Railway Vehicle Wheel Flat Detection with Multiple Records Using Spectral Kurtosis Analysis

applied sciences

Article Railway Vehicle Wheel Flat Detection with Multiple Records Using Spectral Kurtosis Analysis

Araliya Mosleh * , Pedro Aires Montenegro , Pedro Alves Costa and Rui Calçada

CONSTRUCT-LESE, Faculty of Engineering (FEUP), University of Porto, 4200-465 Porto, Portugal; [email protected] (P.A.M.); [email protected] (P.A.C.); [email protected] (R.C.) * Correspondence: [email protected]

Abstract: The gradual deterioration of train wheels can increase the risk of failure and lead to a higher rate of track deterioration, resulting in less reliable railway systems with higher maintenance costs. Early detection of potential wheel damages allows railway infrastructure managers to control railway operators, leading to lower infrastructure maintenance costs. This study focuses on identifying the type of sensors that can be adopted in a wayside monitoring system for wheel flat detection, as well as their optimal position. The study relies on a 3D numerical simulation of the train-track dynamic response to the presence of wheel flats. The shear and acceleration measurement points were defined in order to examine the sensitivity of the layout schemes not only to the type of sensors (strain gauge and accelerometer) but also to the position where they are installed. By considering the shear and accelerations evaluated in 19 positions of the track as inputs, the wheel flat was identified by the envelope spectrum approach using spectral kurtosis analysis. The influence of the type of sensors and their location on the accuracy of the wheel flat detection system is analyzed. Two types of trains were considered, namely the Alfa Pendular passenger vehicle and a freight wagon.

Citation: Mosleh, A.; Montenegro, Keywords: railway vehicles; wheel flat detection; spectral kurtosis analysis; wayside condition P.A.; Costa, P.A.; Calçada, R. Railway monitoring systems; multisensor array Vehicle Wheel Flat Detection with Multiple Records Using Spectral Kurtosis Analysis. Appl. Sci. 2021, 11, 4002. https://doi.org/10.3390/ 1. Introduction app11094002 The presence of a geometric defect in the interface between the rail and the wheel is Academic Editor: Doo-Yeol Yoo one of the sources of dynamic train interaction. Wheel profiles have a significant effect on the safety and dynamic performance of the vehicle, for instance, in terms of the dynamic Received: 19 March 2021 stability of the vehicle, the magnitude of the wheel-rail contact force, and riding comfort. Accepted: 26 April 2021 Wheel tread imperfections are usually divided into two main categories: (i) defect along Published: 28 April 2021 part of the wheel circumference and (ii) defect around the entire wheel. These are both regarded as types of wheel out-of-roundness (OOR) phenomena. The first category includes Publisher’s Note: MDPI stays neutral wheel flat, spalling, shelling, etc., and is mainly caused by braking damage and rolling with regard to jurisdictional claims in contact fatigue cracking. The second category involves wheel corrugation and polygonal published maps and institutional affil- wheel, which are periodic irregularities around the wheel that can be caused by unbalanced iations. loads [1]. The non-roundness of railway wheels creates an unfavorable effect on the components of track and vehicle [2]. In recent decades, several researchers [3–6] have focused on the development mechanism of OOR. However, the main focus of this research is the development of techniques that accurately identify the presence of defective wheels, Copyright: © 2021 by the authors. namely with flats [7–12]. Licensee MDPI, Basel, Switzerland. When the wheelsets are locked and slides along the rails as a result of poorly adjusted This article is an open access article or defective brakes, the wheel flats are created. Therefore, the surface of the wheels becomes distributed under the terms and flat rather than round due to the friction between the wheels and the rails [13]. Wheel flats conditions of the Creative Commons induce high dynamic impact loads on the railway infrastructure, which causes significant Attribution (CC BY) license (https:// damage to the vehicles and the track, such as broken axles, hot axle boxes, damaged rolling creativecommons.org/licenses/by/ bearings, and cracks in the wheels, rails, and sleepers. Moreover, this type of wheel defect 4.0/).

Appl. Sci. 2021, 11, 4002. https://doi.org/10.3390/app11094002 https://www.mdpi.com/journal/applsci Appl. Sci. 2021, 11, 4002 2 of 25

generates high levels of noise and excessive vibration. These large magnitude vibrations are heavily transmitted to the rolling stock, inducing forces way above the permitted values, causing damage to the suspension system, frame, and car body of the rolling stock [8,14]. In addition, the irregularity of the wheel and track also results in the deterioration of the interaction performance of the train and overhead contact line [15]. Therefore, to tackle this issue, railway administrations take the necessary measures to make sure all required precautions and renovations are in place. In terms of precautions, most passenger trains are currently equipped with advanced anti-slip systems that slightly reduce wheel-rail sliding [16]. However, as operating speeds and axle loads increase, wheel flats cannot be completely avoided. Moreover, since freight trains do not have a non-slip system, the condition of the wheels is usually worse, having a significant impact on the long-term service of trains and infrastructure. Several researchers have proposed different in-service methods for measuring wheel defects in the last few decades, namely onboard and wayside measurements. Most onboard techniques are based on vibration, acoustic, image detection, and ultrasonic technolo- gies [17–20]. However, for a comprehensive diagnosis and effective management of the wheels, sensors must be installed on all wheels. The installation of sensors on wheels is seldom used due to cost and maintenance issues, while onboard detection methods are commonly used to monitor the track condition and not for monitoring wheel condition. In contrast, wayside measurement systems are currently a preferable solution to detect wheel flats, as the condition of all wheels evaluates during the passage of trains [21–28]. Liu et al. [29] proposed a quantitative decision approach based on wheel impact data that illustrates the actual condition of the wheels. According to the proposed method, the critical wheels should be removed while wheels that are not in critical condition remain in operation. Liu and Ni [30] developed a fiber Bragg grating (FGB) method to monitor trackside wheel conditions and detect wheel tread defects. The track-side system with more than 20 fiber Bragg grating sensors installed on the rail allows measuring the bending moments of the rail under a defective wheel. An algorithm is developed to identify potential wheel tread defects using the online-monitored rail responses. The results show that when using the proposed method to process the monitoring data, all the defects were identified, and the results were in accordance with those of the static inspection of the wheelsets. Gao et al. [31] presented a parallelogram mechanism based on the wheel flat detection system to perform the dynamic and quantitative measurement of wheel flats. In their study, the wheel flat could be detected by measuring the change in the vertical displacement of the measuring ruler. The experiments performed in the laboratory showed the effectiveness of the proposed system. The main problem in fault diagnosis is differentiating these fault-related frequencies from the signal spectrum. Since the spectral characteristics of the amplitude modulation component (AM) and the frequency modulation component (FM) are relatively simple and closely related to the fault-related frequencies, the wheel flat information can be disclosed in the spectrum of the AM component and the instant frequency spectrum of the FM component. Therefore, to separate these low-frequency modulation components from the high-frequency carrier signal, a demodulation method can be used [32,33]. The amplitude envelope and instantaneous frequency of a signal can be adequately estimated by the Hilbert transform [34]. However, the selection of a center frequency and bandpass filter bandwidth poses a challenge to the practical application of these traditional demodulation methods. In fact, the center frequency estimation has a direct influence on the final demodulation result [35]. In other words, extracting the impulsive signal with amplitude modulation is a key preprocessing step before performing the envelope spectrum. Therefore, the critical step is to extract the signal with the highest kurtosis value for each band frequency. One of the innovations of this research is to obtain the center frequency and the bandwidth frequency to apply this information in the envelope spectrum approach to detect a defective wheel. Appl. Sci. 2021, 11, 4002 3 of 25

In previous studies, for the successful detection of wheel defects, sensors have been installed along an equivalent wheel perimeter length [21,36]. However, the study performed by Mosleh et al. [37] shows that this amount of sensors is not necessary to identify the defective wheel. Installing sensors along an equivalent wheel perimeter length is useful for identifying specific moments when wheel flat impacts occur. Moreover, most previous studies have relied on engineering experiments regarding flat detection rather than sufficient theoretical analysis. With this in mind, this paper established a 3D numerical dynamic model of a vehicle-track coupling system and analyzed the sensitivity and reliability of different sensors and setups to detect wheel flats. A more useful scheme based on multisensor arrays was proposed to capture the abnormal responses caused by wheel flats. A wheel flat recognition methodology was developed using simulated sensor output signals. Another innovation of this research compared to the previous studies from the authors [21,36,37] is reducing the number of installed sensors and optimizing their position.

2. Train-Track Dynamic Interaction Modeling 2.1. Numerical Modeling for Alfa Pendular Vehicle and Freight Wagon In order to enhance the parametric study carried out in the present paper, two different train models were developed, namely a passenger (Alfa Pendular) train and a container freight (Laagrss wagon). The numerical models of both trains were performed in the Finite Element Method (FEM) package ANSYS® [38] using spring-damper elements COMBIN14 to simulate the suspensions that connect the various components of the trains, including car body, bogies, and wheelsets. The mass and the rotary inertia of the mentioned components are taken into account through mass point elements MASS21. The dynamic models adopted for each train are depicted in Figure1. As presented in this ﬁgure, k, c, m, and I denote stiffness, damping, concentrated mass, and rotary inertia. Moreover, the subscripts cb, b, and w refer to the car body, bogies, and wheelsets. rw is the wheel radius, and gauge is indicated by s. Finally, the characters a, b, and h denote the longitudinal, transversal, and vertical distances. The geometrical and mechanical properties of both trains are presented in Tables1 and2. Note that in all analyses, only one car was taken into account to simplify the numerical computations.

Table 1. Parameters for the Alfa Pendular vehicle (adapted from Neto et al. (2020) [39]).

Parameter Value

Car body mass, mcb 35,640 kg 2 Car body roll moment of inertia, Icb,x 55,120 kg·m 2 Car body pitch moment of inertia, Icb,y 1,475,000 kg·m 2 Car body yaw moment of inertia, Icb,z 1,477,000 kg·m Bogie mass, mb 2829 kg 2 Bogie roll moment of inertia, Ib,x 2700 kg·m 2 Bogie pitch moment of inertia, Ib,y 1931.49 kg·m 2 Bogie yaw moment of inertia, Ib,z 3878.76 kg·m Wheelset mass, mw 1711 kg 2 Wheelset roll moment of inertia, Iw,x 733.4303 kg·m 2 Wheelset yaw moment of inertia, Iw,z 733.4303 kg·m Stiffness of the primary longitudinal suspension, k1,x 44,981,000 N/m Stiffness of the primary transversal suspension, k1,y 30,948,200 N/m Stiffness of the primary vertical suspension, k1,z 1,652,820 N/m Damping of the primary vertical suspension, c1,z 16,739 N·s/m Stiffness of the secondary longitudinal suspension, k2,x 4,905,000 N/m Stiffness of the secondary transversal suspension, k2,y 2,500,000 N/m Appl. Sci. 2021, 11, 4002 4 of 25

Appl. Sci. 2021, 11, x FOR PEER REVIEW 3 of 25 Table 1. Cont.

Parameter Value

Stiffness of the secondaryfrequency vertical and the suspension, bandwidthk2,z frequency to apply this information734,832 in N/m the envelope spec- Damping of the secondarytrum approach longitudinal to detect suspension, a defectivec2,x wheel. 400,000 N·s/m Damping of the secondaryIn previous transversal studies, suspension, for thec2,y successful detection of wheel17,500 defect Ns,·s/m sensors have been Damping of theinstalled secondary along vertical an suspension, equivalentc2 ,z wheel perimeter length [21,36]35,000. However, N·s/m the study per- The static loadformed transmitted by Mosleh by each et al. wheel [37] shows that this amount of sensors is 64,000 not necessary N to identify Longitudinal distance between bogies, a 19 m the defective wheel. Installing1 sensors along an equivalent wheel perimeter length is use- Longitudinal distance between wheelsets, a2 2.7 m ful for identifying specific moments when wheel flat impacts occur. Transversal distance between vertical secondary suspensions, b1 2.144 m Transversal distance between longitudinalMoreover, most secondary previous suspensions, studiesb 2have relied on engineering2.846 experiments m regarding Transversal distanceflat between detection primary rather suspensions, than sufficientb3 theoretical analysis. With this2.144 in mind, m this paper es- Vertical distance between cartablished body center a 3Dand numerical secondary dynamic suspension, modelh1 of a vehicle-track coupling0.936 system m and analyzed Vertical distance betweenthe bogie sensitivity center and and secondary reliability suspension, of differenth2 sensors and setups to detect0.142 mwheel flats. A more Vertical distance between bogie center and wheelset center, h 0.065 m useful scheme based on multisensor3 arrays was proposed to capture the abnormal re- Nominal rolling radius, rw 0.43 m sponsesGauge, causeds by wheel flats. A wheel flat recognition methodology1.67 m was developed using simulated sensor output signals. Another innovation of this research compared to the previous studies from the authors [21,36,37] is reducing the number of installed sensors Table 2.andParameters optimizing for thetheir freight position wagon. (adapted from Neto et al. (2019) [40]). Parameter Value 2. Train-Track Dynamic Interaction Modeling Car body mass, mcb 41,100 kg 2.1. Numerical Modeling for Alfa Pendular Vehicle and Freight Wagon 2 Car body roll moment of inertia, Icb,x 48,997.023 kg·m 2 Car body pitchIn moment order ofto inertia,enhanceIcb,y the parametric study carried out in 673,322.463the present kg paper,·m two differ- 2 Car bodyent yaw train moment models of inertia, wereI cb,zdeveloped, namely a passenger (Alfa665,107.6 Pendular) kg· mtrain and a con- Wheelsettainer freight mass, (Laagrssmw wagon). The numerical models of both 1246.52trains were kg performed in · 2 Wheelsetthe roll Finite moment Element of inertia, MethodIw,x (FEM) package ANSYS® [38] using311.839 spring kg-mdamper elements Wheelset yaw moment of inertia, I 311.839 kg·m2 COMBIN14 to simulatew,z the suspensions that connect the various components of the trains, Stiffness of the longitudinal suspension, k1,x 44,981,000 N/m including car body, bogies, and wheelsets. The mass and the rotary inertia of the men- Stiffness of the transversal suspension, k1,y 30,948,200 N/m Stiffness oftioned the vertical components suspension, are ktaken1,z into account through mass point1,860,000 elements N/m MASS21. The dy- Damping ofnamic the vertical models suspension, adopted forc1,z each train are depicted in Figure 1. 16,739As presented N·s/m in this figure, The static loadk, c, transmitted m, and I denote by each stiffness, wheel damping, concentrated mass, and107,000 rotary Ninertia. Moreover, Longitudinal distance between wheelsets, a 6 m the subscripts cb, b, and w refer1 to the car body, bogies, and wheelsets. rw is the wheel Transversal distance between vertical suspensions, b 2.17 m radius, and gauge is indicated by1 s. Finally, the characters a, b, and h denote the longitu- Vertical distance between car body center and suspension, h1 1.867 m dinal, transversal, and vertical distances. The geometrical and mechanical properties of Nominal rolling radius, rw 0.43 m both Gauge,trains sare presented in Tables 1 and 2. Note that in all analyses,1.67 m only one car was taken into account to simplify the numerical computations.

mcb ,Icb

mcb ,Icb a1 /2 a1 /2 k ,c k ,c 2, z 2, z 2, z 2, z a1 /2 a1 /2 k ,c mb , Ib mb ,Ib 1, z 1, z k1, z ,c1, z k1, z ,c1, z k1, z ,c1, z

mw , Iw

a 2

(a) (b)

Figure 1. Cont. Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 of 25 Appl. Sci. 2021, 11, x FOR PEER REVIEW 4 of 25 Appl. Sci. 2021, 11, 4002 5 of 25

Gauge, s 1.67 m

m , I Table 2. Parametersc bforcb the freight wagon (adapted from Neto et al. (2019) [40]).

b1 Parameter mcb ,Icb Value h1 b2 Car body mass, mcb b 41,100 kg k ,c 1 k2, x ,c2, x 2, z 2, z 2 k ,Carc body roll moment of inertia, Icb,x 48,997.023 kg·m 2, y 2, y k1, z ,c1, z h1 m , I h2 Car body pitch bmomentb of inertia, Icb,y 673,322.463 kg·m2 k1, x ,c1,x mw , Iw cb,z 2 k1, z Car,c1, z body yaw moment of inertia, I 665,107.6 kg·m h3 k ,c k c 1, y 1, y 1, x , 1, x Wheelset mass, mw R0 1246.52 kg mw , Iw w,x 2 k1, y ,Wheelsetc1, y roll moment of inertia, I s 311.839 kg·m b3 R Wheelset yaw moment of inertia,0 Iw,z 311.839 kg·m2 s Stiffness of the longitudinal suspension, k1,x 44,981,000 N/m Stiffness of the transversal suspension, k1,y 30,948,200 N/m (c) (d) Stiffness of the vertical suspension, k1,z 1,860,000 N/m FigureFigure 1. DynamicDynamicDamping models models of of theof the therailway railway vertical vehicles: vehicles: suspension, lateral lateral views views c 1of,z of the the ( (aa)) Alfa Alfa Pendular Pendular and and ( (bb)) freight freight wagon, wagon,16,739 and and N front front·s/m views of the (c) Alfa Pendular and (d) freight wagon. views of the (c) AlfaThe Pendularstatic load and transmitted (d) freight wagon. by each wheel 107,000 N LongitudinalTable 1. Parameters distance for the between Alfa Pendular wheelset vehicles, a(adapted1 from Neto et al. (2020) [39]). 6 m 2.2. Railway Track Transversal distance between vertical suspensions, b1 2.17 m The numericalParameter model of the track was developed in ANSYS® [38Value] and is schema- Vertical distance between car body center and suspension, h1 1.867 m tized inCar Figure body2 .mass, It consists mcb of a three-layer model composed by the rail35,640 modeled kg with Nominal rolling radius, rw 0.43 m Carbeam body elements roll moment BEAM181, of inertia, and Icb,x by the ballast and sleepers modeled with55,120 mass kg·m point2 ele- Gauge, s 1.67 m Carments body MASS21.pitch moment These of three inertia, layers Icb,y are connected through linear spring-damper1,475,000 kg·m elements2 CarCOMBIN14, body yaw moment which simulate of inertia, the Icb,z foundation, the ballast and the pad/fastener,1,477,000 kg·m indicated2 f b p 2.2. Railwayin Figure BogieT2rackby mass, the subscripts mb , , and , respectively. The mechanical2829 properties kg of the different elements are presented in Table3. BogieThe numerical roll moment model of inertia, of the Ib,x track was developed in ANSYS® [38]2700 and kg·m is schematized2 In real conditions, the rails are not perfect, which also affects the wheel-rail forces. Bogie pitch moment of inertia, Ib,y 1931.49 kg·m2 in FigureTherefore, 2. It aconsists sample of a the three track-layer irregularity model wascomposed measured by onthe a rail Northern modeled Line with of the beam Bogie yaw moment of inertia, Ib,z 3878.76 kg·m2 elementsPortuguese BEAM181, Railway and Network by the as ballast the initial and excitation sleepers of modeled the wheel-rail with contact. mass Morepoint details elements Wheelset mass, mw 1711 kg MASS21.about the These unevenness three layers profile are measurement connected are through provided linear by Mosleh spring et-damper al. [25]. As elements an 2 COMBIN14,Wheelsetexample, roll Figurewhich moment3 simulatedepicts of inertia, a 220the mIfoundation,w,x stretch (the the total ballast length and of the the simulation) pad/fastener,733.4303 of kg·m the indicated vertical in 2 FigureWheelsetand 2 lateralby yaw the irregularitymomentsubscript ofs inertia, profilesf, b, and Iw,z corresponding p , respectively. to The the rightmechanical rail. The 733.4303properties sample kg·m of of the the track differ- Stiffnessent of elementsirregularity the primary are profile longitudinalpresented shown in insuspension, Table Figure 3.3 is k used1,x for all analyses in this research44,981,000 study. N/m Stiffness of the primary transversal suspension, k1,y 30,948,200 N/m

Stiffness of the primaryRail: verticalAr, Ir, ρ suspension,r, Er k1,z 1,652,820 N/m

DampingPad/Fastner of the primary behavior vertical: kp, cp suspension, c1,z 16,739 N·s/m Stiffness of the secondary longitudinal suspension, k2,x 4,905,000 N/m Sleeper: ρs Stiffness of the secondary transversal suspension, k2,y 2,500,000 N/m Stiffness of theBallast secondary behavior: vertical kb, cb suspension, k2,z 734,832 N/m

Damping of the secondaryBallast longitudinal mass: ρb suspension, c2,x 400,000 N·s/m Damping of the secondary transversal suspension, c2,y 17,500 N·s/m Foundation behavior: kf, cf Damping of the secondary vertical suspension, c2,z 35,000 N·s/m The static load transmitted by each wheel 64,000 N

Longitudinal distance between bogies, a1 19 m LongitudinalFigureFigure 2. Railway 2. distanceRailway track between track modeling modeling. wheelset. s, a2 2.7 m Transversal distance between vertical secondary suspensions, b1 2.144 m Table 3. Mechanical properties of the track. Transversal distance between longitudinal secondary suspensions, b2 2.846 m Transversal distance between primaryParameter suspensions, b3 Value2.144 m Vertical distance between car body center and secondary suspension, h1 0.936 m Ar (m2) 7.67 × 10−4 [41] Vertical distance between bogie center and secondary suspension, h2 0.142 m ρr (kg/m3) 7850 [41] Vertical distance between bogie center and wheelset center, h3 0.065 m Rail Ir (m4) 30.38 × 10−6 [41] Nominal rolling radius, rw 0.43 m νr 0.28 [41] Er (N/m2) 210 × 109 [41] Kp (N/m) 20 × 106 [42] Rail pad, longitudinal Cp (N·s/m) 50 × 103 [42] Appl. Sci. 2021, 11, 4002 6 of 25

Table 3. Mechanical properties of the track. Appl. Sci. 2021, 11, x FOR PEER REVIEW 6 of 25

Parameter Value 2 −4 Ar (m ) 7.67 × 10 [41] 3 ρr (kg/m ) 7850Kp (N/m) [41] 20 × 106 [42] 4 Rail pad, lateral × −6 Rail Ir (m ) 30.38 10 Cp (N·s/m) [41] 50 × 103 [42] νr 0.28 [41] 6 2 9 Kp (N/m) 500 × 10 [43] Er (N/m )Rail pad, vertical 210 × 10 [41] Cp (N·s/m) 200 × 103 [43] 6 Kp (N/m) 20 × 10 3 [42] Rail pad, longitudinal 3 ρs (kg/m ) 2590 Cp (N·s/m) 50 × 10 [42] Sleeper νs 0.2 K (N/m) 20 × 106 [42] p Es (N/m2) 40.9 × 109 Rail pad, lateral 3 Cp (N·s/m) 50 × 10 [42] Kb,x (N/m) 9000 × 103 [44] Ballast, longitudinal 6 K (N/m) × [43] 3 p 500 10 Cb,x (N·s/m) 15 × 10 [45] Rail pad, vertical 3 Cp (N·s/m) 200 × 10 [43] Kb,y (N/m) 2250 × 103 [46] 3 Ballast, lateral ρs (kg/m ) 2590 Cp,y (N·s/m) 15 × 103 [45] Sleeper νs 0.2 6 2 9 Kp,z (N/m) 30 × 10 [46] Es (N/m )Ballast, vertical 40.9 × 10 Cp,z (N·s/m) 15 × 103 [45] 3 Kb,x (N/m) 9000 × 10 [44] 6 Ballast, longitudinal 3 Kf,x (N/m) 20 × 10 [47] Cb,x (N·s/m) Foundation, longitudinal15 × 10 [45] Cf,x (N·s/m) 5.01 × 102 [47] K (N/m) 2250 × 103 [46] b,y Kf,y (N/m) 20 × 106 [47] Ballast, lateral 3 Cb,y (N·s/m) Foundation, lateral15 × 10 [45] Cf,y (N·s/m) 5.01 × 102 [47] 6 × 6 Kb,z (N/m) 30 10 Kf,z (N/m) [46] 20 × 10 [47] Ballast, vertical 3 Cb,z (N·s/m) Foundation, vertical15 × 10 [45] Cf,z (N·s/m) 5.01 × 102 [47] 6 Kf,x (N/m) 20 × 10 [47] Foundation, longitudinal 2 Cf,x (N·s/m)In real conditions, the5.01 rails× 10 are not perfect, which also[47 affects] the wheel-rail forces. 6 Kf,y (N/m)Therefore, a sample of the20 track× 10 irregularity was measured[ 47on] a Northern Line of the Por- Foundation, lateral 2 Cf,y (N·tugueses/m) Railway Network5.01 as× the10 initial excitation of the wheel[47] -rail contact. More details about the unevenness profile measurement6 are provided by Mosleh et al. [25]. As an ex- Kf,z (N/m) 20 × 10 [47] Foundation, vertical ample, Figure 3 depicts a 220 m stretch2 (the total length of the simulation) of the vertical Cf,z (N·s/m) 5.01 × 10 [47] and lateral irregularity profiles corresponding to the right rail. The sample of the track irregularity profile shown in Figure 3 is used for all analyses in this research study.

(a) (b)

Figure 3. Track irregularityFigure 3. Track sample: irregularity (a) vertical sample: and ( b(a)) lateral vertical direction. and (b) lateral direction.

2.3. Train-Track Interaction2.3. Train-Track Interaction The train-track dynamicThe train interaction-track dynamic coupling interaction system iscoupling solved usingsystem the is numerical solved using tool the numerical “VSI-Vehicle-Structuretool “VSI Interaction-Vehicle-Structure Dynamic Interaction Analysis.” Dynamic This tool Analysis.” is based on This the Lagrangetool is based on the La- multipliers method,grange in multipliers which the constraintmethod, in equations which the thatconstraint relate equations the train displacements that relate the train displace- with the nodal displacementsments with the of nodal the track, displacements including theof the track track, irregularities, including the are addedtrack irregularities, to are the equilibriumadded equations, to the forming equilibrium a single equations, system that forming deﬁnes a single the coupling system betweenthat defines the the coupling between the two sub-structures. Note that when including the track irregularities in the constraint equations, it is not necessary to include them explicitly in the FEM model, Appl. Sci. 2021, 11, 4002 7 of 25 Appl. Sci. 2021, 11, x FOR PEER REVIEW 7 of 25

two sub-structures. Note that when including the track irregularities in the constraint equa- which considerably simplifies the analysis. Moreover, as will be mentioned later, it is also tions, it is not necessary to include them explicitly in the FEM model, which considerably possible to include the wheel flats as a periodic rail irregularity. To solve the train struc- simplifies the analysis. Moreover, as will be mentioned later, it is also possible to include ture detachment, Neves et al. [48] and Montenegro et al. [49] extended the original formu- the wheel flats as a periodic rail irregularity. To solve the train structure detachment, Neveslation etdeveloped al. [48] and by MontenegroNeves et al. et[50] al.. Subsequently, [49] extended theMonten originalegro formulation et al. [51] reformulated developed bythe Neves method et to al. c [onsider50]. Subsequently, the lateral interaction Montenegro by ettaking al. [51 into] reformulated account the thewheel method and rail to considergeometry the and lateral the contact interaction model by between taking into them. account In the the last wheel upgrade, and railthe vertical geometry and and lateral the contactcontact modelforces were between computed them. Inand the made last upgrade,the software the possible vertical andto deal lateral with contact different forces sce- werenario computeds. and made the software possible to deal with different scenarios. TheThe wheel-railwheel-rail contact contact model model adopted adopted in in this this numerical numerical tool tool has hasbeen been validated validated with withnumerical numerical examples examples and experimental and experimental data (see data [48 (see,51] [48) and,51]) is and based is based on a specially on a specially devel- developedoped finite finitecontact contact element element that con thatsiders considers the geometry the geometry of the contact of the interface contact by interface param- byeterizing parameterizing the wheel the and wheel rail profiles and rail with profiles cubic with splines. cubic The splines. element The uses element the Lagrange uses the Lagrangemultipliers multipliers method to method guarantee to guarantee the coupling the couplingbetween the between vehicle the and vehicle the track and thestructure track structurewhen contact when occurs, contact but occurs, can also but lead can to also wheel lead-rail to wheel-raildetachment. detachment. The contact Theformulation contact formulationimplemented implemented in this contact in element this contact can be element defined can by bethree defined main bysteps, three namely: main steps,(i) the namely:geometric (i) analysis, the geometric (ii) the analysis, normal (ii)contact the normal analysis, contact and ( analysis,iii) the tangential and (iii) thecontact tangential analy- contactsis. analysis. TheThe geometric geometric analysis analysis consists consists of evaluatingof evaluating the the location location of the of contact the contact points points between be- thetween wheel the andwheel the and rail the based rail on based the geometryon the geometry of their of profiles their profiles previously previously parameterized parame- throughterized through cubic splines. cubic Thesplines. contact The search contact is thensearch carried is then out carried through out a through couple of a nonlinear couple of equations,nonlinear whoseequations, solution whose guarantees solution theguarantees compatibility the compatibility between the twobetween contacting the two bodies con- (seetacting [51] bodies for details). (see [51] Although for details). the present Although study the does present not focus study on does scenarios not focus with on important scenarios lateralwith important loads, it islateral important loads, to it highlightis important that to the highlight wheel profilethat the is wheel parameterized profile is parame- by two surfacesterized by (tread two andsurfaces flange), (tread which and enables flange), the which algorithm enables to the deal algorithm with double to deal contact with point dou- scenariosble contact where point the scenarios flange alsowhere touches the flange the rail. also touches the rail. WithWith thethe locationlocation ofof thethe contactcontact pointspoints determined,determined, thethe computationcomputation ofof thethe contactcontact forcesforces throughthrough the the normal normal and and tangential tangential contact contact algorithms algorithms can can then then be be performed. performed. The The formerformer is is basedbased onon thethe HertzHertz nonlinearnonlinear theory theory [ [52]52] and and aimsaims toto computecompute thethe contactcontact forcesforces perpendicularperpendicular to to the the contacting contacting bodies, bodies, while while the the latter latter uses usesKalker’s Kalker’s USETABUSETAB [ 53[53]] routine routine toto calculatecalculate thethe creepcreep forcesforces that that arisearise from from the the rolling rolling friction friction contact contact between between the the wheel wheel andand thethe rail.rail. TheThe dynamicdynamic coupling coupling formulation formulation used used in in this this work, work, and and schematized schematized in Figurein Figure4, is 4, ® programmedis programmed inMATLAB in MATLAB[54® ],[54] which, which imports imports the structuralthe structural matrices matrices of the of trackthe track and theand ® trainthe train modeled modeled in the in FEMthe FEM package package ANSYS ANSYS[38®]. [38] A detailed. A detailed description description of the of interactionthe interac- model,tion model, as well asas well its validation,as its validation, can be can found be found in Montenegro in Montenegro et al. [51 et] andal. [51] Montenegro and Montene- [55]. Figuregro [55]4 presents. Figure 4 the presents modeling the ofmodeling train-track of train interaction-track interaction numerical numerical modeling. modeling.

Train dynamic model

Wheel-rail contact model

Track model FigureFigure 4.4.The The modelingmodeling ofof train-tracktrain-trackinteraction. interaction.

2.4.2.4. SystemSystem DescriptionDescription 2.4.1.2.4.1. LayoutLayout SchemeScheme ofof MultisensorMultisensor ArraysArrays TheThe wheelwheel ﬂatflat detection system consists consists of of a a range range of of strain strain gauges gauges (SGs) (SGs) and and accel- ac- celerometerserometers installed installed along along a astretch stretch of of the the track. track. The The shear shear and accelerations areare evaluatedevaluated forfor the the 1919 positionspositions ofof thethe tracktrack illustratedillustrated inin FigureFigure5 ,5 which, which represent represent the the positions positions where where thethe sensorssensors wouldwould bebe installed in in a a real real wayside wayside system. system. In In this this figure, ﬁgure, numbers numbers 1 to 1 to12 12represent represent shear shear measurement measurement or or acceleration acceleration points, points, which which simulate the positionposition ofof thethe Appl. Sci. 2021, 11, x FOR PEER REVIEW 8 of 25 Appl. Sci. 2021, 11, 4002 8 of 25

strain gauges (layout scheme 1) or accelerometers (layout scheme 2) installed on the rail, whilestrain numbers gauges (layout 13 to 19 scheme represent 1) or only accelerometers acceleration (layout measurement scheme 2)points installed (layout on thescheme rail, 3),while which numbers represent 13 to the 19 representposition of only the accelerationaccelerometers measurement installed on points the rail (layout over schemethe sleep- 3), ers.which represent the position of the accelerometers installed on the rail over the sleepers.

Figure 5. Schematic of a vehicle containing the damaged wheel.wheel.

Although strain gauges are widely used in field field tests for the experimental evaluation of wheel wheel-rail-rail forces, they are not suitable for long long-term-term monitoring of wheel flats, flats, as they are vulnerable to electromagnetic interference interference from from environme environmentalntal conditions. conditions. Moreover, Moreover, base drift for strain gauges is unavoidable during long long-term-term service, making it impossible toto maintain maintain consistency for a long period. For For these these reasons reasons,, and and unlike unlike the the previous previous work carried out by Mosleh et al. [37] [37],, in which only shear responses were considered to detect wheel flats, flats, a a parametric parametric study study was was performed performed to tocompare compare the the accuracy accuracy of the of thesystem system us- ingusing acceleration acceleration responses. responses. The The wheel wheel flat flat was was identified identified through through the the envelope envelope spectrum approach by considering the evaluated shear and the accelerations as inputs. The results were compared considering different track responses to determine the precise wheel flatflat due to the type and position of the sensors and the wheel flatflat level. 2.4.2. Flat Geometry 2.4.2. Flat Geometry Figure5 presented a vehicle passing through the virtual sensors. The front wheel of Figure 5 presented a vehicle passing through the virtual sensors. The front wheel of the vehicle has a defect with the following characteristics: wheel flat length (L) is 150 mm the vehicle has a defect with the following characteristics: wheel flat length (L) is 150 mm and flat depth (D) is 1.5 mm. The impact of the wheel flat occurred before position 9. The and flat depth (D) is 1.5 mm. The impact of the wheel flat occurred before position 9. The speed of the vehicle is 60 m/s. The wheel flat vertical profile deviation is defined as: speed of the vehicle is 60 m/s. The wheel flat vertical profile deviation is defined as: D 2π2x Z =−= − 1−1 −cos H (x−−(2(π2rw−−L))),, 00≤≤ x ≤≤ 2πrw (1)(1) 2 2 L

where D isis the the depth depth of of the the flat, flat,L is theis the length length of theof the flat, flat,rw is the is radiusthe radius of the of wheel,the wheel and, andH represents H represents the Heavisidethe Heaviside function. function. When a defective wheel rotates, a periodic impulse occurs due to the wheel flat flat that isis applied to the track by the wheel at a specificspecific frequency.frequency. AfterAfter determiningdetermining thethe wheelwheel perimeter ( (rW)) and and the the speed speed of of the the vehicle vehicle ( (),S the), the frequency frequency of ofthe the periodic periodic impulse impulse re- ferringreferring to tothe the flat flat impact impact frequency frequency can can be be ca calculatedlculated as as follows: follows: S f = = (2)(2) f (2) (2πrW ) Appl. Sci. 2021, 11, 4002 9 of 25

The wheel flat is incorporated in the unevenness profile of the track defined in Figure3 as a periodic irregularity with the shape given by Equation (1).

3. Methodology for Flat Detection 3.1. Deﬁnition of Kurtosis The shape and density of the signal distribution can be visualized by moments about the mean [56]. The mean of a set of data xm, is calculated through the following equation:

1 N xm = ∑ xi (3) N j=1

in which xi is the recorded data sample and N is the number of samples. Therefore, the k-th central moment is calculated as:

N 1 k µk = ∑ xj − x (4) N j=1

where x is the true mean. Kurtosis is a non-dimensional quantity that measures the relative intensity of the signal compared to a Gaussian distribution. Signals with impulsive-type responses lead to high kurtosis values (K). It is deﬁned as the fourth central moment (µ4) 2 divided by the square of variance µ2 as follow [57]: µ = 4 K 2 (5) µ2 For signals containing impulse features, the kurtosis value can be high. Therefore, high amounts of kurtosis indicate that the data include spikes [58]. Since the defect in the wheel causes a spike in the signal, one way to detect the defect in the wheel is to calculate the kurtosis for various frequency bands rather than the original signal.

3.2. Envelope Spectrum Approach to Detect a Wheel with Spectral Kurtosis Analysis Envelope spectrum detection is a technique to perform complex demodulation by shifting each frequency to zero and then applying a low-pass ﬁlter [59]. Therefore, the signal is multiplied by the following factor to perform a complex signal deformation.

X(t) = R(t) exp(2πi f0t) (6)

where, f0 is the center frequency of the band. After the frequency band with the highest kurtosis level is determined, a passband filter is applied to the raw signal to obtain a higher impulse signal to analyze the envelope spectrum. The process to identify a defective wheel is shown in Figure6. As shown in this figure, the method to detect a wheel flat is presented with two main blocks. In the first block of the flowchart, the necessary equations for evaluating the envelope spectrum in each sensor of the installation system are explained, and furthermore, the wheel flat is identified in the second block based on the following criteria. If the responses obtained by all sensors are coincident, the amplitude of the envelope spectrum is the same and indicates the passage of a healthy wheel through the system. However, a significant lag in the amplitude of the envelope spectrum infers a passage of a defective wheel. Appl. Sci. 20202121,, 11,, x 4002 FOR PEER REVIEW 1010 of of 25 25

FigureFigure 6. 6. TheThe methodology methodology to to identify identify a a defective defective wheel wheel.. The selection of the demodulation band for the envelope analysis of a defective wheel The selection of the demodulation band for the envelope analysis of a defective wheel is frequently performed by comparing the spectrum for a healthy wheel to select the is frequently performed by comparing the spectrum for a healthy wheel to select the par- particular frequencies at which the largest change occurred as a result of the fault. It has ticular frequencies at which the largest change occurred as a result of the fault. It has been been identified that spectral kurtosis (SK) gives a very similar indication of the band to be identified that spectral kurtosis (SK) gives a very similar indication of the band to be de- demodulated without requiring historical data. The spectral kurtosis of a signal is obtained modulated without requiring historical data. The spectral kurtosis of a signal is obtained by dividing the main signal into different frequency bands and obtaining the kurtosis of by dividing the main signal into different frequency bands and obtaining the kurtosis of each frequency band [60]. This demonstrates how the peak in the signal changes with each frequency band [60]. This demonstrates how the peak in the signal changes with frequencies and can be used to identify a frequency band within which a signal has the frequenciesmost impulsive and behavior.can be used This to information identify a frequency is needed band to select within the optimalwhich a frequency signal has band the mostto demodulate impulsive andbehavior. perform This an information envelope analysis is needed on to the select recorded the optimal signals. frequency The optimum band tocenter demodulate frequency and and perform bandwidth an envelope combination analysis of a bandpasson the recorded filter to signals. maximize The the optimum kurtosis centerof the filterfrequency output and is describedbandwidth by combination a kurtogram of [ 61a bandpass]. filter to maximize the kurtosis of the filter output is described by a kurtogram [61]. Appl. Sci. 2021, 11, 4002 11 of 25

The kurtogram shows the kurtosis results for a range of window lengths and frequencies and is used to visualize the spectral kurtosis. The principle of the proposed algorithm to obtain the kurtogram is based on the multirate filter-bank structure [60,61]. The algorithm is described first in the case of a binary tree structure and then extended to a 1/3-binary tree structure. The procedure to perform a kurtogram is described in the following steps: 1. Time-series signal R(t) is considered, as the response recorded (shear or acceleration) in positions 1 to 19 (presented in Figure5). 2. Consider h as a low-pass prototype filter. Then construct two low-pass and high- 1 1 1 pass analysis filters h1 and h2, from h, in the frequency bands [0 4 ] and [ 4 2 ] of the sampling frequency, respectively, as follow:

jπ/4 h1 = he j3π/4 (7) h2 = he

Filters h1 and h2 are used to perform the low-pass/high-pass decomposition indicated in Figure7, which is iterated in a pyramidal manner to produce the filter-bank tree, l i where each level has 2 bands. As illustrated in this figure, let Rl(t) be a sequence of coefficients issued from the filter i-th. i is considered as 0, 1, 2, ... , 2l−1 at each level (l) in the decomposition tree. After filtration, the main signal with h1 and h2, kurtosis values would be obtained for each signal. 3. The idea of this step is to define three additional bandpass filters g with frequency 1 1 1 1 1 bands [0 6 ], [ 6 3 ] and [ 3 2 ] of the sampling frequency, as follow:

jπ/6 g1 = ge j3π/6 g2 = ge (8) j5π/6 g3 = ge

i These filters are then used to further decompose each sequence Rl(t) into three sub- sequences, regarding the low, medium, and high frequencies. After decomposition, the kurtosis of all these sub-sequences is computed according to Equation (5). 4. After filtering with the above low-pass and bandpass filters, each signal obtained i from section (ii) is taken into account as the main signal (Rl(t)) and this process is iterated from l = 0 down to level l − 1. The kurtogram would finally be estimated by computing the kurtosis for all sequence signals. The framework to obtain the kurtogram is depicted in Figure7. Moreover, the center frequency and bandwidth frequency for the level with the highest kurtosis can be obtained by the following equations [60,61]: (−l−1) ∆ f = 2 Fs (9)

fo = (i + 0.5)∆ f (10)

where ∆ f is the bandwidth, Fs is the sampling frequency, and fo is the center frequency. Figure8a shows a signal contaminated with noise obtained from position 9, considering layout scheme 1 (see Figure5). Figure8b presents the kurtogram using the above framework for the mentioned signal corresponding to the Alfa Pendular train. Appl.Appl. Sci. Sci. 20202121, 11, 11, x, FOR 4002 PEER REVIEW 12 12of of25 25

FigureFigure 7. 7.FrameworkFramework to obtain to obtain the thekurtogram, kurtogram, the thecentral central frequency frequency,, and andthe frequency the frequency band band for for eacheach signal. signal.

FigureAs shown 8a shows in Figure a signal8, the contaminated maximum kurtosis with noise is obtained obtained as 60.25 from at position level 3, with 9, consid- the cor- 1 1 eringresponding layout scheme band frequency 1 (see Figure equal 5). to Figure [ 16 8 ]. 8b Therefore, presents considering the kurtogramEquations using (9)the andabove (10) , frameworkthe center for and the bandwidth mentioned frequencies signal corresponding for the corresponding to the Alfa signal Pendular are calculated train. as follow: ∆ f = 2(−3−1) × 2000 = 125 Hz (11) fo = (1 + 0.5) × 125 = 187.5 Hz

where the sampling frequency (Fs) is 2000 Hz, the bandwidth frequency (∆ f ) is 125 Hz and the center frequency ( fo) is 187.5 Hz. More details regarding the envelope spectrum approach for wheel ﬂat detection are presented by Mosleh et al. [37]. In the following sections, the envelope spectrum approach is employed to detect a wheel ﬂat. For this purpose, spectral kurtosis is used to calculate the center frequency and bandwidth for each signal as a key preprocessing step before performing the envelope spectrum. The sensitivity of the layout schemes (presented in Appl. Sci. 2021, 11, 4002 13 of 25

Appl. Sci. 2021, 11, x FOR PEER REVIEW 13 of 25

Figure5) is veriﬁed according to the type of sensors, the type of train, and the position of the sensors.

(a)

(b)

FigureFigure 8. 8.(a):(a Main): Main signal signal evaluated evaluated from from position position 9 9regarding regarding layout layout scheme scheme 1 1.. ( (bb):): Spectral Spectral kurto- kurtosis sisfor for a a signal signal corresponding corresponding to to the the Alfa Alfa Pendular Pendular train, train, K maxKmax= = 60.25 60.25 Hz, Hz, optimal optimal window window length length = = 16, 16,center center frequency frequency = = 0.1875 0.1875 kHz, kHz, bandwidth bandwidth frequency frequency = = 0.125 0.125 kHz. kHz.

4. SensitivityAs shown in of Figure Layout 8, Schemes the maximum to the kurtosis Unevenness is obtained Profile as of 60.25 the Rail at level and 3, the with the Different Types of Trains corresponding band frequency equal to [ ]. Therefore, considering Equations (9) and 4.1. Track Response Obtained by the Strain Gauge Setup (10), the center and bandwidth frequencies for the corresponding signal are calculated as follow:The passage of one vehicle of the Alfa Pendular and a freight wagon in a track stretch with the irregularity profile presented in Figure3 is investigated. In all analyses where a ∆ = 2() × 2000 = 125 Hz wheel flat is considered, only the right wheel of the first wheelset is damaged with this (11) type of defect. As shown in = Figure(1 +50,. the5) × three125 layout= 187.5 schemes Hz considered in this study aim to optimize the number and position of the sensors. As illustrated in this figure, layout ∆ wschemehere the 1 sampling includes 12frequency virtual strain( ) is gauges2000 Hz, installed the bandwidth on the rail frequency from a track ( ) stretch is 125 withHz anda lengththe center equivalent frequency to a( wheel) is 187.5 perimeter. Hz. The envelope spectrum approach is used to verifyMore the details sensitivity regarding of layout the envelope scheme 1,spectrum considering approach the unevenness for wheel flat profile detection presented are presentedin Figure by3. MoslehThe speed et al. is [37] considered. In the follo as 60wing m/s sections, for both the the envelope Alfa Pendular spectrum and approach the freight is wagon.employed Figure to detect9 presents a wheel the flat. power For this spectrum purpose, of thespectral envelope kurtosis signal is used for the to calculate defective theand center healthy frequency wheels correspondingand bandwidth to for the each 12 shear signal responses as a key obtained preprocessing on the step rail(position before performing the envelope spectrum. The sensitivity of the layout schemes (presented in Appl. Sci. 2021, 11, x FOR PEER REVIEW 14 of 25

Figure 5) is verified according to the type of sensors, the type of train, and the position of the sensors.

4. Sensitivity of Layout Schemes to the Unevenness Profile of the Rail and the Differ- ent Types of Trains 4.1. Track Response Obtained by the Strain Gauge Setup The passage of one vehicle of the Alfa Pendular and a freight wagon in a track stretch with the irregularity profile presented in Figure 3 is investigated. In all analyses where a wheel flat is considered, only the right wheel of the first wheelset is damaged with this type of defect. As shown in Figure 5, the three layout schemes considered in this study aim to optimize the number and position of the sensors. As illustrated in this figure, layout scheme 1 includes 12 virtual strain gauges installed on the rail from a track stretch with a length equivalent to a wheel perimeter. The envelope spectrum approach is used to verify the sensitivity of layout scheme 1, considering the unevenness profile presented in Figure Appl. Sci. 2021, 11, 4002 3. The speed is considered as 60 m/s for both the Alfa Pendular and the freight wagon.14 of 25 Figure 9 presents the power spectrum of the envelope signal for the defective and healthy wheels corresponding to the 12 shear responses obtained on the rail (position 1 to 12 shown in Figure 5), due to the passage of the Alfa Pendular and freight wagon, consider- ing1 tothe 12 unevenness shown inFigure profile5 ),of due the totrack. the passageIn the following of the Alfa figures, Pendular the vertical and freight dash wagon, lines indicateconsidering the wheel the unevenness flat frequency profile as 22.21 of the Hz. track. In the following figures, the vertical dash lines indicate the wheel flat frequency as 22.21 Hz. e d u t i l p m A

k a e P

(a) (b) ) N (

e d u t i l p m A

k a e P

FigureFigure 9. 9.EnvelopeEnvelope spectrum spectrum analysis analysis for for the the 12 12 strain strain gauges gauges (layout (layout scheme scheme 1) for the Alfa PendularPen- dularand and the the freight freight wagon, wagon, considering considering the the unevenness unevenness proﬁle profile of theof the track, track, (a) ( a) defective a defective wheel wheel Alfa AlfaPendular, Pendular, (b) ( ab) healthy a healthy wheel wheel Alfa Alfa Pendular, Pendular, (c) ( ac) defective a defective wheel wheel freight freight wagon, wagon, and and (d) ( ad) healthy a healthywheel wheel freight freight wagon. wagon.

By applying the envelop spectrum analysis, it is possible to observe a lag between the responses evaluated at the several SGs in the defective wheel, showing that a flat has been detected. On the other hand, as seen in Figure9b,d, the responses in terms of the envelope spectrum in all SGs are very similar (no lag), indicating that there are no flats in both the Alfa Pendular and the freight wagon’s wheels. Moreover, the amplitude variation of the envelope spectral signal can also be seen as an additional indicator to evaluate whether the wheel is healthy or not. As shown in this figure, the amplitude variation of the envelope spectral signal of a defective wheel is higher than a healthy one. The frequency content provides some interesting information. There are some high magnitude peaks corresponding to the bogies passing frequency (Figure9b,d), which can be calculated as S fb = (12) a1

where fb is the bogie frequency and a1 is the regular spacing (see Figure1), which for the Alfa Pendular corresponds to the spacing between bogies (a1 = 19 m) and for the freight wagon, since it has no bogies, corresponds to the spacing between axles (a1 = 6 m). Therefore, fb was calculated as 3.15 and 6 Hz for the Alfa Pendular and the freight wagon, respectively. In other words, Figure9b corresponds to the passage of the healthy wheel Appl. Sci. 2021, 11, x FOR PEER REVIEW 15 of 25

By applying the envelop spectrum analysis, it is possible to observe a lag between the responses evaluated at the several SGs in the defective wheel, showing that a flat has been detected. On the other hand, as seen in Figure 9b,d, the responses in terms of the envelope spectrum in all SGs are very similar (no lag), indicating that there are no flats in both the Alfa Pendular and the freight wagon’s wheels. Moreover, the amplitude variation of the envelope spectral signal can also be seen as an additional indicator to evaluate whether the wheel is healthy or not. As shown in this figure, the amplitude variation of the envelope spectral signal of a defective wheel is higher than a healthy one. The frequency content provides some interesting information. There are some high magnitude peaks corresponding to the bogies passing frequency (Figure 9b,d), which can be calculated as = (12)

where is the bogie frequency and is the regular spacing (see Figure 1), which for the Alfa Pendular corresponds to the spacing between bogies ( = 19 m) and for the freight wagon, since it has no bogies, corresponds to the spacing between axles ( = 6

Appl. Sci. 2021, 11, 4002 m). Therefore, was calculated as 3.15 and 6 Hz for the Alfa Pendular15 of 25 and the freight wagon, respectively. In other words, Figure 9b corresponds to the passage of the healthy wheel of the Alfa Pendular and presents the bogie passing frequency as 3.15 Hz and the ofwheel the Alfa passing Pendular frequenc and presentsy as the22.17 bogie Hz. passing However, frequency in Figure as 3.15 Hz 9a, and corresponded the wheel to the passage passingof a defective frequency wheel as 22.17 of Hz.the Alfa However, Pendular, in Figure only9a, correspondedthe frequency to theof the passage defective wheel (22.21 ofHz) a defectivecan be wheelseen. ofFigure the Alfa 10 Pendular, shows the only kurtosis the frequency values of thefor defective healthy wheel and defective wheels (22.21 Hz) can be seen. Figure 10 shows the kurtosis values for healthy and defective wheels correspondingcorrespondin tog 12 to positions 12 positions of the strain of the gauges strain (layout gauges scheme (layout 1) for thescheme Alfa Pendular. 1) for the Alfa Pendular. s i s o t r u K

Figure 10. 10.Kurtosis Kurtosis values values for healthy for healthy and defective and defective wheels corresponding wheels corresponding to 12 positions of to the 12 positions of the strainstrain gauges gauges for thefor passagethe passage of the Alfa of the Pendular. Alfa Pendular. As presented in this figure, a signal corresponding to a defective wheel has signifi- cantlyAs greater presented impulsiveness, in this especially figure, for a strainsignal gauge corresponding 9, which is located to aftera defective the wheel wheel has signifi- flatcantly impact greater location. impulsiveness, However, since the especially unevenness for profile strain does gauge not generate 9, which strong is varia-located after the wheel tionsflat impact in the signal, location. compared However, to the flat since wheel, the all unevenness kurtosis values profile obtained does for a healthynot generate strong var- wheel are the same for the 12 signals. iations in the signal, compared to the flat wheel, all kurtosis values obtained for a healthy 4.2.wheel Track are Response the same Obtained for by the the Accelerometers12 signals. The previous section has shown that the envelope spectrum approach is able to detect a4.2. defective Track wheel Response from aO healthybtained one by for the 12 A virtualccelerometers strain gauges located in the rail (layout scheme 1). However, as mentioned above, the use of strain gauges for long-term moni- toringThe of wheel previous flat is not section suitable has due shown to their that vulnerability the envelope to water spectrum and electromagnetic approach is able to detect interferences.a defective wheel To compare from the a trackhealthy response one offor strain 12 virtual gauges strain and accelerometers, gauges located both in the rail (layout were placed in the same location. This is shown in layout scheme 2, in Figure5. schemeFigure 1). 11 However,shows the envelope as mentioned spectrum correspondingabove, the use to the of 12 strain acceleration gauges responses for long-term monitor- obtaineding of wheel on the rail flat (position is not 1 tosuitable 12 shown due in Figure to their5) for thevulnerability passage of the to Alfa water Pendular and electromagnetic and the freight wagon, considering the track irregularities presented before. As for the SGs, it is possible to observe that, in the presence of a wheel flat, the responses of each accelerometer in terms of envelope spectrum present significant lag, while for the scenarios with a healthy wheel, this lag is practically insignificant. Moreover, the amplitude of the envelope spectral signal is also considerably different in the two situations, being much higher when the wheel is defective. Appl. Sci. 2021, 11, x FOR PEER REVIEW 16 of 25

interferences. To compare the track response of strain gauges and accelerometers, both were placed in the same location. This is shown in layout scheme 2, in Figure 5. Figure 11 shows the envelope spectrum corresponding to the 12 acceleration responses obtained on the rail (position 1 to 12 shown in Figure 5) for the passage of the Alfa Pendular and the freight wagon, considering the track irregularities presented before. As for the SGs, it is possible to observe that, in the presence of a wheel flat, the responses of each accelerometer in terms of envelope spectrum present significant lag, while for the scenarios with a healthy wheel, this lag is practically insignificant. Moreover, the ampli- Appl. Sci. 2021, 11, 4002 16 of 25 tude of the envelope spectral signal is also considerably different in the two situations, being much higher when the wheel is defective. ) 2 s / m (

e d u t i l p m A

k a e P

(a) (b) ) 2 s / m (

e d u t i l p m A

k a e P

Figure 11. 11. EnvelopeEnvelope spectrum spectrum analysis analysis for for the the 12 12accelerometers accelerometers (layout (layout scheme scheme 2) for 2) the for Alfa the Alfa Pendular and the freight wagon, wagon, considering considering the the unevenness unevenness profile proﬁle of of the the track track for for ( (aa)) a a defective defective wheel Alfa Pendular, Pendular, ( (bb)) a a healthy healthy wheel wheel Alfa Alfa Pendular, Pendular, (c (c) )a adefective defective wheel wheel freight freight wagon, wagon, and and (d) a healthy wheel freight wagon. (d) a healthy wheel freight wagon.

5. Sensitivity of the L Layoutayout S Schemeschemes to the S Signalignal N Noiseoise The signal noise can significantlysignificantly perturb the assessment of the track response. To evaluate its influence influence in the proposed methodology to detect wheel flats, flats, an artificial artificial noise is generated generated due to the the maximum maximum response (shear or acceleration) of the signal and added to the mainmain signal.signal. Therefore, Therefore, new new signals signals considering considering different different noise noise levels levels of 5%, of 5%, 10%, 10% and, and20% 20% of the of maximum the maximum amplitude amplitude of the of initial the initial signal signal are generated. are generated. The unevenness The unevenness profile profileof the trackof the (presented track (presented in Figure in 3Figure) is considered. 3) is considered. Moreover, Moreover, the effectiveness the effectiveness of using of usingshear andshear acceleration and acceleration as input as parametersinput parameters for flat for detection flat detection is compared, is compared, with respect with re- to spectstrain to gauges strain (layout gauges scheme (layout 1) scheme and accelerometers 1) and accelerometers (layout scheme (layout 2), scheme considering 2), consider- the Alfa ingPendular the Alfa and Pendular the freight and wagon. the freight wagon. Figures 12 and 13 show an envelope spectrum detection analysis for 12 shear and acceleration responses obtained on the rail (position 1 to 12 shown in Figure5) for different noise intensities and considering a defective and a healthy wheel of the Alfa Pendular train and the freight wagon. As mentioned before, there are usually two indicators that can be used to detect wheel flats, namely the lag between signals and the amplitude value. It can be observed that the first one can also be used in the present scenario since a clear lag between the signals in the presence of a wheel flat is observed, while in the healthy wheel, all signals coincide. However, when looking at the amplitudes of the signals, it is possible to conclude that this indicator is not suitable when the shear is considered as input. In other words, when the signal is contaminated by noise and track responses evaluate by SGs, the only indicator that effectively distinguishes a defective from a healthy wheel is the clear lag between the amplitude of the envelope spectrum obtained with different sensors. On the other hand, the use of accelerometers is clearly more beneficial than using strain gauges Appl.Appl. Sci. Sci. 20202121, 11, 11, x, 4002FOR PEER REVIEW 17 17of of25 25

to performFigures an12 envelopeand 13 show spectrum an envelope analysis spectrum for the detection detection of analysis defective for wheels, 12 shear since and the accelerationamplitude observedresponses inobtained the scenario on the with rail defective(position wheel1 to 12 scenarioshown in is Figure considerably 5) for differ- higher entthan noise that intensities obtained and with considering a healthy wheel,a defective making and a the healthy amplitude wheel indicator of the Alfa also Pendular suitable trainwhen and using the thisfreigh kindt wagon. of layout As scheme.mentioned before, there are usually two indicators that can be Asused demonstrated to detect wheel in aflats, previous namely study the lag [37 ],between the time-domain signals and response the amplitude is not suitablevalue. Itto can distinguish be observed a healthythat the fromfirst one a defective can also wheelbe used due in tothe the present noise scenario contamination since a clear of the lagsignal. between Therefore, the signals some in signal the presence processing of a is wheel required flat to is extractobserved, fault while detection in the under healthy this wheelsituation., all signals Figure coincide. 14 shows However, the kurtosis when values looking for signalsat the amplitudes with and without of the signals, noise for it theis possibleAlfa Pendular to conclude passage that for this layout indicator scheme is 2. not As presentedsuitable when in this the ﬁgure, shear the is noise-freeconsidered signal as input.is signiﬁcantly In other words, more impulsive, when the signal causing is thecontaminated envelope spectrum by noise analysis and track to responses effectively eval- detect uatethe by fault. SGs, As the an only example, indicator regarding that effectively accelerometer distinguishes 9, when a the defective signal is from not markeda healthy by wheelnoise, is the the kurtosis clear lag value between is approximately the amplitude 2400. of the envelope spectrum obtained with different Whensensors. a On 20% the noise other contamination hand, the use of is accelerometers imputed tothe is clearly signal, more the beneficial kurtosis valuethan usingdecreases strain togauges 400 in to relation perform to an the envelope noise-free spectrum signal, analysis but the for pattern the detection remains of constant, defec- tiveand wheels, the maximum since the kurtosisamplitude value observed corresponds in the scenario to accelerometer with defective 9, whichwheel scenario is closer is to considerablythe impact higher of the than flat. that This obtained means with that a the healthy framework wheel, making is effective the amplitude even when indi- the catornoise-to-signal also suitable ratio when is using high. this kind of layout scheme.

Freight wagon- layout scheme 1 (5% noise)

Alfa Pendular train- layout scheme 1 (10% noise) e d u t i l p m A

k a e P

FigureFigure 12. 12. EnvEnvelopeelope spectrum spectrum analysis analysis for forthe the12 strain 12 strain gauges gauges (layout (layout scheme scheme 1) with 1) withdifferent different noisenoise intensities intensities and and considering considering a defective a defective and and healthy healthy wheel wheel of ofthe the Alfa Alfa Pendular Pendular train train and and the the freightfreight wagon. wagon.

Appl. Sci. 2021, 11, x FOR PEER REVIEW 18 of 25

Freight wagon-layout scheme 2 (5% noise)

Appl.Appl. Sci. Sci. 20202121, 11, 11, x, 4002FOR PEER REVIEW 18 18of of25 25

Freight wagon-layout scheme 2 (5% noise)

Alfa Pendular train-layout scheme 2 (10% noise)

Figure 13. Envelope spectrum analysis for the 12 accelerometers (layout scheme 2) with different noise intensities and considering a defective and healthy wheel of the Alfa Pendular train and the freight wagon.

As demonstrated in a previous study [37], the time-domain response is not suitable to distinguish a healthy from a defective wheel due to the noise contamination of the signal. Therefore, some signal processing is required to extract fault detection under this sit- uation. Figure 14 shows the kurtosis values for signals with and without noise for the Alfa Pendular passage for layout scheme 2. As presented in this figure, the noise-free signal is FigureFigure 13. 13. EnvelopeEnvelope spectrum spectrum analysis analysis for for the the 12 12accelerometers accelerometers (layout (layout scheme scheme 2) with 2) with different different noisenoisesignificantly intensities intensities moreand and considering impulsive, considering a causingdefective a defective the and and envelope healthy healthy wheel spectrum wheel of ofthe the analysisAlfa Alfa Pendular Pendular to effectively train train and and detectthe the freightfreightthe fault. wagon. wagon. As an example, regarding accelerometer 9, when the signal is not marked by noise, the kurtosis value is approximately 2400. As demonstrated in a previous study [37], the time-domain response is not suitable to distinguish a healthy from a defective wheel due to the noise contamination of the signal. Therefore, some signal processing is required to extract fault detection under this sit- uation. Figure 14 shows the kurtosis values for signals with and without noise for the Alfa Pendular passage for layout scheme 2. As presented in this figure, the noise-free signal is significantly more impulsive, causing the envelope spectrum analysis to effectively detect s i s o

the t fault. As an example, regarding accelerometer 9, when the signal is not marked by r u

noise,K the kurtosis value is approximately 2400.

s i s o t Figurer 14. KurtosisKurtosis values values for for signals signals with with and and without without noise noise for for the the Alfa Alfa Pendular Pendular passage passage for for u layoutK layout schemescheme 2.2.

6. Sensitivity of the Layout Schemes to the Number and Location of the Accelerometers As stated above, in order to reduce installation and maintenance costs, it is of the utmost importance to reduce the number of sensors without compromising the quality of the output results. Moreover, it is important to demonstrate that the system, regardless of the position of the sensors, is always effective in detecting wheel ﬂats. Therefore, Figure 14. Kurtosis values for signals with and without noise for the Alfa Pendular passage for this section aims to demonstrate the sensitivity of the layout schemes to the number layout scheme 2. Appl. Sci. 2021, 11, x FOR PEER REVIEW 19 of 25

When a 20% noise contamination is imputed to the signal, the kurtosis value de- creases to 400 in relation to the noise-free signal, but the pattern remains constant, and the maximum kurtosis value corresponds to accelerometer 9, which is closer to the impact of the flat. This means that the framework is effective even when the noise-to-signal ratio is high.

6. Sensitivity of the Layout Schemes to the Number and Location of the Accelerome- ters As stated above, in order to reduce installation and maintenance costs, it is of the utmost importance to reduce the number of sensors without compromising the quality of Appl. Sci. 2021, 11, 4002 the output results. Moreover, it is important to demonstrate that the system, regardless19 of of 25 the position of the sensors, is always effective in detecting wheel flats. Therefore, this section aims to demonstrate the sensitivity of the layout schemes to the number of sensors (strain gauges or accelerometers) and the location where they are installed. Consequently, of sensors (strain gauges or accelerometers) and the location where they are installed. inConsequently, addition to the in addition 12 positions to the selected 12 positions on the selected rail between on the the rail two between sleepers, the two seven sleepers, more accelerometersseven more accelerometers (positions 13 (positionsto 19 shown 13 in to Figure 19 shown 5) were in Figureconsidered5) were on consideredthe rail located on abovethe rail each located sleeper above (layout each scheme sleeper 3). (layout The robustness scheme 3). of Thethe proposed robustness method of the for proposed detect- ingmethod a wheel for detectingflat with fewer a wheel sensors flat with without fewer compromising sensors without data compromising quality is discussed data quality in this is section.discussed in this section. Figure 15 shows the envelope spectrum detection analysis for seven accelerometersaccelerometers (positions 13 to 19 shown in Figure 5 5)) consideringconsidering aa highhigh noisenoise intensityintensity levellevel inin thethe signalsignal obtained in a scenario with a healthy and defective wheel of the Alfa Pendular and the freight wagon. wagon. The The results results are are consistent consistent with with those those obtained obtained in inprev previousious sections. sections. As de- As picteddepicted in inFigure Figure 15, 15 there, there is a is lag a lagbetween between the thesignals signals from from a defective a defective wheel, wheel, and andthe maximumthe maximum amplitude amplitude corresponds corresponds to accelerometer to accelerometer 17. As 17. presented As presented in Figure in Figure5, accel-5, erometeraccelerometer 17 is located 17 is located just after just the after flat theimpact flat impactlocation. location. Again, another Again, indicator another indicatorincluded inincluded this study in this to studydistinguish to distinguish defective defective wheels from wheels healthy from healthyones is onesthe amplitude is the amplitude of the envelopeof the envelope spectrum. spectrum. As expected, As expected, the amplitude the amplitude of the envelope of the envelopespectrum spectrumfor a defective for a wheeldefective is significantly wheel is significantly higher than higher for a than healthy for a one. healthy Regardless one. Regardless of the position of the positionof the sen- of sorsthe sensors (between (between two sleepers two sleepers or above or above each each sleeper), sleeper), the the system system is isefficient efficient in in detecting wheel flats. flats. Figure 16 presents an additional optimization of the sensor layout, specifying the optimum number and positions of thethe accelerometers.

Alfa Pendular train-layout scheme 3 10% noise

Appl. Sci. 2021, 11, x FOR PEER REVIEW 20 of 25

Freight wagon-layout scheme 3 10% noise

FigureFigure 15. 15. EnvelopeEnvelope spectrum spectrum analysis analysis for for seven seven accelerometers accelerometers (layout (layout scheme scheme 3) considering 3) considering a defectivea defective and and a healthy a healthy wheel wheel of ofthe the Alfa Alfa Pendular Pendular and and the the freight freight wagon wagon (noise (noise = 10%). = 10%). ) ) 2 2 s s / / m m ( (

e e d d u u t t i i l l p p m m A A

k k a a e e P P

Figure 16. Optimization of the sensor layout.

The installation along an equivalent wheel perimeter length (2 = 2.7 m) is useful for monitoring the entire perimeter of the wheel. Figure 16 shows that from an economic point of view, two accelerometers are needed to distinguish a defective from a healthy wheel. Regarding the position of the sensors, it is better to install them so that, with every rotation of the wheel, each sensor monitors half the length of the wheel perimeter. The length of the wheel perimeter is 2 = 2.7 m. Therefore, each time the defective wheel contacts the surface of the rail, the same impact will occur again 2.7 m later. For example, if the flat impact location occurs before accelerometer 19 (see Figure 5), it means that in the previous rotation of the wheel, the accelerometer 14, which is located 3.0 m away, could also detect the impact at a close distance. For this reason, it is recommended that accelerometers should be installed in such a way that for each wheel rotation, one accelerometer could monitor half of the wheel circumference (2.7/2 = 1.35 m), which is the distance between three sleepers, to guarantee that at least one accelerometer is close to the location of the flat impact. In the actual installation, it is not necessary to install 12 sensors along the rail to identify the profiles of the wheels during the operation of the vehicle.

7. Sensitivity of the Layout Schemes to the Severity of the Flat As previously stated, when the envelope spectrum analysis is carried out using acceleration signals, the amplitudes of the envelope spectrum obtained in a scenario with Appl. Sci. 2021, 11, x FOR PEER REVIEW 20 of 25

Freight wagon-layout scheme 3 10% noise

Figure 15. Envelope spectrum analysis for seven accelerometers (layout scheme 3) considering a Appl. Sci. 2021, 11, 4002 defective and a healthy wheel of the Alfa Pendular and the freight wagon (noise = 10%). 20 of 25

10 Accelerometer 13-defective 9 Accelerometer 16-defective Accelerometer 13-healthy 8 Accelerometer 16-healthy

) Wheel flat frequency 2 7

Peak Amplitude (m/s Amplitude Peak 3

0 0 20 40 60 80 100 120 140 Frequency (Hz) Figure 16. Optimization of the sensor layout. Figure 16. Optimization of the sensor layout.

The installation along an equivalent wheel perimeter length (2πrw = 2.7 m) is useful The installation along an equivalent wheel perimeter length (2𝜋𝑟 =2.7 m) is useful for monitoring the entire perimeter of the wheel. Figure 16 shows that from an economic forpoint monitoring of view, the two entire accelerometers perimeter of are the needed wheel. to Figure distinguish 16 shows a defective that from from an economic a healthy pointwheel. of Regardingview, two theaccelerometers position of the are sensors, needed it to is distinguish better to install a defective them so from that, witha healthy every wheel.rotation Regarding of the wheel, the position each sensor of the monitorssensors, it half is better the length to install of thethem wheel so that, perimeter. with every The rotation of the wheel, each sensor monitors half the length of the wheel perimeter. The length of the wheel perimeter is 2πrw = 2.7 m. Therefore, each time the defective wheel lengthcontacts of thethe surfacewheel perimeter of the rail, is the 2𝜋𝑟 same =2.7 impact m. Therefore, will occur againeach 2.7time m the later. defective For example, wheel if contactsthe flat the impact surface location of the occurs rail, the before same accelerometer impact will occur 19 (see again Figure 2.75 m), itlater. means For thatexample, in the ifprevious the flat impact rotation location of the wheel, occurs the before accelerometer accelerometer 14, which 19 (see is located Figure 3.0 5), m it away, means could that also in thedetect previous the impact rotation at a closeof the distance. wheel, the For acce thislerometer reason, it is14, recommended which is located thataccelerometers 3.0 m away, couldshould also be detect installed the inimpact such aat way a close that distance. for each For wheel this rotation, reason, oneit is accelerometerrecommended couldthat accelerometersmonitor half of should the wheel be circumferenceinstalled in such (2.7/2 a way = 1.35 that m), for which each is wheel the distance rotation, between one accelerometerthree sleepers, could to guarantee monitor thathalf atof least the wheel one accelerometer circumference is close(2.7/2 to = 1.35 the location m), which of theis the flat distanceimpact. between In the actual three installation, sleepers, to guarantee it is not necessary that at least to install one accelerometer 12 sensors along is close the to rail the to locationidentify of the the profiles flat impact. of the In wheels the actual during installation, the operation it is not of thenecessary vehicle. to install 12 sensors along the rail to identify the profiles of the wheels during the operation of the vehicle. 7. Sensitivity of the Layout Schemes to the Severity of the Flat 7. SensitivityAs previously of the Layout stated, Schemes when the to envelopethe Severity spectrum of the Flat analysis is carried out using accelerationAs previously signals, stated, the amplitudes when the of envelope the envelope spectrum spectrum analysis obtained is carried in a scenario out using with accelerationhealthy wheels signals, and inthe another amplitudes with defectiveof the envelope ones are spectrum significantly obtained different. in a scenario The amplitude with of the envelope spectrum for a defective wheel is greater than for a healthy one. The present section aims to illustrate the sensitivity of the layout schemes to the severity of the flat. Hence, the influence of the flat geometry on the acceleration response is investigated in this section. In addition to the flat geometry used in the previous sections (L = 150 mm and D = 1.5 mm), other flat geometries are also considered in this section with different flat lengths (25 mm < L < 80 mm) and flat depths (0.1 mm < D < 1 mm). Moreover, signal contamination with 10% noise and the track irregularity profiles presented before are also considered in the analyses discussed in the present section. Figure 17 shows the envelope spectrum detection analysis for 12 acceleration responses obtained on the rail (position 1 to 12 shown in Figure5) considering a healthy and a defective wheel of the Alfa Pendular with 10% noise incorporated in the signals. Again, the layout composed by accelerometers proved to be efficient since both the lag and the amplitude indicators are suitable for distinguishing a healthy from a defective wheel, regardless of the flat geometry properties. From the results presented in the above figure, it can be concluded that the system is effective for the range of small to severe flat properties, considering a combination of noise level and rail unevenness. Appl. Sci. 2021, 11, x FOR PEER REVIEW 21 of 25

healthy wheels and in another with defective ones are significantly different. The amplitude of the envelope spectrum for a defective wheel is greater than for a healthy one. The present section aims to illustrate the sensitivity of the layout schemes to the severity of the flat. Hence, the influence of the flat geometry on the acceleration response is investigated in this section. In addition to the flat geometry used in the previous sections (L = 150 mm and D = 1.5 mm), other flat geometries are also considered in this section with different flat lengths (25 mm < L < 80 mm) and flat depths (0.1 mm < D < 1 mm). Moreover, signal contamination with 10% noise and the track irregularity profiles presented before are also considered in the analyses discussed in the present section. Figure 17 shows the envelope spectrum detection analysis for 12 acceleration responses obtained on the rail (position 1 to 12 shown in Figure 5) considering a healthy and a defective wheel of the Alfa Pendular with 10% noise incorporated in the signals. Again, the layout composed by accelerometers proved to be efficient since both the lag and the amplitude indicators are suitable for distinguishing a healthy from a defective wheel, regardless of the flat geometry properties. From the results presented in the above figure, it Appl. Sci. 2021, 11, 4002 21 of 25 can be concluded that the system is effective for the range of small to severe flat properties, considering a combination of noise level and rail unevenness.

L = 80 mm, D = 1 mm L = 60 mm, D = 0.5 mm ) 2 s / m (

e d u t i l p m A

k a e P

L = 40 mm, D = 0.2 mm L = 30 mm, D = 0.15 mm

Appl. Sci. 2021, 11, x FOR PEER REVIEW 22 of 25

Healthy wheel ) 2 s / m (

e d u t i l p m A

k a e P

Figure 17. Envelope spectrum analysis for 12 accelerometers (layout scheme 2) considering a defective and a healthy wheel Figure 17. Envelope spectrum analysis for 12 accelerometers (layout scheme 2) considering a de- of the Alfa Pendular due to the severity of the flat considering 10% noise. fective and a healthy wheel of the Alfa Pendular due to the severity of the flat considering 10% noise. 8. Conclusions 8. Conclusions This paper aimed to identify the best options regarding the type of sensors and their location in a wayside system for wheel flat detection. Based on this, a multisensory This paper aimedlayout to scheme identify has the been best proposed options to regarding detect the the presence type of of sensors wheel flats and on their both passenger location in a waysideandfreight system trains. for wheel Hence, flat detection. a wheel flat Based recognition on this, a methodology multisensory has layout been developed scheme has beenbased proposed on simulated to detect signals the presence obtained of with wheel a vehicle-track flats on both interaction passenger model. and Both shear freight trains. Hence, a wheel flat recognition methodology has been developed based on simulated signals obtained with a vehicle-track interaction model. Both shear and acceleration measurement points were considered to examine the sensitivity of the layout schemes not only to the type of sensors (strain gauge and accelerometer) but also to the position where they are installed. By considering the evaluated shear and accelerations in 19 positions of the track as inputs, the wheel flat was identified by the envelope spectrum approach with spectral kurtosis analysis. The influence of the type of sensors and installation location on the accuracy of the wheel flat detection system is analyzed. Two types of trains, namely the Alfa Pendular passenger train and a freight wagon, were considered. This study allowed us to draw the following conclusions: 1. The methodology proposed in this work to detect wheel flats and distinguish defective from healthy wheels is based on the envelope spectrum analysis. In this regard, extracting the impulsive signal with amplitude modulation proved a critical preprocessing step before performing the envelope spectrum. In this study, the center frequency and the bandwidth frequency of the signal were obtained first, and then this information was used as input to detect defective wheels through the envelope spectrum approach. 2. In situations where the signal is significantly contaminated by noise, the use of a layout scheme composed of accelerometers is clearly more advantageous than using the strain gauges to perform an envelope spectrum to detect defective wheels. When the envelope spectrum is performed with the acceleration as input, there are significant differences in the amplitudes of the envelope spectrum response obtained in a scenario with healthy wheels and in a scenario with a defective one. 3. This finding confirms the importance of kurtosis to obtain the center frequency because the impulse signal obtained with the accelerometer is more obvious than that obtained with the strain gauge. Moreover, in noised contaminated signals, the impulse signal is even more visible due to the defective wheel, and with kurtosis, the center frequency and bandwidth would be obtained more accurately. 4. To detect a defective wheel, there is no need to install 12 strain gauges or accelerometers along the perimeter of a wheel since, according to the results obtained in this work, it is clear that the same results are achieved with only two accelerometers on the rail web. Appl. Sci. 2021, 11, 4002 22 of 25

and acceleration measurement points were considered to examine the sensitivity of the layout schemes not only to the type of sensors (strain gauge and accelerometer) but also to the position where they are installed. By considering the evaluated shear and accelerations in 19 positions of the track as inputs, the wheel flat was identified by the envelope spectrum approach with spectral kurtosis analysis. The influence of the type of sensors and installation location on the accuracy of the wheel flat detection system is analyzed. Two types of trains, namely the Alfa Pendular passenger train and a freight wagon, were considered. This study allowed us to draw the following conclusions: 1. The methodology proposed in this work to detect wheel flats and distinguish defective from healthy wheels is based on the envelope spectrum analysis. In this regard, extracting the impulsive signal with amplitude modulation proved a critical preprocessing step before performing the envelope spectrum. In this study, the center frequency and the bandwidth frequency of the signal were obtained first, and then this information was used as input to detect defective wheels through the envelope spectrum approach. 2. In situations where the signal is significantly contaminated by noise, the use of a layout scheme composed of accelerometers is clearly more advantageous than using the strain gauges to perform an envelope spectrum to detect defective wheels. When the envelope spectrum is performed with the acceleration as input, there are significant differences in the amplitudes of the envelope spectrum response obtained in a scenario with healthy wheels and in a scenario with a defective one. 3. This finding confirms the importance of kurtosis to obtain the center frequency because the impulse signal obtained with the accelerometer is more obvious than that obtained with the strain gauge. Moreover, in noised contaminated signals, the impulse signal is even more visible due to the defective wheel, and with kurtosis, the center frequency and bandwidth would be obtained more accurately. 4. To detect a defective wheel, there is no need to install 12 strain gauges or accelerometers along the perimeter of a wheel since, according to the results obtained in this work, it is clear that the same results are achieved with only two accelerometers on the rail web. 5. The results show that the system, regardless of the position of the sensors and the severity of the flat, is always effective in detecting wheel flats, which is a major advantage regarding the installing process. For the final development of the proposed methodology, it is imperative to develop an automatic indicator to distinguish a defective wheel from a healthy one. Moreover, based on artificial intelligence techniques, a threshold should be defined to detect a damaged wheel. Finally, the application of the proposed method will be studied considering different speeds and unevenness profiles and tested under real field conditions by considering envi- ronmental disturbance, measurement error, and electronic interference. The optimization of the system should be thoroughly evaluated with respect to the various characteristics of wheel flats and several types of trains.

Author Contributions: Conceptualization, A.M.; methodology, A.M.; software, P.A.M., A.M.; validation, A.M., P.A.M.; formal analysis, P.A.C., R.C.; investigation, A.M.; writing—original draft preparation, A.M.; writing—review and editing, P.A.M., P.A.C., R.C.; supervision, R.C., P.A.C. All authors have read and agreed to the published version of the manuscript. Funding: This work was ﬁnancially supported by: Base Funding-UIDB/04708/2020 and Program- matic Funding-UIDP/04708/2020 of the CONSTRUCT-Instituto de I&D em Estruturas e Construções funded by national funds through the FCT/MCTES (PIDDAC). Moreover, the results presented here are within the scope of the project PEDDIR DEMO, with reference NORTE-01-0247-FEDER-006397, founded by ANI (P2020|COMPETE—Projetos Demonstradores em Copromoção). The authors also thank the ﬁnancial sponsorship provided by the European Commission to the H2020 MARIE SKŁODOWSKA-CURIE RISE Project “RISEN: Rail Infrastructure Systems Engineering Network” Appl. Sci. 2021, 11, 4002 23 of 25

(Grant No. 691135), and to the H2020 SHIFT2RAIL project “IN2TRACK2—Research into enhanced track and switch and crossing system 2”. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. Nielsen, J.C.O.; Johansson, A. Out-of-round railway wheels-a literature survey. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2000, 214, 79–91. [CrossRef] 2. Barke, D.W.; Chiu, W.K. A Review of the Effects of Out-Of-Round Wheels on Track and Vehicle Components. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2005, 219, 151–175. [CrossRef] 3. Johansson, A.; Andersson, C. Out-of-round railway wheels—A study of wheel polygonalization through simulation of three- dimensional wheel–rail interaction and wear. Veh. Syst. Dyn. 2005, 43, 539–559. [CrossRef] 4. Nielsen, J. Out-of-Round Railway Wheels; Chalmers University of Technology: Göteborg, Sweden, 2009. 5. Fesharakifard, R.; Dequidt, A.; Tison, T.; Coste, O. Dynamics of railway track subjected to distributed and local out-of-round wheels. Mech. Ind. 2013, 14, 347–359. [CrossRef] 6. Lan, Q.; Dhanasekar, M.; Handoko, Y.A. Wear damage of out-of-round wheels in rail wagons under braking. Eng. Fail. Anal. 2019, 102, 170–186. [CrossRef] 7. Dukkipati, R.V.; Dong, R. Impact Loads due to Wheel Flats and Shells. Veh. Syst. Dyn. 1999, 31, 1–22. [CrossRef] 8. Wu, T.; Thompson, D. A hybrid model for the noise generation due to railway wheel flats. J. Sound Vib. 2002, 251, 115–139. [CrossRef] 9. Uzzal, R.U.A.; Ahmed, W.; Rakheja, S. Dynamic analysis of railway vehicle-track interactions due to wheel flat with a pitch-plane vehicle model. J. Mech. Eng. 2008, 39, 86–94. [CrossRef] 10. Vale, C. Influência da Qualidade dos Sistemas Ferroviários no Comportamento Dinâmico e no Planeamento da Manutenção Preventiva de Vias de Alta Velocidade. Ph.D. Thesis, Faculty of Engineering of the University of Porto, Porto, Portugal, 2010. (In Portuguese). 11. Li, Y.; Liu, J.; Wang, Y. Railway Wheel Flat Detection Based on Improved Empirical Mode Decomposition. Shock. Vib. 2016, 2016, 1–14. [CrossRef] 12. Alemi, A.; Corman, F.; Pang, Y.; Lodewijks, G. Reconstruction of an informative railway wheel defect signal from wheel–rail contact signals measured by multiple wayside sensors. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2018, 233, 49–62. [CrossRef] 13. Palo, M. Condition Monitoring of Railway Vehicles, A Study on Wheel Condition for Heavy Haul Rolling Stock; Operation and Maintenance Engineering Luleå University of Technology: Luleå, Sweden, 2012. 14. Baasch, B.; Heusel, J.; Roth, M.; Neumann, T. Train Wheel Condition Monitoring via Cepstral Analysis of Axle Box Accelerations. Appl. Sci. 2021, 11, 1432. [CrossRef] 15. Song, Y.; Wang, Z.; Liu, Z.; Wang, R. A spatial coupling model to study dynamic performance of pantograph-catenary with vehicle-track excitation. Mech. Syst. Signal Process. 2021, 151, 107336. [CrossRef] 16. Luo, R. Anti-sliding control simulation of railway vehicle braking. Chin. J. Mech. Eng. 2008, 44, 35–40. [CrossRef] 17. Bosso, N.; Gugliotta, A.; Zampieri, N. Wheel flat detection algorithm for onboard diagnostic. Measurement 2018, 123, 193–202. [CrossRef] 18. Cavuto, A.; Martarelli, M.; Pandarese, G.; Revel, G.; Tomasini, E.P. Train wheel diagnostics by laser ultrasonics. Measurement 2016, 80, 99–107. [CrossRef] 19. Amini, A.; Entezami, M.; Papaelias, M. Onboard detection of railway axle bearing defects using envelope analysis of high frequency acoustic emission signals. Case Stud. Nondestruct. Test. Eval. 2016, 6, 8–16. [CrossRef] 20. Zhang, Z.; Entezami, M.; Stewart, E.; Roberts, C. Enhanced fault diagnosis of roller bearing elements using a combination of empirical mode decomposition and minimum entropy deconvolution. Proc. Inst. Mech. Eng. Part C J. Mech. Eng. Sci. 2015, 231, 655–671. [CrossRef] 21. Meixedo, A.; Goncalves, A.; Calcada, R.; Gabriel, J.; Fonseca, H.; Martins, R. Weighing in motion and wheel defect detection of rolling stock. In Proceedings of the 2015 3rd Experiment International Conference (exp.at’15), Ponta Delgada, Portugal, 2–4 June 2015. 22. Amini, A.; Entezami, M.; Huang, Z.; Rowshandel, H.; Papaelias, M. Wayside detection of faults in railway axle bearings using time spectral kurtosis analysis on high-frequency acoustic emission signals. Adv. Mech. Eng. 2016, 8.[CrossRef] 23. Colaço, A.; Costa, P.A.; Connolly, D.P. The influence of train properties on railway ground vibrations. Struct. Infrastruct. Eng. 2016, 12, 517–534. [CrossRef] 24. Mosleh, A.; Meixedo, A.; Costa, P.; Calçada, R. Trackside Monitoring Solution for Weighing in Motion of Rolling Stock. In Proceedings of the TESTE2019—2nd Conference on Testing and Experimentations in Civil Engineering—Proceedings, Porto, Portugal, 19–21 February 2019. [CrossRef] 25. Mosleh, A.; Costa, P.A.; Calçada, R. A new strategy to estimate static loads for the dynamic weighing in motion of railway vehicles. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2020, 234, 183–200. [CrossRef] 26. Kouroussis, G.; Kinet, D.; Moeyaert, V.; Dupuy, J.; Caucheteur, C. Railway structure monitoring solutions using fibre Bragg grating sensors. Int. J. Rail Transp. 2016, 4, 135–150. [CrossRef] Appl. Sci. 2021, 11, 4002 24 of 25

27. Alexandrou, G.; Kouroussis, G.; Verlinden, O. A comprehensive prediction model for vehicle/track/soil dynamic response due to wheel flats. J. Rail Rapid Transit 2016, 230, 1088–1104. [CrossRef] 28. Mosleh, A.; Costa, P.; Calçada, R. Development of a Low-Cost Trackside System for Weighing in Motion and Wheel Defects Detection. Int. J. Railw. Res. 2020, 7, 1–9. 29. Stratman, B.; Yongming, L.; Sankaran, M. Structural Health Monitoring of Railroad Wheels Using Wheel Impact Load Detectors. J. Fail. Anal. Preven. 2007, 7, 218–225. [CrossRef] 30. Liu, X.; Ni, Y. Wheel tread defect detection for high-speed trains using FBG-based online monitoring techniques. Smart Struct. Syst. 2018, 21, 687–694. 31. Gao, R.; He, Q.; Feng, Q. Railway Wheel Flat Detection System Based on a Parallelogram Mechanism. Sensors 2019, 19, 3614. [CrossRef] 32. Liang, M.; Bozchalooi, I.S. An energy operator approach to joint application of amplitude and frequency-demodulations for bearing fault detection. Mech. Syst. Signal Process. 2010, 24, 1473–1494. [CrossRef] 33. Zhou, C.; Gao, L.; Xiao, H.; Hou, B. Railway Wheel Flat Recognition and Precise Positioning Method Based on Multisensor Arrays. Appl. Sci. 2020, 10, 1297. [CrossRef] 34. Merainani, B.; Benazzouz, D.; Rahmoune, C. Early detection of tooth crack damage in gearbox using empirical wavelet transform combined by Hilbert transform. J. Vib. Control. 2017, 23, 1623–1634. [CrossRef] 35. Lv, Y.; Ge, M.; Zhang, Y.; Yi, C.; Ma, Y. A Novel Demodulation Analysis Technique for Bearing Fault Diagnosis via Energy Separation and Local Low-Rank Matrix Approximation. Sensors 2019, 19, 3755. [CrossRef] 36. Meixedo, A.; Goncalves, A.; Calcada, R.; Gabriel, J.; Fonseca, H.; Martins, R. On-line monitoring system for tracks. In Proceedings of the 3rd Experiment International Conference: Online Experimentation, Ponta Delgada, Portugal, 2–4 June 2015. 37. Mosleh, A.; Montenegro, P.; Costa, P.A.; Calçada, R. An approach for wheel flat detection of railway train wheels using envelope spectrum analysis. Struct. Infrastruct. Eng. 2020, 1–20. [CrossRef] 38. ANSYS®. Academic Research, Release 19.2; ANSYS Inc.: Canonsburg, PA, USA, 2018. 39. Neto, J.; Montenegro, P.A.; Vale, C.; Calçada, R. Evaluation of the train running safety under crosswinds—A numerical study on the influence of the wind speed and orientation considering the normative Chinese hat model International. J. Rail Transp. 2020, 1–28. [CrossRef] 40. Neto, J.; Montenegro, P.A.; Calçada, R. An innovative approach for an inverse dynamic model of a freight wagon. In Proceedings of the 1st International Symposium on Risk Analysis and Safety of Complex Structures and Components, Porto, Portugal, 1–2 July 2019. 41. European Standard. Railway Applications Railway Applications—Track-Rail-Part1: Vignole Railway 46 kg/m and above; peEN 13674–1, Final Draft; European Union: Brussels, Belgium, 2002. 42. Zhai, W.; Wang, K.; Cai, C. Fundamentals of vehicle-track coupled dynamics. Veh. Syst. Dyn. 2009, 47, 1349–1376. [CrossRef] 43. ERRI D 214/RP 5. Rail Bridges for Speeds > 200 km/h: NUMERICAL Investigation of the Effect of Track Irregularities at Bridge Resonance; European Rail Research Institute: Utrecht, The Netherlands, 1999. 44. UIC 774-3-R. Track/Bridge Interaction—Recommendations for Calculation, 2nd ed.; International Union of Railways (UIC): Paris, France, 2001. 45. Wu, Y.D.; Yang, Y.B. Steady-state response and riding comfort of trains moving over a series of simply supported bridges. Eng. Struct. 2003, 25, 251–265. [CrossRef] 46. ERRI D 202/RP 11. Improved Knowledge of Forces in CWR Track (Including Switches): Parametric Study and Sensivity Analysis of Cwerri; European Rail Research Institute: Utrecht, The Netherlands, 1999. 47. Auersch, L. Dynamic interaction of various beams with the underlying soil—Finite and infinite, half-space and Winkler models. Eur. J. Mech. A/Solids 2008, 27, 933–958. [CrossRef] 48. Neves, S.; Montenegro, P.; Azevedo, A.; Calçada, R. A direct method for analyzing the nonlinear vehicle–structure interaction. Eng. Struct. 2014, 69, 83–89. [CrossRef] 49. Montenegro, P.A.; Neves, S.G.M.; Azevedo, A.F.M.; Calçada, R. A nonlinear vehicle-structure interaction methodology with wheel-rail detachment and reattachment. In Proceedings of the COMPDYN 2013—4th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering, Kos, Greece, 12–14 June 2013. 50. Neves, S.; Azevedo, A.; Calçada, R. A direct method for analyzing the vertical vehicle–structure interaction. Eng. Struct. 2012, 34, 414–420. [CrossRef] 51. Montenegro, P.; Neves, S.; Calçada, R.; Tanabe, M.; Sogabe, M. Wheel-rail contact formulation for analyzing the lateral train- structure dynamic interaction. Comput. Struct. 2015, 152, 200–214. [CrossRef] 52. Hertz, H. Ueber die Berührung fester elastischer Körper [On the contact of elastic solids]. J. Reine Angew. Math. 1882, 92, 156–171. 53. Kalker, J.J. Book of Tables for the Hertzian Creep-Force Law; Delft University of Technology: Budapest, Hungary, 1996. 54. MATLAB®. Release R2018a; The MathWorks Inc.: Natick, MA, USA, 2018. 55. Montenegro, P.A. A Methodology for the Assessment of the Train Running Safety on Bridges; Faculty of Engineering of the University of Porto: Porto, Portugal, 2015. 56. Abell, M.L.; Braselton, J.P.; Rafter, J.A. Statistics with Mathematica; Academic Press: New York, NY, USA, 1999. 57. Brandt, A. Noise and Vibration Analysis: Signal Analysis and Experimental Procedures; John Wiley & Sons: New York, NY, USA, 2011. Appl. Sci. 2021, 11, 4002 25 of 25

58. Entezami, M.; Roberts, C.; Weston, P.; Stewart, E.; Amini, A.; Papaelias, M. Perspectives on railway axle bearing condition monitoring. Proc. Inst. Mech. Eng. Part F J. Rail Rapid Transit 2019, 234, 17–31. [CrossRef] 59. Hasan, T. Complex Demodulation: Some Theory and Applications. Handb. Stat. 1983, 3, 125–156. 60. Antoni, J. Fast computation of the kurtogram for the detection of transient faults. Mech. Syst. Signal Process. 2007, 21, 108–124. [CrossRef] 61. Antoni, J. The spectral kurtosis: A useful tool for characterising non-stationary signals. Mech. Syst. Signal Process. 2006, 20, 282–307. [CrossRef]