestagestage demonstrateddemonstrateddemonstrated in inin Figure FigureFigure5 b).5b).5b). Combined CombCombinedined with withwith the thethe results resultsresults of ofof the thethe wheel wheelwheel impacts impactsimpacts such suchsuch a afatiguea fatiguefatigue area areaarea further furtherfurther confirmed confirmedconfirmed the thethe instability instabilityinstability of ofof the thethe wheel-rail wheel-railwheel-rail contact contactcontact in inin the thethe monitored monitoredmonitored crossing. crossing.crossing. InIn aa previousprevious studystudy [27],[27], itit waswas foundfound thatthat thethe crossingcrossing degradationdegradation waswas accompaniedaccompanied byby thethe increaseincrease ofof wheel-railwheel-rail impactsimpacts andand thethe reductionreduction inin thethe fatiguefatigue area.area. TheThe largelarge numbernumber ofof extremelyextremely highhigh wheel-railwheel-rail impactsimpacts andand relativelyrelatively widewide fatigufatiguee areaarea clearlyclearly indicateindicate thethe abnormalabnormal performanceperformance

Sensors 2020, 20, 2278 8 of 19

In a previous study [27], it was found that the crossing degradation was accompanied by the increaseSensors 2020 of, 20 wheel-rail, x FOR PEER impacts REVIEW and the reduction in the fatigue area. The large number of extremely8 of 18 high wheel-rail impacts and relatively wide fatigue area clearly indicate the abnormal performance ofof thethe monitoredmonitored crossing.crossing. FindingFinding thethe rootroot causescauses ofof suchsuch abnormalityabnormality isis thethe keykey toto improvingimproving thethe dynamicdynamic performanceperformance ofof thethe crossing.crossing.

3.3.3.3. BallastBallast SettlementSettlement TheThe measured vertical displacement displacement of of the the crossing crossing rail rail is ispresented presented in inFigure Figure 13. 13It .can It be can seen be seenthat thatthe vertical the vertical rail raildisplacement displacement was was around around 4 4mm. mm. The The measured measured displacement displacement result result can bebe consideredconsidered to to have have two two main main parts: parts: the the elastic elastic deformation deformation and theand gap the between gap between the sleeper the sleeper and ballast. and Consideringballast. Considering that the that ballast the settlement ballast settlement is the accumulated is the accumulated effect due effect to multiple due to wheelmultiple passages, wheel thepassages, plastic deformationthe plastic deformation caused by each caused passing by each train passing can be neglected.train can be Due neglected. to the high Due impacts to the in high the crossingimpacts panel,in the thecrossing ballast panel, is usually the settledballast unevenly,is usually whichsettled results unevenly, in hanging which sleepers.results in Using hanging the validatedsleepers. Using MBS model,the validated it was MBS calculated model, thatit was the calcul rail displacementated that the rail in displacement the reference in condition the reference was 1.4condition mm (Figure was 1.4 13 mm), which (Figure only 13), consisted which only of the consisted elastic of deformation the elastic deformation part. By comparing part. By comparing these two results,these two it could results, be calculatedit could be that calculated the gap that between the gap the sleeperbetween and the ballast sleeper was and 2.6 ballast mm, whichwas 2.6 can mm, be estimatedwhich can as be the estimated settlement as ofthe ballast. settlement It was of observed ballast. thatIt was the observed rail displacement that the obtainedrail displacement from the MBSobtained simulation from the was MBS much simulation higher than was that much in a normalhigher trackthan (lessthat thanin a normal 1 mm [27 track,31]), (less which than indicates 1 mm the[27,31]), vulnerability which indicates of the ballast the vulnerability in the railway of crossings.the ballast in the railway crossings.

FigureFigure 13.13. BallastBallast settlementsettlement inin thethe monitoredmonitored crossing.crossing.

InIn aa previousprevious studystudy [[27],27], itit waswas foundfound thatthat tracktrack irregularitiesirregularities suchsuch asas railrail jointsjoints andand turnoutturnout crossingscrossings cancan leadlead toto thethe fastfast deteriorationdeterioration ofof thethe ballast,ballast, andand thethe ballastballast settlementsettlement willwill inin turnturn accelerateaccelerate thethe degradationdegradation procedureprocedure ofof otherother relatedrelated tracktrack components.components. In this study, thethe 2.62.6 mmmm ballast settlement was already higher than those in the previously monitored welded joints ( 1.5 mm) ballast settlement was already higher than those in the previously monitored welded joints (≈≈1.5 mm) and movable crossings ( 2 mm), which revealed the seriously deteriorated ballast condition. and movable crossings (≈≈2 mm), which revealed the seriously deteriorated ballast condition. ItIt can can be be concluded concluded that that the monitored the monitored crossing crossing was suffering was su fromffering rapidly from occurring, rapidly occurring,extremely extremelyhigh wheel-rail high wheel-rail impacts and impacts severe and ballast severe settl ballastement. settlement. For a recently For a recently renovated renovated crossing, crossing, such suchperformance performance is quite is quite abnormal. abnormal.

4.4. EEffectivenessffectiveness AnalysisAnalysis ofof thethe MaintenanceMaintenance ActionsActions TheThe constantly constantly occurring occurring extremely extremely high high wheel-rai wheel-raill impacts impacts as well as as serious well as ballast serious settlement ballast settlementclearly indicate clearly the indicate degraded the condition degraded of the condition crossing. of In the order crossing. to improve In order such toa situation, improve various such a situation,maintenance various actions maintenance were implemented actions were in this implemented location including in this location ballast includingtamping, fastening ballast tamping, system fasteningrenovation, system etc. In renovation, this section, etc. the In thiseffectiveness section, the of etheffectiveness maintenance of the actions maintenance are briefly actions discussed, are briefly as discussed,presented asbelow. presented below. 4.1. Ballast Tamping 4.1. Ballast Tamping Considering that the crossing rail was lately renovated with limited wear or plastic deformation, Considering that the crossing rail was lately renovated with limited wear or plastic deformation, the severe ballast settlement was suspected to be the main cause for the high wheel-rail impacts. the severe ballast settlement was suspected to be the main cause for the high wheel-rail impacts. Therefore, ballast tamping actions were frequently performed in this location by the local contractor. However, due to the lack of maintenance facilities, the tamping actions were mainly performed using the squeezing machine (Figure 14a) without track geometry correction. It can be imagined that the settled ballast cannot be fully recovered with such tamping action. As shown in Figure 14b, after tamping, the rail displacement was not dramatically reduced.

Sensors 2020, 20, 2278 9 of 19

Therefore, ballast tamping actions were frequently performed in this location by the local contractor. However, due to the lack of maintenance facilities, the tamping actions were mainly performed using the squeezing machine (Figure 14a) without track geometry correction. It can be imagined that the settled ballast cannot be fully recovered with such tamping action. As shown in Figure 14b,

SensorsafterSensors tamping, 2020 2020, ,20 20, ,x x FOR theFOR railPEER PEER displacement REVIEW REVIEW was not dramatically reduced. 99 of of 18 18

FigureFigure 14.14. ((aa)) SqueezingSqueezing machinemachine usedused forfor ballastballast tampingtamping inin thethe monitoredmonitoredmonitored crossing.crossing.crossing. ((bb)) MeasuredMeasured railrail displacement displacement before before andand afterafter ballastballast tamping. tamping.

TheThe development development of of the the wheel-rail wheel-rail impacts impactsimpacts before beforebefore and andand after after tamping tamping are are presented presented in in Figure Figure 15.1515.. InIn this this figure,figure,figure, eacheach pointpoint representsrepresents thethe meanmean valuevalue ofof thethe impactimpactimpact accelerationsaccelerations basedbased onon multiplemultiple wheelwheel passagespassages in inin one oneone monitoring monitoringmonitoring day. day.day. It ItIt was waswas discussed discusdiscussedsed in inin a apreviousa previousprevious study studystudy [28 [28][28]] that thatthat the thethe fluctuation fluctuationfluctuation of oftheof the the wheel wheel wheel impacts impacts impacts was was was highly highly highly aff affected affectedected by by by external external external disturbances disturbances disturbances such such such as as as the the the weather.we weather.ather. Still,Still, Still, itit cancan be be seenseen thatthat thethe regressionregression valuesvalues beforebefore andan andd afterafter tamping tampingtamping were werewere both bothboth around aroundaround 100100 100 g. g.g.

FigureFigure 15. 15. Development DevelopmentDevelopment of of the the wheel-rail wheel-rail impacts impacts before before and and after after ballast ballast tamping. tamping.

ItIt cancan bebe concludedconcluded thatthat suchsuch frequentlyfrequently implementedimplemented ballastballast tampingtamping hadhad nono improvementimprovement inin eithereither thethe ballast ballastballast condition conditioncondition or or or the thethe dynamic dynamicdynamic performance perforperformancemance of of theof thethe monitored monitoredmonitored crossing. crossing.crossing. Without WithoutWithout figuring figuring figuring out outtheout rootthethe causesrootroot causescauses for the forfor fast thethe crossing fastfast crossingcrossing degradation, degradation,degradation, such ine ffsuchsuchective ineffectiveineffective ballast tamping ballastballast should tampingtamping be suspended.shouldshould bebe suspended.suspended. 4.2. Fastening System Renovation 4.2.4.2. Fastening FasteningDuring the System System monitoring RenovationRenovation period, the fastening system was found to be degraded with some broken bolts.DuringDuring Such degradationthethe monitoring monitoring aff period, ectedperiod, the the the lateral fastening fastening stability system system of was was the found track.found to to Therefore, be be degraded degraded the with fasteningwith some some system,broken broken bolts.mainlybolts. SuchSuch the bolts degradationdegradation in the guard affectedaffected rails the andthe laterallateral the clips, stabilstabil wasityity renovated, ofof thethe track.track. as shown Therefore,Therefore, in Figure thethe fastening fastening16. system,system, mainlymainly thethe boltsbolts inin thethe guardguard railsrails and and thethe clips,clips, waswas renovated,renovated, as as shown shown inin FigureFigure 16. 16.

Sensors 2020, 20, 2278 10 of 19 Sensors 2020, 20, x FOR PEER REVIEW 10 of 18

Figure 16.16. Fastening system renovation. ( aa)) Remove Remove the the broken bolts. ( (bb)) Reposition Reposition the the guard guard rail. rail. ((cc)) InstallInstall newnew bolts.bolts.

The development of the wheel-rail impacts before and after renovation is is shown in in Figure Figure 17.17. The upperupper figure figure is is the the development development of theof meanthe mean value, value, and theand lower the figurelower givesfigure the gives ratio the of diratiofferent of impactdifferent levels impact in eachlevels monitoring in each monitoring day, corresponding day, corresponding to the value to inthe the value upper in the figure. upper figure.

Figure 17. EffectEffect of fastening system renovation on the dynamic performance ofof thethe crossing.crossing.

It cancan bebe seenseen fromfrom FigureFigure 1717 thatthat beforebefore thethe renovation,renovation, thethe wheel-railwheel-rail impactimpact showed a clearclear increasingincreasing trendtrend withwith thethe impactimpact valuesvalues widelywidely distributeddistributed fromfrom 00 toto 450450 g.g. Such a degradation trend indicatesindicates thatthat maintenancemaintenance is urgentlyurgently requiredrequired duedue toto thethe defectsdefects ofof thethe fasteningfastening system.system. AfterAfter thethe renovation, thethe wheel-rail wheel-rail impacts impacts were were dramatically dramatically reduced reduced in terms in ofterms the mean of the value mean and value separated and intoseparated two distribution into two modes,distribution which modes, is similar which to those is similar shown into Figure those 11showna. Such in improvement Figure 11a. isSuch due toimprovement the enhancement is due to in the the enhancement track integrity. in However,the track integrity. the wheel-rail However, impacts the wheel-rail above 300 impacts g were above still a large300 g proportionwere still a after large maintenance, proportion after which maintenance, means that thewhich sources means for that such the high sources wheel-rail for such impacts high werewheel-rail not found. impacts were not found. In practice,practice, ballast tamping is currently one of thethe fewfew optionsoptions forfor contractorscontractors toto maintainmaintain thethe track. However,However, the the unimproved unimproved crossing crossing performance performance clearly clearly indicates indicates the ine fftheectiveness ineffectiveness of tamping. of Thetamping. fastening The systemfastening renovation system renovation was a forced was action a forced to repair action damaged to repair components. damaged Althoughcomponents. the crossingAlthough performance the crossing wasperformanc improved,e was the improved, extremely the high extremely wheel-rail high impacts wheel-rail were notimpacts reduced, were thus not thereduced, sources thus for the the sources fast crossing for the fast degradation crossing degr wereadation not eliminated. were not eliminated. To figure To out figure the rootout the causes root forcauses the for crossing the crossing damage, damage, the track the inspection track inspecti wason extended was extended to the bridge to the in bridge front ofin thefront crossing of the (Figurecrossing 10 (Figureb). The 10b). results The for results the track for the inspection, track inspection, as well as as the well numerical as the numerical verification verification using the using MBS model,the MBS are model, presented are presented in the next insection. the next section.

5. Damage Sources Investigation In this section, the track inspection, including the whole turnout and the adjacent bridge, is presented. The inspected degradations will be input into the MBS model to verify the influence on

Sensors 2020, 20, 2278 11 of 19

5. Damage Sources Investigation In this section, the track inspection, including the whole turnout and the adjacent bridge, isSensors presented. 2020, 20 The, x FOR inspected PEER REVIEW degradations will be input into the MBS model to verify the influence11 onof 18 the crossing performance. As a reference, the dynamic responses in the designed condition with no track the crossing performance. As a reference, the dynamic responses in the designed condition with no degradations were also simulated and compared with those in degraded conditions. The verification track degradations were also simulated and compared with those in degraded conditions. The results, followed by the analysis, are also presented. verification results, followed by the analysis, are also presented. 5.1. Track Inspection 5.1. TrackSensors Inspection 2020, 20, x FOR PEER REVIEW 11 of 19 In the field inspection, it was found that the bridge was not well aligned in the track, but deviated In thethe field crossing inspection, performance. it was As afound reference, that the the dyna bridgemic responseswas not wellin the aligned designed in condition the track, with but no deviated around 15 cm, as shown in Figure 18a. Such deviation introduced a curve into the track, which was around 15track cm, degradations as shown inwere Figure also simulated18a. Such and deviation compared introduced with those ain curve degraded into conditions. the track, The which was likelylikely to to be verificationbe out out of of design results,design sincefollowed since no no by elevation elevation the analysis, waswas are also setset uppresented. in the outer outer rail. rail. It It can can be be imagined imagined that that the the passingpassing trains trains could could not not pass pass the the tracktrack alongalong thethe centralcentral line line but but tended tended to to have have wheel wheel flange flange contact contact 5.1. Track Inspection withwith the the outer outer rail, rail, which which eventually eventually leads leads toto thethe severesevere wear in the switch switch blade blade (Figure (Figure 18b). 18b). In the field inspection, it was found that the bridge was not well aligned in the track, but deviated around 15 cm, as shown in Figure 18a. Such deviation introduced a curve into the track, which was likely to be out of design since no elevation was set up in the outer rail. It can be imagined that the passing trains could not pass the track along the central line but tended to have wheel flange contact with the outer rail, which eventually leads to the severe wear in the switch blade (Figure 18b).

FigureFigure 18. 18.Track Track deviationdeviation in front front of of the the crossing. crossing. (a) (Inspecteda) Inspected curve curve introd introduceduced by the by bridge. the bridge. (b) (bWorn) Worn switch switch rail. rail. (c) ( cDemonstration) Demonstration of the of the bridge bridge deviation. deviation. Figure 18. Track deviation in front of the crossing. (a) Inspected curve introduced by the bridge. (b) TheThe accumulated accumulatedWorn switch e ffrail.effectect (c of) Demonstration of the the track track deviation of deviation the bridge was deviation. was also also reflected reflected in in the the variated variated track track gauge. gauge. It wasIt shownwas shown in the measurementin the measurement results results that the that gauge the gaug variationse variations along along the whole the whole turnout turnout were were up to up 3 mm,to The accumulated effect of the track deviation was also reflected in the variated track gauge. It as3 presented mm, aswas presented inshown Table in 1thein. ConsideringTablemeasurement 1. Considering results that thethat thatthe monitored gaug thee monitored variations crossing along crossing is the located whole is located turnout quite closewerequite up toclose to the to bridge the (Figurebridge 18 (Figure3c), mm, such as 18c), presented track such misalignment, intrack Table misalignment, 1. Considering including that including th thee monitored track the deviationtrack crossing deviation is inlocated the in bridgequite the close bridge and to the trackand track gauge variationgauge variation alongbridge the(Figure along turnout, 18c), the such turnout, may track aff misalignment,ectmay the affect wheel-rail the incl wheel-railuding interaction the track interaction deviation in the crossing.inin thethe bridgecrossing. and track gauge variation along the turnout, may affect the wheel-rail interaction in the crossing. Table 1. Track gauge measurement results in the critical sections along the turnout. Table 1.TableTrack 1. gaugeTrack gauge measurement measurement results results in the the critical critical sections sections along alongthe turnout. the turnout.

LocationLocationLocation A B A A CBB C DC D D E E E F F F G G G Deviation (mm) +2 +3 2 2 +2 +3 0 − − DeviationDeviation (mm) (mm) +2 +2 +3 +3 −2− 2 −2 −2 +2 +2 +3 +30 0 5.2. Numerical Verification and Analysis 5.2. Numerical Verification and Analysis 5.2.In Numerical order to Verification verify the eandffect Analysis of the track lateral misalignment on the performance of the crossing, both the bridge-introducedIn order to verify curve the effect and of the thetrack track lateral gauge misalignment variation were on the input performance into the ofMBS the crossing, vehicle-crossing In orderboth the to verifybridge-introduced the effect curveof the and track the lateratrack gaugel misalignment variation were on theinput performance into the MBS ofvehicle- the crossing, modelboth (Figurethecrossing bridge-introduced9). model The equivalent (Figure 9).curve The track equivalent and lateral the track irregularities lateral gauge irregularities variation as the modelas were the model input input into are are shownthe shown MBS in in Figurevehicle- 19. crossingIn theFigure model MBS 19. model, (Figure the 9). crossingThe equivalent type is track the same lateral as irregularities the monitored as 1:9 the crossing model input with are the shown rail type in of UIC54Figure E1. 19. The vehicle model is consistent with the recorded VIRM train with the wheel profile of S1002. The initial track parameters of Dutch railways [32] applied in the model are given in Table2.

Sensors 2020, 20, x FOR PEER REVIEW 11 of 18

the crossing performance. As a reference, the dynamic responses in the designed condition with no track degradations were also simulated and compared with those in degraded conditions. The verification results, followed by the analysis, are also presented.

5.1. Track Inspection In the field inspection, it was found that the bridge was not well aligned in the track, but deviated around 15 cm, as shown in Figure 18a. Such deviation introduced a curve into the track, which was likely to be out of design since no elevation was set up in the outer rail. It can be imagined that the passing trains could not pass the track along the central line but tended to have wheel flange contact with the outer rail, which eventually leads to the severe wear in the switch blade (Figure 18b).

Figure 18. Track deviation in front of the crossing. (a) Inspected curve introduced by the bridge. (b) Worn switch rail. (c) Demonstration of the bridge deviation.

The accumulated effect of the track deviation was also reflected in the variated track gauge. It was shown in the measurement results that the gauge variations along the whole turnout were up to 3 mm, as presented in Table 1. Considering that the monitored crossing is located quite close to the bridge (Figure 18c), such track misalignment, including the track deviation in the bridge and track gauge variation along the turnout, may affect the wheel-rail interaction in the crossing.

Table 1. Track gauge measurement results in the critical sections along the turnout.

Location A B C D E F G

Deviation (mm) +2 +3 −2 −2 +2 +3 0

5.2. Numerical Verification and Analysis In order to verify the effect of the track lateral misalignment on the performance of the crossing, both the bridge-introduced curve and the track gauge variation were input into the MBS vehicle- Sensorscrossing2020 ,model20, 2278 (Figure 9). The equivalent track lateral irregularities as the model input are shown12 of in 19 Figure 19.

Sensors 2020, 20, x FOR PEER REVIEW 12 of 18

Figure 19. Equivalent lateral irregularities in the track.

In the MBS model, the crossing type is the same as the monitored 1:9 crossing with the rail type

of UIC54 E1. The vehicle model is consistent with the recorded VIRM train with the wheel profile of

S1002. The initial track parametersFigure 19. ofEquivalent Dutch railways lateral irregularities [32] applied in in the the track. model are given in Table 2.

Table 2.2. Track parameters.

TrackTrack Components Components Stiffness,Stiffness, MN MN/m/m Damping, Damping, kN kN·s/ms/m · VerticalVertical 13001300 45 45 RailRail pad/Clips pad/Clips Lateral 280280 580 580 Roll 360360 390 390 BallastBallast Vertical Vertical & laterallateral 45 45 32 32

With the track misalignment taken into account, the crossing condition was considered as degraded.With the The track simulation misalignment results taken of into both account, wheel thes crossingin the bogie, condition including was considered the wheel as degraded. impact Theaccelerations simulation and results transition of both regions, wheels were in thecompared bogie, with including the results the wheel in the impactdesigned accelerations condition [29], and transitionas shown regions, in Figure were 20. compared with the results in the designed condition [29], as shown in Figure 20.

FigureFigure 20. 20.Vertical Vertical impact impact acceleration acceleration responses responses and and transition transition regions. regions. (a) Front (a) Front wheel. wheel. (b) Rear (b) wheel.Rear wheel. It can be seen from Figure 20a that with the lateral irregularity taken into account, the impact of theIt front can be wheel seen was from dramatically Figure 20a that increased with the to late 247ral g, irregularity which was 4taken times into higher account, than the the impact reference of valuethe front (around wheel 62 was g) in dramatically the designed increased condition. to 247 While g, which for the was rear 4 wheeltimes higher from thethan same the bogie,reference the impactvalue (around was 48 g,62 whichg) in the was designed even lower condition. than the While reference for the value. rear wheel Despite from the the slight same di bogie,fference the in theimpact absolute was values,48 g, which the simulationwas even lower results than were the consistent reference withvalue. the Despite measurement the slight results difference (Figure in the 11 ). Meanwhile,absolute values, the transition the simulationregion ofresults the front were wheel consistent was 0.176–0.182 with the measurement m from the TP results with a(Figure size of 11). only 0.006Meanwhile, m. Compared the transition with the region reference of the levelfront (0.196–0.217wheel was 0.176–0.182 m with a sizem from of 0.031 the TP m, with [29]), a itsize was of muchonly narrower0.006 m. andCompared closer towith the the TP, reference indicating level earlier (0.196–0.217 wheel impact m with and a much size of sharper 0.031 m, wheel [29]), load it was transition much innarrower the crossing. and closer For the to rear the wheel,TP, indicating although earlier the transition wheel impact region wasand locatedmuch sharper farther wheel from the load TP, thetransition size was in almost the crossing. the same For as the the rear reference wheel, although value. the transition region was located farther from the TP,Such the results size was clearly almost show the that same the as curve the reference and lateral value. track misalignment in front of the crossing can lead toSuch unstable results wheel-rail clearly show contact that inthe the curve crossing and late andral sometimes track misalignment result in extremelyin front of highthe crossing impacts. Additionally,can lead to theunstable front andwheel-rail rear wheels contact pass in through the crossing the crossing and sometimes quite differently, result whichin extremely indicates high that theimpacts. performance Additionally, of the rearthe front wheel and is not rear independent, wheels pass butthrough is aff ectedthe crossing by the frontquite wheel.differently, which indicatesFor the that wheel-rail the performance contact forces, of the rear thetendency wheel is not was independent, similar to the but acceleration is affected responses,by the front as wheel. shown in FigureFor 21the. Withwheel-rail the degraded contact trackforces, condition, the tendency the maximum was similar contact to the force acceleration of the front responses, wheel in theas degradedshown in conditionFigure 21. wasWith 468 the kN,degraded which track was twicecondition, as high the asmaximum that in the contact designed force condition of the front (235 wheel kN). in the degraded condition was 468 kN, which was twice as high as that in the designed condition (235 kN). While for the rear wheel, the difference between the degraded condition and the designed condition was limited.

Sensors 2020, 20, 2278 13 of 19

While for the rear wheel, the difference between the degraded condition and the designed condition Sensors 2020, 20, x FOR PEER REVIEW 13 of 18 wasSensors limited. 2020, 20, x FOR PEER REVIEW 13 of 18

FigureFigure 21. 21. VerticalVertical Vertical wheel-rail wheel-rail contact contact responsesresponses ofof thethe facingfacing crossing.crossing. ( a) Front wheel. wheel. ( (bb)) RearRear Rear wheel. wheel. wheel.

ToTo understand understand how howhow the the track track track misalignment misalignment misalignment affectsaffects affects the the wheel-rail wheel-rail interaction interaction inin thethe in crossing, thecrossing, crossing, the the relationshiptherelationship relationship between between between the the thewheel wheel wheel laterallateral lateral displacemedisplaceme displacementsntsnts and and wheel-rail wheel-rail contact contact forcesforces forces werewere were analyzed.analyzed. analyzed. BeforeBefore that,that, that, thethe the wheelwheel wheel lateral lateral lateral displacement displacement displacement in inin the thethe designed dedesignedsigned condition condition is presentedis presented in Figure inin FigureFigure 22. 22. 22. When When When the thetrainthe train train enters enters enters the the crossingthe crossing crossing panel, panel, panel, the the the variated variated variated rail railrail geometry geometrygeometry will will lead lead to to the thethe lateral laterallateral movementmovement of of the the wheel.wheel. The The maximum maximum lateral lateral disp displacementdisplacementlacement waswas around around 0.70.7 mm.

FigureFigure 22. 22. Wheel WheelWheel lateral lateral displacement displacement in the designed condition.

InIn the thethe degraded degraded condition condition condition with with with track track track lateral lateral lateral irreguirregu irregularities,larities,larities, the lateral the lateral displacementsdisplacements displacements of of the the wheels wheels of the wheelswerewere dramatically dramatically were dramatically changed, changed, changed,as as shown shown as inin shownFigureFigure 23.23. in FigureItIt can be 23 seen. It that can both be seen thethe frontfront that wheel bothwheel the andand front thethe wheelrearrear wheel wheel and theshowed showed rear wheelactivated activated showed hunting hunting activated oscillationoscillation hunting bebeforefore oscillation andand after before passing and throughthrough after passing thethe crossing,crossing, through butbut the thecrossing,the trajectories trajectories but the were were trajectories quite quite different. different. were quiteFor For thethe di ff fronfronerent.tt wheel,wheel, For thethethe lateral front wheel, movement the lateral waswas moremore movement intenseintense and wasand ranmoreran toward toward intense the the and crossing crossing ran toward nose nose rail therail crossingnearnear thethe TP.noseTP. TheThe rail maximummaximum near the TP. lateral The maximumdisplacement lateral correspondingcorresponding displacement to to the position with the highest contact force was 2.3 mm, which means that compared with that in the thecorresponding position with to the positionhighest contact with the force highest was contact 2.3 mm, force which was means 2.3 mm, that which compared means with that that compared in the designed condition, the wheel flange was around 1.6 mm closer to the nose rail. Comparatively withdesigned that condition, in the designed the wheel condition, flange thewas wheel around flange 1.6 wasmm aroundcloser to 1.6 the mm nose closer rail. toComparatively the nose rail. speaking, such displacement of the rear wheel was only 0.3 mm. Such results indicate that the wheel- speaking,Comparatively such displacement speaking, such of displacement the rear wheel of was the rearonly wheel0.3 mm. was Such only results 0.3 mm. indicate Such resultsthat the indicate wheel- rail impact was profoundly affected by the movement of the wheel. When the wheel approaches railthat impact the wheel-rail was profoundly impact was affected profoundly by the a ffmoectedvement by the of movementthe wheel. ofWhen the wheel. the wheel When approaches the wheel closer to the crossing nose, the wheel-rail impact is likely to be increased. It can be concluded that the closerapproaches to the crossing closer to nose, the crossing the wheel-rail nose, impact the wheel-rail is likely impactto be increased. is likely It to can be be increased. concluded It that can the be train hunting activated by the lateral track misalignment in front of the crossing is the main cause of trainconcluded hunting that activated the train by hunting the lateral activated track bymisalignme the lateralnt track in front misalignment of the crossing in front is the of themain crossing cause of is the extremely high wheel-rail impacts. the extremely main cause high of the wh extremelyeel-rail impacts. high wheel-rail impacts. The train hunting effect also explains the unstable wheel-rail impacts. For the rear wheel, the lateral movement was affected not only by the track misalignment but also by the front wheel from the same bogie. As a result, these two wheels led to quite different wheel trajectories. It can be imagined that in the real-life situation, there are much more factors that may affect the hunting motion of each passing wheelset such as the initial position of the wheel when entering the misaligned track section, the mutual interaction between the adjacent wheelsets, the lateral resistance of the track, and even the weather condition [28], etc. The combined effect of all these factors ultimately resulted in the polarized distribution of the impact acceleration responses (Figure 11a).

Figure 23. Wheel-rail contact forces and wheel lateral displacements. (a) Front wheel. (b) Rear wheel. Figure 23. Wheel-rail contact forces and wheel lateral displacements. (a) Front wheel. (b) Rear wheel.

Sensors 2020, 20, x FOR PEER REVIEW 13 of 18

Figure 21. Vertical wheel-rail contact responses of the facing crossing. (a) Front wheel. (b) Rear wheel.

To understand how the track misalignment affects the wheel-rail interaction in the crossing, the relationship between the wheel lateral displacements and wheel-rail contact forces were analyzed. Before that, the wheel lateral displacement in the designed condition is presented in Figure 22. When the train enters the crossing panel, the variated rail geometry will lead to the lateral movement of the wheel. The maximum lateral displacement was around 0.7 mm.

Figure 22. Wheel lateral displacement in the designed condition.

In the degraded condition with track lateral irregularities, the lateral displacements of the wheels were dramatically changed, as shown in Figure 23. It can be seen that both the front wheel and the rear wheel showed activated hunting oscillation before and after passing through the crossing, but the trajectories were quite different. For the front wheel, the lateral movement was more intense and ran toward the crossing nose rail near the TP. The maximum lateral displacement corresponding to the position with the highest contact force was 2.3 mm, which means that compared with that in the designed condition, the wheel flange was around 1.6 mm closer to the nose rail. Comparatively speaking, such displacement of the rear wheel was only 0.3 mm. Such results indicate that the wheel- railSensors impact 2020, 20 was, x FOR profoundly PEER REVIEW affected by the movement of the wheel. When the wheel approaches14 of 18 closer to the crossing nose, the wheel-rail impact is likely to be increased. It can be concluded that the The train hunting effect also explains the unstable wheel-rail impacts. For the rear wheel, the trainSensors hunting2020, 20, 2278activated by the lateral track misalignment in front of the crossing is the main cause14 of of 19 lateral movement was affected not only by the track misalignment but also by the front wheel from the extremely high wheel-rail impacts. the same bogie. As a result, these two wheels led to quite different wheel trajectories. It can be imagined that in the real-life situation, there are much more factors that may affect the hunting motionSensors 2020 of, 20each, x FOR passing PEER REVIEW wheelset such as the initial position of the wheel when entering14 ofthe 18 misaligned track section, the mutual interaction between the adjacent wheelsets, the lateral resistance of theThe track, train and hunting even effectthe weather also explains condition the unstable[28], etc. wheel-rail The combined impacts. effect For ofthe all rear these wheel, factors the ultimatelylateral movement resulted was in the affected polarized not onlydistribution by the tr ofack the misalignment impact acceleration but also responses by the front (Figure wheel 11a). from the same bogie. As a result, these two wheels led to quite different wheel trajectories. It can be 5.3.imagined Respective that Effect in the of Lateralreal-life Curve situation, or Track there Gauge are Deviation much more factors that may affect the hunting motion of each passing wheelset such as the initial position of the wheel when entering the It can be noticed that in the previous analysis, the input track misalignment consisted of two misaligned track section, the mutual interaction between the adjacent wheelsets, the lateral resistance parts: the lateral curve introduced by the bridge and the track gauge deviation. In order to understand of the track, and even the weather condition [28], etc. The combined effect of all these factors the effect of each part in the wheel-rail interaction, these two parts were further analyzed, and the ultimately resulted in the polarized distribution of the impact acceleration responses (Figure 11a). results are presented below. Figure 23. Wheel-rail contact forces and wheel lateral displacements. (a) Front wheel. (b) Rear wheel. FigureConsidering 23. Wheel-rail the bridge-introduced contact forces and wheellateral lateral curve, displacements. the wheel-rail (a) Frontcontact wheel. forces (b) Rearand wheel.the lateral 5.3. Respective Effect of Lateral Curve or Track Gauge Deviation 5.3.wheel Respective displacements Effect of were Lateral calculated, Curve or Trackas presented Gauge Deviation in Figure 24. It can be seen that in the front wheel, the bridge-introducedIt can be noticed thatcurve in mainlythe previous resulted analysis in the, thelateral input shift track of themisalignment wheel trajectory consisted due ofto twothe It can be noticed that in the previous analysis, the input track misalignment consisted of two parts: centripetalparts: the lateral force. curve Such introduced a shift was by only the bridge0.5 mm an dne thear thetrack crossing gauge deviation. nose when In ordercompared to understand with the the lateral curve introduced by the bridge and the track gauge deviation. In order to understand the designedthe effect condition,of each part and in the the effect wheel-rail on the interactwheel impaction, these was two limited. parts For were the further rear wheel, analyzed, the combined and the effect of each part in the wheel-rail interaction, these two parts were further analyzed, and the results effectresults of are the presented curve and below. the motion of the front wheel resulted in the lateral deviation of 0.9 mm, which are presented below. was quiteConsidering close to thethat bridge-introduced in the designed condition lateral curve, and had the nowheel-rail significant contact influence forces on and the thewheel-rail lateral Considering the bridge-introduced lateral curve, the wheel-rail contact forces and the lateral impact.wheel displacements were calculated, as presented in Figure 24. It can be seen that in the front wheel, wheelthe bridge-introduced displacements were curve calculated, mainly resulted as presented in the in Figurelateral 24shift. It of can the be wheel seen that trajectory in the front due wheel,to the thecentripetal bridge-introduced force. Such curvea shift mainly was only resulted 0.5 mm in thenear lateral the crossing shift of thenose wheel when trajectory compared due with to thethe centripetaldesigned condition, force. Such and a shift the waseffect only on the 0.5 mmwheel near impact the crossing was limited. nose when For the compared rear wheel, with the the combined designed condition,effect of the and curve the and effect the on motion the wheel of the impact front wheel was limited. resulted For in thethe rearlateral wheel, deviation the combined of 0.9 mm, eff whichect of thewas curve quite andclose the to motionthat in the of the designed front wheel condition resulted and inhad the no lateral significant deviation influence of 0.9 on mm, the which wheel-rail was quiteimpact. close to that in the designed condition and had no significant influence on the wheel-rail impact.

Figure 24. Wheel-rail contact forces and lateral wheel displacements. (a) Front wheel. (b) Rear wheel.

The effect of the track gauge deviation on the wheel-rail interaction is demonstrated in Figure 25. Different from the effect of the bridge-introduced curve, the deviated track gauge activated the hunting motion of the passing wheels. Still, the resulted lateral wheel displacements were not large enough to amplify the wheel-rail impact. The maximum displacements corresponding to the wheel impactsFigureFigure were 24.24. 1 Wheel-rail mm in the contact front forces wheel andand and laterallateral 0.4 mm wheelwheel indisplacements. displacements.the rear wheel, ((aa ))respectively. FrontFront wheel.wheel. ((b)) RearRear wheel.wheel.

The effect of the track gauge deviation on the wheel-rail interaction is demonstrated in Figure 25. Different from the effect of the bridge-introduced curve, the deviated track gauge activated the hunting motion of the passing wheels. Still, the resulted lateral wheel displacements were not large enough to amplify the wheel-rail impact. The maximum displacements corresponding to the wheel impacts were 1 mm in the front wheel and 0.4 mm in the rear wheel, respectively.

Figure 25. Wheel-rail contact forces and wheel lateral displacements. (a) Front wheel. (b) Rear wheel.

Sensors 2020, 20, 2278 15 of 19

The effect of the track gauge deviation on the wheel-rail interaction is demonstrated in Figure 25. Different from the effect of the bridge-introduced curve, the deviated track gauge activated the hunting motionSensors 2020 of, the 20, x passing FOR PEER wheels. REVIEW Still, the resulted lateral wheel displacements were not large enough15 of to18 amplify the wheel-rail impact. The maximum displacements corresponding to the wheel impacts were 1 mmFigure in the 25. front Wheel-rail wheel andcontact 0.4 forces mm inand the wheel rear lateral wheel, displacements. respectively. (a) Front wheel. (b) Rear wheel.

5.4.5.4. Summary BasedBased onon thethe aboveabove analysis, analysis, it it can can be be concluded concluded that that the the extremely extremely high high wheel-rail wheel-rail impacts impacts in the in monitoredthe monitored crossing crossing were were caused caused by the by hunting the huntin oscillationg oscillation of the passingof the passing trains. Such trains. train Such hunting train washunting the combinedwas the combined effect of the effect bridge-introduced of the bridge-int curveroduced in front curve of the in crossing front of and the thecrossing deviated and track the gaugedeviated along track the gauge turnout. along When the turnout. the maximum When th wheele maximum lateral displacementwheel lateral displacement reaches a certain reaches level a (e.g.,certain 2.3 level mm), (e.g., the wheel-rail2.3 mm), the impact wheel-rail will be impact dramatically will be amplified.dramatically amplified. ItIt hashas toto bebe notednoted thatthat althoughalthough thethe curvecurve inin frontfront ofof thethe crossingcrossing diddid notnot directlydirectly activateactivate traintrain hunting,hunting, the activated lateral shift of the passing wheels resulted in the wearwear inin thethe switchswitch bladeblade (Figure(Figure 1818b)b) andand contributedcontributed toto thethe tracktrack gaugegauge deviation.deviation. Therefore, such a curve can be considered asas thethe rootroot causecause of of the the fast fast degradation degradation of of the the monitored monitored crossing. crossing. To To improve improve the the performance performance of the of crossing,the crossing, this this curve curve has toha bes to first be first eliminated. eliminated. InIn thethe previousprevious studystudy [[28],28], itit waswas provenproven thatthat highhigh railrail temperaturetemperature duedue toto thethe longlong durationduration ofof sunshinesunshine wouldwould amplifyamplify thethe existingexisting tracktrack geometrygeometry deviation in turnout and lead to the increase inin thethe wheel-railwheel-rail impacts.impacts. The train hunting activated by the track gauge deviation inin thisthis studystudy further confirmedconfirmed thesethese results.results.

6.6. EEffectffect ofof Maintenance-RelatedMaintenance-Related DegradationDegradation AccordingAccording toto the the measurement measurement results, results, the monitored the monitored crossing crossing also suff alsoered fromsuffered ballast from settlement ballast andsettlement broken and clips. broken In order clips. to betterIn order simulate to better the simula real-lifete situation,the real-life these situation, track defects these track were defects respectively were addedrespectively to the degradedadded to the MBS degraded model developed MBS model in Sectiondeveloped 5.2. in The Section combined 5.2. The effects combined were simulated effects were and analyzed,simulated asand presented analyzed, below. as presented below.

6.1.6.1. EEffectffect of Ballast Settlement ItIt isis shownshown inin FigureFigure 1313 thatthat thethe detecteddetected ballastballast settlementsettlement waswas around 2.62.6 mm.mm. To simplify the problem,problem, aa verticalvertical irregularityirregularity was introduced in the MBS model to simulate the ballast settlement, asas shownshown inin FigureFigure 2626.. In this irregularityirregularity function,function, thethe amplitudeamplitude waswas 1.31.3 mm,mm, andand thethe wavelengthwavelength waswas 1010 m.m. The trough of the wave was located 0.3 m from the TP of the crossing, which was consistent withwith thethe instrumentedinstrumented accelerometeraccelerometer andand thethe installedinstalled displacementdisplacement target.target.

FigureFigure 26.26. Ballast settlement introduced inin thethe MBSMBS model.model.

WithWith thethe ballastballast settlementsettlement takentaken intointo accountaccount inin thethe MBSMBS model,model, thethe dynamicdynamic performanceperformance ofof thethe crossingcrossing waswas simulated.simulated. The representative resultsresults areare shownshown inin FigureFigure 2727.. ItIt can be seen thatthat the simulationsimulation results were almostalmost thethe samesame asas thosethose withoutwithout ballastballast settlementsettlement (Figure(Figure 2323),), despitedespite thethe slightlyslightly increasedincreased impact impact force force of of the the front front wheel wheel (from (from 468 468 kN kN to 487to 487 kN). kN). It can It can be concludedbe concluded that that the existencethe existence of ballast of ballast settlement settlement had a had limited a limite influenced influence on the dynamicon the dynamic performance performance of the crossing. of the Fromcrossing. another From point another of view, point the of ballast view, settlementthe ballast wassettlement more likely was more to be thelikely accumulated to be the accumulated effect of the effect of the high wheel-rail impacts. Such results further explain the ineffectiveness of the frequently performed ballast tamping since ballast settlement is not the main cause of the extremely high wheel- rail impacts.

Sensors 2020, 20, 2278 16 of 19 high wheel-rail impacts. Such results further explain the ineffectiveness of the frequently performed ballastSensors 2020 tamping, 20, x FOR since PEER ballast REVIEW settlement is not the main cause of the extremely high wheel-rail impacts.16 of 18

Figure 27. Wheel-rail contact forces and lateral wheel displacements.displacements. (a)) FrontFront wheel.wheel. (b) Rear wheel.

6.2. Influence Influence of Reduced Lateral Support It is shown in Figure 16 that the defects of the fastening system can increase the instability of the wheel-railwheel-rail impactimpact inin thethe crossing. crossing. Combined Combined with with the the maintenance maintenance action action and and the the simulation simulation results results in Sectionin Section5, it 5, can it can be inferred be inferred that thisthat ethisffect effect was causedwas caused by the by reduced the reduced lateral lateral track resistance.track resistance. To verify To thisverify inference this inference in the degraded in the degraded model (Sectionmodel (Section 5.2), the 5.2), input the lateral input sti lateralffness stiffness of the clips of the in the clips crossing in the panelcrossing was panel reduced was from reduced 280 MNfrom/m 280 (Table MN/m2) to 2.8(Table N / m,2) andto 2.8 the N/m, corresponded and the corresponded damping was damping reduced from 580 kN s/m to 5.8 N s/m. Based on these inputs, the wheel-rail contact forces and the lateral wheel was reduced· from 580 kN·s/m· to 5.8 N·s/m. Based on these inputs, the wheel-rail contact forces and displacementsthe lateral wheel were displacements calculated, aswere presented calculated, in Figure as presented 28. in Figure 28.

Figure 28. Wheel-rail contact forces and lateral wheel displacements. ( a) Front wheel. ( (bb)) Rear Rear wheel. wheel. Note: Ballast settlement was not taken into account.

It can be be seen seen from from Figure Figure 28 28 that that with with the the reduced reduced lateral lateral stiffness stiffness and and damping damping of the of theclips, clips, the theimpacts impacts of both of both the front the front wheel wheel and the and rear the wheel rear wheel were slightly were slightly increased increased (compared (compared with the with results the resultsin Figure in Figure23). Moreover, 23). Moreover, the hunting the hunting motion motion of wheels of wheels in the incrossing the crossing panel panelwas more was moreintense. intense. As a Asresult, a result, the lateral the lateral deviation deviation of the of therear rear wheel wheel increased increased from from 0.3 0.3 mm mm to to 0.8 0.8 mm. mm. It It can can be be imagined that with thethe impacts of thethe passing trains, the tracktrack alignment will continuously be changing due to the reducedreduced structuralstructural integrity.integrity. The changed tracktrack alignment will, in return, act on thethe wheel-railwheel-rail interactioninteraction andand eventuallyeventually leadlead toto moremore unstableunstable wheelwheel impactsimpacts inin thethe crossing (Figure(Figure 1717).). From this point ofof view, view, renovating renovating the the defected defected fastening fastenin systemg system is necessary is necessary for a monitored for a monitored crossing. crossing. Enough trackEnough lateral track resistance lateral resistance can help can to maintain help to maintain better crossing better performance.crossing performance.

7. Conclusions In this study, the root cause of the fast degradation of a 1:9 crossing in the Dutch railway system was investigated. The effectiveness of some typical track maintenance actions was also assessed and verified. Based on the measurement and simulation results, the following conclusions can be drawn: • The fast crossing degradation was directly caused by the extremely high wheel-rail impacts, and the root cause for such high impacts was the hunting of the passing trains that were

Sensors 2020, 20, 2278 17 of 19

7. Conclusions In this study, the root cause of the fast degradation of a 1:9 crossing in the Dutch railway system was investigated. The effectiveness of some typical track maintenance actions was also assessed and verified. Based on the measurement and simulation results, the following conclusions can be drawn: The fast crossing degradation was directly caused by the extremely high wheel-rail impacts, • and the root cause for such high impacts was the hunting of the passing trains that were activated by the track lateral misalignment in front of the crossing. When the lateral deviation of the passing wheel exceeds a certain extent (e.g., 2.3 mm), the wheel-rail contact situation will change and the wheel impacts will be dramatically increased. To improve the current situation, such track misalignment needs to be eliminated; Ballast settlement is likely to be the accumulated effect of the high wheel-rail impacts. The influence • on the crossing performance is somewhat limited. Ballast tamping, especially with only the squeezing machine, cannot improve the dynamic performance of the crossing. In the case of not knowing the sources of damage, it is better to take no action, rather than implement ballast tamping; Fastening system renovation helped improved the crossing performance by providing better • lateral support in the track but was not targeted to the fundamental problem. Therefore, such damage repair action is useful, but not enough for an improvement in the crossing performance. This study further verified the effectiveness of the previously proposed condition indicators in the investigation of the damage sources of the crossing. Since the root causes for the fast degradation were the deviated track in front of the crossing, this means that the degradation detection is not only restricted to the crossing itself but can also take the adjacent structures into account. The activated train hunting reasonably explained the instability of wheel-rail interaction in the crossing, which pointed out a possible direction to maintain the problematic crossings in the Dutch railway network. As part of the Structural Health Monitoring System for railway crossings developed in TU Delft, the findings in this study will help improve the current maintenance philosophy from “failure reactive” to “failure proactive”, and eventually lead to sustainable railway crossings.

Author Contributions: This article is written by X.L. and supervised by V.L.M. All authors have read and agreed to the published version of the manuscript. Funding: The field measurements in this research were funded by ProRail and performed within the framework of the joint ProRail and TU Delft project “Long-Term Monitoring of Railway Turnouts”. Acknowledgments: The authors would like to thank the support from I.Y. Shevtsov from ProRail and technical support from G.H. Lambert and other colleagues from ID2. Furthermore, the fruitful cooperation with Strukton Rail is highly appreciated. Conflicts of Interest: The authors declare no conflicts of interest.

References

1. ProRail, Ontwerpvoorschrift (Baan en Bovenbouw) Deel 6.1: Wissels en Kruisingen Alignement, Tussenafstanden, Overgangsvrije Zones en Bedieningsapparatuur, V006; ProRail: Utrecht, The Netherlands, 2015; p. 12. 2. Shevtsov,I. Rolling Contact Fatigue Problems at Railway Turnouts. Available online: http://www.sim-flanders. be/sites/default/files/events/Meeting_Materials_Nov2013/mm_13112013_ivan_shevtsov_prorail.pdf (accessed on 13 November 2013). 3. Cornish, A.T. Life-Time Monitoring of in Service Switches and Crossings through Field Experimentation. Ph.D. Thesis, Imperial College London, London, UK, May 2014. 4. Molina, L.F.; Resendiz, E.; Edwards, J.R.; Hart, J.M.; Barkan, C.P.L.; Ahuja, N. Condition Monitoring of Railway Turnouts and Other Track Components Using Machine Vision. In Proceedings of the Transportation Research Board 90th Annual Meeting, Washington, DC, USA, 23–27 January 2011. 5. Chen, C.; Xu, T.; Wang, G.; Li, B. Railway turnout system RUL prediction based on feature fusion and genetic programming. Measurement 2020, 151, 107162. [CrossRef] Sensors 2020, 20, 2278 18 of 19

6. Wang, P.; Wang, L.; Chen, R.; Xu, J.; Gao, M. Overview and outlook on railway track stiffness measurement. J. Mod. Transp. 2016, 24, 89–102. [CrossRef] 7. Yeo, G.J. Monitoring Railway Track Condition Using Inertial Sensors on an in-Service Vehicle. Ph.D. Thesis, University of Birmingham, Birmingham, UK, June 2017. 8. De Rosa, A.; Alfi, S.; Bruni, S. Estimation of lateral and cross alignment in a railway track based on vehicle dynamics measurements. Mech. Syst. Signal Process. 2019, 116, 606–623. [CrossRef] 9. Balouchi, F.; Bevan, A.; Formston, R. Detecting Railway Under-Track Voids Using Multi-Train in-Service Vehicle Accelerometer. In Proceedings of the 7th IET Conference on Railway Condition Monitoring, Birmingham, UK, 27–28 September 2016. 10. Tsunashima, H. Condition Monitoring of Railway Tracks from Car-Body Vibration Using a Machine Learning Technique. Appl. Sci. 2019, 9, 2734. [CrossRef] 11. Kassa, E.; Nielsen, J.C.O. Stochastic analysis of dynamic interaction between train and railway turnout. Veh. Syst. Dyn. 2008, 46, 429–449. [CrossRef] 12. Alfi, S.; Bruni, S. Mathematical modelling of train-turnout interaction. Veh. Syst. Dyn. 2009, 47, 551–574. [CrossRef] 13. Ren, Z.; Sun, S.; Xie, G. A method to determine the two-point contact zone and transfer of wheel rail forces in a turnout. Veh. Syst. Dyn. 2010, 48, 1115–1133. [CrossRef] 14. Pletz, M.; Daves, W.; Ossberger, H. A wheel set/crossing model regarding impact, sliding and deformation explicit finite element approach. Wear 2012, 294, 446–456. [CrossRef] 15. Anyakwo, A.; Pislaru, C.; Ball, A. A New Method for Modelling and Simulation of the Dynamic Behaviour of the Wheel-rail contact. Int. J. Autom. Comput. 2012, 9, 237–247. [CrossRef] 16. Pålsson, B.A. Optimisation of Railway Switches and Crossings. Ph.D. Thesis, Chalmers University of Technology, Göteborg, Sweden, February 2014. 17. Wei, Z.; Chen, S.; Li, Z.; Dollevoet, R. Wheel-Rail Impact at Crossings-Relating Dynamic Frictional Contact to Degradation. J. Comput. Nonlinear Dyn. 2017, 12, 1–11. [CrossRef] 18. Xin, L. Long-Term Behaviour of Railway Crossings Wheel-Rail Interaction and Rail Fatigue Life Prediction. Ph.D. Thesis, Delft University of Technology, Delft, The Netherlands, June 2017. 19. Chiou, S.; Yen, J. Modeling of railway turnout geometry in the frog area with the vehicle wheel trajectory. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2018, 232, 1598–1614. [CrossRef] 20. Skrypnyk, R.; Nielsen, J.C.O.; Ekh, M.; Pålsson, B.A. Metamodelling of wheel-rail normal contact in railway crossings with elasto-plastic material behavior. Eng. Comput. 2019, 35, 139–155. [CrossRef] 21. Torstensson, P.T.; Squicciarini, G.; Krüger, M.; Pålsson, B.A.; Nielsen, J.C.O.; Thompson, D.J. Wheel-rail impact loads and noise generated at railway crossings influence of vehicle speed and crossing dip angle. J. Sound Vib. 2019, 456, 119–136. [CrossRef] 22. Wiest, M.; Daves, W.; Fischer, F.D.; Ossberger, H. Deformation and damage of a crossing nose due to wheel passages. Wear 2008, 265, 1431–1438. [CrossRef] 23. Johansson, A.; Pålsson, B.; Ekh, M.; Nielsen, J.; Ander, M.; Brouzoulis, J.; Kassa, E. Simulation of wheel-rail contact and damage in switches & crossings. Wear 2010, 271, 472–481. 24. Nielsen, J.C.O.; Li, X. Railway track geometry degradation due to differential settlement of ballast/subgrade Numerical prediction by an iterative procedure. J. Sound Vib. 2018, 412, 441–456. [CrossRef] 25. Skrypnyk, R.; Ekh, M.; Nielsen, J.C.O.; Pålsson, B.A. Prediction of plastic deformation and wear in railway crossings—Comparing the performance of two rail steel grades. Wear 2019, 428–429, 302–314. [CrossRef] 26. Wan, C. Optimisation of Vehicle-Track Interaction at Railway Crossings. Ph.D. Thesis, Delft University of Technology, Delft, The Netherlands, September 2016. 27. Liu, X.; Markine, V.L.; Wang, H.; Shevtsov, I.Y. Experimental tools for railway crossing condition monitoring. Measurement 2018, 129, 424–435. [CrossRef] 28. Liu, X.; Markine, V.L. Correlation Analysis and Application in the Railway Crossing Condition Monitoring. Sensors 2019, 19, 4175. [CrossRef] 29. Liu, X.; Markine, V.L. Validation and Verification of the MBS Models for the Dynamic Performance Study of Railway Crossings. Eng. Struct. 2019, submitted. 30. Wang, H.; Markine, V.L.; Liu, X. Experimental analysis of railway track settlement in transition zones. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2018, 232, 1774–1789. [CrossRef][PubMed] Sensors 2020, 20, 2278 19 of 19

31. Xu, L.; Yu, Z.; Shi, C. A matrix coupled model for vehicle-slab track-subgrade interactions at 3-D space. Soil Dyn. Earthq. Eng. 2020, 128, 105894. [CrossRef] 32. Hiensch, M.; Nielsen, J.C.O.; Verheijen, E. Rail corrugation in The Netherlands measurements and simulations. Wear 2002, 253, 140–149. [CrossRef]

© 2020 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/).