- Email: [email protected]
Investigations to introduce the probability of detection method for ultrasonic inspection of hollow axles at Deutsche Bahn
Available online at www.sciencedirect.com Available online at www.sciencedirect.com
ScienceDirect ScienceDirect
Structural Integrity 00 (2017) 000–000 Available online at www.sciencedirect.com Available online atProcedia www.sciencedirect.com Structural Integrity Procedia 00 (2017) 000–000
ScienceDirect ScienceDirect Procedia Structural Integrity 00 4 (2017) Structural Integrity Procedia (2016)79–86 000–000
www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ESIS TC24 Workshop "Integrity of Railway Structures", 24-25 October 2016, Leoben, Austria ESIS TC24 Workshop "Integrity of Railway Structures", 24-25 October 2016, Leoben, Austria
Investigations to introduce the probability of detection method for Investigations to introduce the probability of detection method for ultrasonic inspection ofPCF hollow axlesFebruary at Deutsche XV Portuguese Conference on Fracture, 2016, 10-12 2016, PaçoBahn de Arcos, Portugal ultrasonic inspection of hollow axles at Deutsche Bahn a b Mato Pavlovicaa*, Andreas Zoëgabbof , Christina Zanotelli H. Kurz a Jochen Thermo-mechanical modeling a high pressure turbine blade Mato Pavlovic *, Andreas Zoëga , Christina Zanotelli Jochen H. Kurzb of an Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205 Berlin, Germany airplane gas 74,turbine engine Germany Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205 Berlin, Germany DB Systemtechnik GmbH, Bahntechnikerring 14774 Brandenburg-Kirchmöser, a a
b b
DB Systemtechnik GmbH, Bahntechnikerring 74, 14774 Brandenburg-Kirchmöser, Germany
P. Brandãoa, V. Infanteb, A.M. Deusc*
AbstractaDepartment of Mechanical Engineering, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1, 1049-001 Lisboa, Portugal Abstract b IDMEC, Department of Mechanical Engineering, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1, 1049-001 Lisboa, The vast experience with the automated, ultrasonic system for the inspection of hollow railway axles used by Deutsche Bahn The withflaws the automated, ultrasonic system This forPortugal the inspection of hollow railway axles calls used lead by Deutsche Bahn c vast shows thatexperience much smaller are detectable thanInstituto required. inUniversidade a number ofdefalse calls. unnecessary CeFEMA, Department of Mechanical Engineering, Superiorresults Técnico, Lisboa, Av.False Rovisco Pais, 1, to 1049-001 Lisboa, shows that much smaller flaws are detectable than required. This results in additional a number of falseIncalls. callsthe leadsensitivity to unnecessary demounting and disassembling of wheelsets, which generates unnecessary costs. orderFalse to adjust of the Portugal demounting and disassembling wheelsets, which generates additional costs. In orderof to the adjust the sensitivity of the inspection system to reduce theofnumber of false calls withoutunnecessary compromising safety, the capability system to detect cracks inspection to reduce the number ofThis false calls without the capability of the system to curves detect for cracks needs to besystem comprehensively established. capability can becompromising quantified bysafety, using probability of detection (POD) the needs to The be comprehensively capability can be by using probability of detection (POD) curves for the Abstract system. multi-parameter established. POD modelThis makes it possible to quantified include several factors that influence the crack detection in the system. The POD makes it possible to include factorsdepth that influence in the analysis. Themulti-parameter analysis presented in model this paper shows that crack position,several orientation, extension,the andcrack shapedetection as well as analysis. analysis presented in this paper that crack position, orientation, depth work, extension, and shape as POD well as the During The their operation, modern aircraft components are amplitude. subjected to demanding operating conditions, geometry of the axle all have influence on engine theshows ultrasonic response Forincreasingly future calculation of the using especially turbine blades. conditions these For partsfuture to undergo types geometry of the the high axle pressure all have influence on the ultrasonic response cause amplitude. work,different calculation of of thetime-dependent POD using multi-parameter POD model with these(HPT) parameters isSuch planned. degradation, one of model which is creep. model using finite element method (FEM) was developed, in order to be able to predict multi-parameter POD with theseAparameters is the planned. the creep behaviour of HPT blades. Flight data records (FDR) for a specific aircraft, provided by a commercial aviation © 2017 The Authors. Published by Elsevier B.V. Copyright © 2017. The Authors. Published by Elsevier B.V. company, were used to obtain thermal and mechanical data for three different flight cycles. In order to create the 3D model © 2017 The Authors. Published by Elsevier B.V. Peer-review underresponsibility responsibility of Scientific the Scientific Committee ofTC24. ESIS TC24. Peer-review under of the Committee of ESIS needed for the FEM analysis, HPT blade scrap was scanned, and its chemical composition and material properties were Peer-review under responsibility ofa the Scientific Committee of ESIS TC24. obtained. The data that was gathered was fed into the FEM model and different simulations were run, first with a simplified 3D Keywords: Non-destructive testing; NDT; Ultrasonic; Cracks; Reliability; Probability of Detection; POD rectangular block shape, in order to Ultrasonic; better establish model, and then with the real 3D mesh obtained from the blade scrap. The Keywords: Non-destructive testing; NDT; Cracks;the Reliability; Probability of Detection; POD overall expected behaviour in terms of displacement was observed, in particular at the trailing edge of the blade. Therefore such a model can be useful in the goal of predicting turbine blade life, given a set of FDR data.
1. Introduction 1. ©Introduction 2016 The Authors. Published by Elsevier B.V. Deutsche Bahn mechanized inspectionofsystems to inspect hollow railway axles. Currently, about Peer-review under uses responsibility of theultrasonic Scientific Committee PCF 2016. usestomechanized ultrasonic inspection to inspect about 140Deutsche devices Bahn are used inspect more than 130,000 axlessystems per year. Currenthollow testingrailway results axles. show Currently, that the devices 140 devices arePressure used to inspect more than 130,000 axles per year. Simulation. Currentistesting results thatofthe devices Keywords: Turbine Creep; Finite Element Method; 3Doversensitivity Model; detect evenHigh smaller flaws thanBlade; required by the standard. This resulting in a show number false calls. detect even smaller flaws than required by the standard. This oversensitivity is resulting in a number of false calls. * Tel.: +49-30-8104-4616. * Tel.: +49-30-8104-4616. E-mail address: [email protected] E-mail address: [email protected] 2452-3216 © 2017 The Authors. Published by Elsevier B.V. 2452-3216 © 2017 Authors. Published Elsevier B.V. Peer-review underThe responsibility of theby Scientific Committee of ESIS TC24. Peer-review underauthor. responsibility the Scientific Committee of ESIS TC24. * Corresponding Tel.: +351of218419991. E-mail address: [email protected] 2452-3216 © 2016 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the Scientific Committee of PCF 2016. 2452-3216 Copyright 2017. The Authors. Published by Elsevier B.V. Peer-review under responsibility of the Scientific Committee of ESIS TC24 10.1016/j.prostr.2017.07.002
80 2
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
The sources of false calls are various. Accumulations of dirt, press fits and flawed coatings can all generate ultrasonic indications. False calls lead to unnecessary demounting and disassembling of the wheelsets which generates unnecessary additional costs. The goal of the joint project between DB-Sytemtechnik and BAM is to determine the capability of the ultrasonic system to detect flaws. Once determined, the sensitivity of the inspection can be adjusted so that the number of false calls is reduced. Indications from non-destructive testing (NDT) systems, especially when trying to detect very small defects, are inconsistent. Therefore, in applications where a missed flaw can have catastrophic consequences, the reliability of the system has to be established. The probability of detection (POD) curve is considered to be a standard tool to quantify the reliability of NDT [1]. A signal response model [2] assumes that the response signal from the defect is linearly correlated with the flaw size with normally distributed deviations. Setting the decision threshold, the POD and a lower 95% confidence bend are calculated from this correlation. The a90/95 point determines the limit of reliable use of the NDT system. When there are more factors that influence the POD of the flaw significantly, the signal response model is not applicable any more. The multi-parameter POD model, allows several influencing factors to be included in the analysis and enables the POD to be expressed as a function of these factors [3]. 2. Automated system for the inspection of hollow axles One of the ultrasonic inspection systems for the inspection of hollow railway axles used by Deutsche Bahn is shown in Fig. 1. The system is docked to the side of the axle and the probe module is pushed into the borehole. The axles are inspected from the bore surface.
Fig. 1. Example for a mechanized ultrasonic inspection system for the inspection of hollow axles used by Deutsche Bahn
The section view of the axle with the probe module inside is illustrated in Fig. 2. The ultrasonic transducers are located in the probe module. The electro-mechanical drive moves the probe module in the axial direction and at the same time it rotates around the longitudinal axis, so that the transducers describe a helical path. The electromechanical drive also delivers coupling oil to the probe module and transfers the signals from the transducers to the control unit. Several transducers with different angles of incidence are mounted in the probe head with the aim of detecting cracks on the outer surface of the axle. The detection of the planar, surface breaking flaws that are perpendicular to the surface is mainly based on the corner reflection effect. The ultrasonic wave that hits the back surface of the inspected component under a favorable angle is reflected to the flaw and then back to the transducer. Since the outer surface of the axle has a variable diameter, the cracks which propagate perpendicular to the surface of the axle, will have a 90° angle to the axis only in the regions of the axle where the diameter is constant. The angle between the flaw and the ultrasonic wave will have an influence on the corner reflection effect and therefore also on the detectability of the flaw. The transition region between the wheel seat and the body of the axle is considered particularly susceptible to cracking. The investigation of the detectability of cracks by the ultrasonic inspection system was concentrated in this region.
Pavlovic et al. / Procedia Integrity (2017) 79–86 MatoMato Pavlovic / Structural Integrity Structural Procedia 00 (2017) 4000–000
813
Fig. 2. Section view of the axle with the probe module
3. Investigation of the influencing factors on the response amplitude To properly asses the reliability of the inspection system, several factors that influence the response amplitude, have been investigated. As a first, relatively fast and inexpensive tool for investigation, ultrasonic simulation of the inspection was used. Simulations were performed with software CIVA 2015 (11.1). The exact geometry and material of the axle, the geometry of cracks and the properties of ultrasonic transducers were defined in CIVA. The sound field was calculated and response amplitudes analyzed. In Fig. 3. the sound field generated in the axle by one of the transducers is shown.
Fig. 3. Simulation of the ultrasonic inspection of the hollow axle
3.1. Crack position The first factor whose influence has been investigated was the position of the crack in the axle. The response amplitudes from the crack, plotted against the position of the crack in the axle for a 35° transducer are plotted in Fig. 4 a). The position of the crack in the transition region from the shaft to the wheel seat is illustrated in Fig. 4 b). It can be seen that the crack of the same size will have a different response amplitude depending on its location. The highest response amplitude is at the position 485 and the smallest at the position 465.
82 4
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
a)
b)
Fig. 4. a) Maximum response amplitude as a function of the flaw position in the axle; b) Position of the crack in the transition region from the free shaft to the wheel seat
3.2. Crack orientation As a next factor, the influence of the orientation of the crack on the response amplitude was investigated. The detection of the cracks is primarily based on the corner reflection effect. If the angle between the crack and the ultrasonic wave generated by the transducer is changed, it will influence the corner effect and hence the response amplitude. In Fig. 5 a) the response amplitudes for the crack oriented perpendicular to the surface and a crack oriented perpendicular to the longitudinal axis are plotted for four transducers with different angles. It can be seen that the maximum response amplitude is higher for the cracks oriented perpendicular to the surface in all four cases. In Fig 5 b) the influence of the change of orientation on the corner effect is illustrated.
a)
b)
Fig. 5. (a) Maximum response amplitude for different orientations of the crack; (b) Influence of the crack orientation on the corner reflection effect.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
83 5
3.3. Crack depth extension The crack depth extension, considered as the size of the crack, is usually seen as the most influencing factor on the response amplitude. In Fig. 6., the diagram from Fig. 4 a) is shown again, this time with two significant points marked with dashed circles. It is normally assumed that the response amplitude will be higher for larger cracks. Here it can be seen that the smaller crack with the depth extension 1.5 mm at the position 495 will have a larger response amplitude than the larger flaw with depth extension 2.0 mm at the position 465. Obviously, it is not enough just to consider one influencing factor i.e. the size of the crack when calculating the probability of detection. Other important influencing factors also have to be considered.
Fig. 6. Maximum response amplitude for different crack depth extension.
3.4. Crack shape According to the standard for the non-destructive inspection of railway axles, the capability of the NDT system to detect cracks is determined on the saw-cut type of reflectors [4]. However, the cracks that can occur in the axles are semi-elliptical with an aspect ratio of minor and major axes of 0.8. Therefore, the influence of the shape of the crack on the response amplitude was also investigated. In Fig. 7 the amplitude for both types of reflectors is plotted against the crack depth extension. It can be seen that for the depth extensions smaller than 2 mm, the saw-cut reflector has a larger response amplitude than the semi-elliptical reflector. Hence, it is important to perform investigation with the appropriate reflector, when establishing the capability of the inspection system to detect cracks smaller than 2 mm.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
84 6
Fig. 7. Response amplitude for a saw-cut reflector (solid lines) and for a semi-elliptical notch with the aspect ratio 0.8 (dashed lines)
3.5. Geometry of the axle The geometry of the axle also influences the response amplitude. In Fig. 8. two axles with different curvature in the transition region from the free shaft to the wheel seat are shown. In Fig. 8 a) the old design of the axle is shown. In Fig. 8 b) from the fracture mechanics point of view improved design is shown. To investigate the influence of the changed design on the response amplitude of the ultrasonic inspection system the crack with the same characteristic is placed on the same location in both axles and the ultrasonic response is calculated. For the 35° inspection angle, the response amplitude in the new axle was -10 dB smaller than in the old. The inspectability of the axle for a given example has been reduced through the new design.
a)
b)
Fig. 8. (a) Ultrasonic B-scan for the old design of the axle; (b) Ultrasonic B-scan for the new design of the axle.
4. Influence of the amplitude drop on the probability of detection The change in the response amplitude will have a direct influence on the probability of detection. It will be shown here how a drop of amplitude influences the probability of detection. In Fig. 9 a) the signal distribution (solid line), the noise distribution (dashed line) and a decision threshold (red line) are shown for the case of a good signalto-noise (SNR) ratio.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
a)
85 7
b)
Fig. 9. (a) Signal and noise distribution with good signal-to-noise ratio; (b) Signal and noise distributions for bad signal-to-noise ratio.
The SNR is 8 and the decision threshold is set to 3 times noise level. Below the probability density distributions this is schematically shown as it would look like on the screen of the ultrasonic inspection device. For this case, the calculated POD is practically 100% and the probability of false calls is very small, PFC = 2%. In Fig. 9 b) the case is illustrated when the amplitude of the signal drops -10 dB. Signal-to-noise ratio is only 2.5. The calculated POD is 12% and the PFC stays the same. Obviously, the drop of amplitude of -10 dB has also drastically reduced the POD. To improve the POD the decision threshold can be decreased. This is illustrated in Fig. 10). The decision threshold is reduced to 2 times the noise level. The POD calculated for this case is 91%. However, due to the decrease of threshold, the probability of false call increases to 16%.
Fig. 10. Decreasing the decision threshold increases the POD but increases the PFC.
86 8
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
5. Conclusions and outlook The investigation showed that several factors have an influence on the response amplitude of the ultrasonic inspection system. Furthermore, the change of amplitude will have a direct influence on the probability of detection of the crack. The analysis showed that crack position, crack orientation, crack depth extension, crack shape and geometry of the axle are all influencing factors. All these factors have to be included in the reliability analysis to properly estimate the capability of the inspection system to detect cracks. After the initial investigation with the ultrasonic simulation software, a series of experiments on realistic and real cracks will be performed. Finally, by using both the experimental and the simulation results, applying the multiparameter model, the POD as a function of all influencing parameters will be calculated. This should give a comprehensive view of the capability of the inspection system to detect cracks. The sensitivity of the inspection can be adjusted so that the false alarms are minimised without reduction of the probability of detection. References [1] Rummel, W.D., 1998. Probability of Detection As a Quantitative Measure of Nondestructive Testing End-To-End Process Capabilities, Materials Evaluation, 56. [2] Berens A.P., 1988. NDE Reliability Data Analysis, ASM Metals Handbook, Volume 17, 9th Edition: Nondestructive Evaluation and Quality Control. ASM International, Materials Park, Ohio, pp. 689-701. [3] Pavlovic M, K Takahashi and Müller, C., 2012. Probability of Detection as a Function of Multiple Influencing Parameters, Insight 54(11):606-611. [4] DIN 27201-7., 2014, State of railway vehicles – Basic principles and production technology – Part 7: Non-destructive testing, Beuth.
ScienceDirect ScienceDirect
Structural Integrity 00 (2017) 000–000 Available online at www.sciencedirect.com Available online atProcedia www.sciencedirect.com Structural Integrity Procedia 00 (2017) 000–000
ScienceDirect ScienceDirect Procedia Structural Integrity 00 4 (2017) Structural Integrity Procedia (2016)79–86 000–000
www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ESIS TC24 Workshop "Integrity of Railway Structures", 24-25 October 2016, Leoben, Austria ESIS TC24 Workshop "Integrity of Railway Structures", 24-25 October 2016, Leoben, Austria
Investigations to introduce the probability of detection method for Investigations to introduce the probability of detection method for ultrasonic inspection ofPCF hollow axlesFebruary at Deutsche XV Portuguese Conference on Fracture, 2016, 10-12 2016, PaçoBahn de Arcos, Portugal ultrasonic inspection of hollow axles at Deutsche Bahn a b Mato Pavlovicaa*, Andreas Zoëgabbof , Christina Zanotelli H. Kurz a Jochen Thermo-mechanical modeling a high pressure turbine blade Mato Pavlovic *, Andreas Zoëga , Christina Zanotelli Jochen H. Kurzb of an Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205 Berlin, Germany airplane gas 74,turbine engine Germany Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205 Berlin, Germany DB Systemtechnik GmbH, Bahntechnikerring 14774 Brandenburg-Kirchmöser, a a
b b
DB Systemtechnik GmbH, Bahntechnikerring 74, 14774 Brandenburg-Kirchmöser, Germany
P. Brandãoa, V. Infanteb, A.M. Deusc*
AbstractaDepartment of Mechanical Engineering, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1, 1049-001 Lisboa, Portugal Abstract b IDMEC, Department of Mechanical Engineering, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1, 1049-001 Lisboa, The vast experience with the automated, ultrasonic system for the inspection of hollow railway axles used by Deutsche Bahn The withflaws the automated, ultrasonic system This forPortugal the inspection of hollow railway axles calls used lead by Deutsche Bahn c vast shows thatexperience much smaller are detectable thanInstituto required. inUniversidade a number ofdefalse calls. unnecessary CeFEMA, Department of Mechanical Engineering, Superiorresults Técnico, Lisboa, Av.False Rovisco Pais, 1, to 1049-001 Lisboa, shows that much smaller flaws are detectable than required. This results in additional a number of falseIncalls. callsthe leadsensitivity to unnecessary demounting and disassembling of wheelsets, which generates unnecessary costs. orderFalse to adjust of the Portugal demounting and disassembling wheelsets, which generates additional costs. In orderof to the adjust the sensitivity of the inspection system to reduce theofnumber of false calls withoutunnecessary compromising safety, the capability system to detect cracks inspection to reduce the number ofThis false calls without the capability of the system to curves detect for cracks needs to besystem comprehensively established. capability can becompromising quantified bysafety, using probability of detection (POD) the needs to The be comprehensively capability can be by using probability of detection (POD) curves for the Abstract system. multi-parameter established. POD modelThis makes it possible to quantified include several factors that influence the crack detection in the system. The POD makes it possible to include factorsdepth that influence in the analysis. Themulti-parameter analysis presented in model this paper shows that crack position,several orientation, extension,the andcrack shapedetection as well as analysis. analysis presented in this paper that crack position, orientation, depth work, extension, and shape as POD well as the During The their operation, modern aircraft components are amplitude. subjected to demanding operating conditions, geometry of the axle all have influence on engine theshows ultrasonic response Forincreasingly future calculation of the using especially turbine blades. conditions these For partsfuture to undergo types geometry of the the high axle pressure all have influence on the ultrasonic response cause amplitude. work,different calculation of of thetime-dependent POD using multi-parameter POD model with these(HPT) parameters isSuch planned. degradation, one of model which is creep. model using finite element method (FEM) was developed, in order to be able to predict multi-parameter POD with theseAparameters is the planned. the creep behaviour of HPT blades. Flight data records (FDR) for a specific aircraft, provided by a commercial aviation © 2017 The Authors. Published by Elsevier B.V. Copyright © 2017. The Authors. Published by Elsevier B.V. company, were used to obtain thermal and mechanical data for three different flight cycles. In order to create the 3D model © 2017 The Authors. Published by Elsevier B.V. Peer-review underresponsibility responsibility of Scientific the Scientific Committee ofTC24. ESIS TC24. Peer-review under of the Committee of ESIS needed for the FEM analysis, HPT blade scrap was scanned, and its chemical composition and material properties were Peer-review under responsibility ofa the Scientific Committee of ESIS TC24. obtained. The data that was gathered was fed into the FEM model and different simulations were run, first with a simplified 3D Keywords: Non-destructive testing; NDT; Ultrasonic; Cracks; Reliability; Probability of Detection; POD rectangular block shape, in order to Ultrasonic; better establish model, and then with the real 3D mesh obtained from the blade scrap. The Keywords: Non-destructive testing; NDT; Cracks;the Reliability; Probability of Detection; POD overall expected behaviour in terms of displacement was observed, in particular at the trailing edge of the blade. Therefore such a model can be useful in the goal of predicting turbine blade life, given a set of FDR data.
1. Introduction 1. ©Introduction 2016 The Authors. Published by Elsevier B.V. Deutsche Bahn mechanized inspectionofsystems to inspect hollow railway axles. Currently, about Peer-review under uses responsibility of theultrasonic Scientific Committee PCF 2016. usestomechanized ultrasonic inspection to inspect about 140Deutsche devices Bahn are used inspect more than 130,000 axlessystems per year. Currenthollow testingrailway results axles. show Currently, that the devices 140 devices arePressure used to inspect more than 130,000 axles per year. Simulation. Currentistesting results thatofthe devices Keywords: Turbine Creep; Finite Element Method; 3Doversensitivity Model; detect evenHigh smaller flaws thanBlade; required by the standard. This resulting in a show number false calls. detect even smaller flaws than required by the standard. This oversensitivity is resulting in a number of false calls. * Tel.: +49-30-8104-4616. * Tel.: +49-30-8104-4616. E-mail address: [email protected] E-mail address: [email protected] 2452-3216 © 2017 The Authors. Published by Elsevier B.V. 2452-3216 © 2017 Authors. Published Elsevier B.V. Peer-review underThe responsibility of theby Scientific Committee of ESIS TC24. Peer-review underauthor. responsibility the Scientific Committee of ESIS TC24. * Corresponding Tel.: +351of218419991. E-mail address: [email protected] 2452-3216 © 2016 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the Scientific Committee of PCF 2016. 2452-3216 Copyright 2017. The Authors. Published by Elsevier B.V. Peer-review under responsibility of the Scientific Committee of ESIS TC24 10.1016/j.prostr.2017.07.002
80 2
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
The sources of false calls are various. Accumulations of dirt, press fits and flawed coatings can all generate ultrasonic indications. False calls lead to unnecessary demounting and disassembling of the wheelsets which generates unnecessary additional costs. The goal of the joint project between DB-Sytemtechnik and BAM is to determine the capability of the ultrasonic system to detect flaws. Once determined, the sensitivity of the inspection can be adjusted so that the number of false calls is reduced. Indications from non-destructive testing (NDT) systems, especially when trying to detect very small defects, are inconsistent. Therefore, in applications where a missed flaw can have catastrophic consequences, the reliability of the system has to be established. The probability of detection (POD) curve is considered to be a standard tool to quantify the reliability of NDT [1]. A signal response model [2] assumes that the response signal from the defect is linearly correlated with the flaw size with normally distributed deviations. Setting the decision threshold, the POD and a lower 95% confidence bend are calculated from this correlation. The a90/95 point determines the limit of reliable use of the NDT system. When there are more factors that influence the POD of the flaw significantly, the signal response model is not applicable any more. The multi-parameter POD model, allows several influencing factors to be included in the analysis and enables the POD to be expressed as a function of these factors [3]. 2. Automated system for the inspection of hollow axles One of the ultrasonic inspection systems for the inspection of hollow railway axles used by Deutsche Bahn is shown in Fig. 1. The system is docked to the side of the axle and the probe module is pushed into the borehole. The axles are inspected from the bore surface.
Fig. 1. Example for a mechanized ultrasonic inspection system for the inspection of hollow axles used by Deutsche Bahn
The section view of the axle with the probe module inside is illustrated in Fig. 2. The ultrasonic transducers are located in the probe module. The electro-mechanical drive moves the probe module in the axial direction and at the same time it rotates around the longitudinal axis, so that the transducers describe a helical path. The electromechanical drive also delivers coupling oil to the probe module and transfers the signals from the transducers to the control unit. Several transducers with different angles of incidence are mounted in the probe head with the aim of detecting cracks on the outer surface of the axle. The detection of the planar, surface breaking flaws that are perpendicular to the surface is mainly based on the corner reflection effect. The ultrasonic wave that hits the back surface of the inspected component under a favorable angle is reflected to the flaw and then back to the transducer. Since the outer surface of the axle has a variable diameter, the cracks which propagate perpendicular to the surface of the axle, will have a 90° angle to the axis only in the regions of the axle where the diameter is constant. The angle between the flaw and the ultrasonic wave will have an influence on the corner reflection effect and therefore also on the detectability of the flaw. The transition region between the wheel seat and the body of the axle is considered particularly susceptible to cracking. The investigation of the detectability of cracks by the ultrasonic inspection system was concentrated in this region.
Pavlovic et al. / Procedia Integrity (2017) 79–86 MatoMato Pavlovic / Structural Integrity Structural Procedia 00 (2017) 4000–000
813
Fig. 2. Section view of the axle with the probe module
3. Investigation of the influencing factors on the response amplitude To properly asses the reliability of the inspection system, several factors that influence the response amplitude, have been investigated. As a first, relatively fast and inexpensive tool for investigation, ultrasonic simulation of the inspection was used. Simulations were performed with software CIVA 2015 (11.1). The exact geometry and material of the axle, the geometry of cracks and the properties of ultrasonic transducers were defined in CIVA. The sound field was calculated and response amplitudes analyzed. In Fig. 3. the sound field generated in the axle by one of the transducers is shown.
Fig. 3. Simulation of the ultrasonic inspection of the hollow axle
3.1. Crack position The first factor whose influence has been investigated was the position of the crack in the axle. The response amplitudes from the crack, plotted against the position of the crack in the axle for a 35° transducer are plotted in Fig. 4 a). The position of the crack in the transition region from the shaft to the wheel seat is illustrated in Fig. 4 b). It can be seen that the crack of the same size will have a different response amplitude depending on its location. The highest response amplitude is at the position 485 and the smallest at the position 465.
82 4
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
a)
b)
Fig. 4. a) Maximum response amplitude as a function of the flaw position in the axle; b) Position of the crack in the transition region from the free shaft to the wheel seat
3.2. Crack orientation As a next factor, the influence of the orientation of the crack on the response amplitude was investigated. The detection of the cracks is primarily based on the corner reflection effect. If the angle between the crack and the ultrasonic wave generated by the transducer is changed, it will influence the corner effect and hence the response amplitude. In Fig. 5 a) the response amplitudes for the crack oriented perpendicular to the surface and a crack oriented perpendicular to the longitudinal axis are plotted for four transducers with different angles. It can be seen that the maximum response amplitude is higher for the cracks oriented perpendicular to the surface in all four cases. In Fig 5 b) the influence of the change of orientation on the corner effect is illustrated.
a)
b)
Fig. 5. (a) Maximum response amplitude for different orientations of the crack; (b) Influence of the crack orientation on the corner reflection effect.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
83 5
3.3. Crack depth extension The crack depth extension, considered as the size of the crack, is usually seen as the most influencing factor on the response amplitude. In Fig. 6., the diagram from Fig. 4 a) is shown again, this time with two significant points marked with dashed circles. It is normally assumed that the response amplitude will be higher for larger cracks. Here it can be seen that the smaller crack with the depth extension 1.5 mm at the position 495 will have a larger response amplitude than the larger flaw with depth extension 2.0 mm at the position 465. Obviously, it is not enough just to consider one influencing factor i.e. the size of the crack when calculating the probability of detection. Other important influencing factors also have to be considered.
Fig. 6. Maximum response amplitude for different crack depth extension.
3.4. Crack shape According to the standard for the non-destructive inspection of railway axles, the capability of the NDT system to detect cracks is determined on the saw-cut type of reflectors [4]. However, the cracks that can occur in the axles are semi-elliptical with an aspect ratio of minor and major axes of 0.8. Therefore, the influence of the shape of the crack on the response amplitude was also investigated. In Fig. 7 the amplitude for both types of reflectors is plotted against the crack depth extension. It can be seen that for the depth extensions smaller than 2 mm, the saw-cut reflector has a larger response amplitude than the semi-elliptical reflector. Hence, it is important to perform investigation with the appropriate reflector, when establishing the capability of the inspection system to detect cracks smaller than 2 mm.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
84 6
Fig. 7. Response amplitude for a saw-cut reflector (solid lines) and for a semi-elliptical notch with the aspect ratio 0.8 (dashed lines)
3.5. Geometry of the axle The geometry of the axle also influences the response amplitude. In Fig. 8. two axles with different curvature in the transition region from the free shaft to the wheel seat are shown. In Fig. 8 a) the old design of the axle is shown. In Fig. 8 b) from the fracture mechanics point of view improved design is shown. To investigate the influence of the changed design on the response amplitude of the ultrasonic inspection system the crack with the same characteristic is placed on the same location in both axles and the ultrasonic response is calculated. For the 35° inspection angle, the response amplitude in the new axle was -10 dB smaller than in the old. The inspectability of the axle for a given example has been reduced through the new design.
a)
b)
Fig. 8. (a) Ultrasonic B-scan for the old design of the axle; (b) Ultrasonic B-scan for the new design of the axle.
4. Influence of the amplitude drop on the probability of detection The change in the response amplitude will have a direct influence on the probability of detection. It will be shown here how a drop of amplitude influences the probability of detection. In Fig. 9 a) the signal distribution (solid line), the noise distribution (dashed line) and a decision threshold (red line) are shown for the case of a good signalto-noise (SNR) ratio.
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
a)
85 7
b)
Fig. 9. (a) Signal and noise distribution with good signal-to-noise ratio; (b) Signal and noise distributions for bad signal-to-noise ratio.
The SNR is 8 and the decision threshold is set to 3 times noise level. Below the probability density distributions this is schematically shown as it would look like on the screen of the ultrasonic inspection device. For this case, the calculated POD is practically 100% and the probability of false calls is very small, PFC = 2%. In Fig. 9 b) the case is illustrated when the amplitude of the signal drops -10 dB. Signal-to-noise ratio is only 2.5. The calculated POD is 12% and the PFC stays the same. Obviously, the drop of amplitude of -10 dB has also drastically reduced the POD. To improve the POD the decision threshold can be decreased. This is illustrated in Fig. 10). The decision threshold is reduced to 2 times the noise level. The POD calculated for this case is 91%. However, due to the decrease of threshold, the probability of false call increases to 16%.
Fig. 10. Decreasing the decision threshold increases the POD but increases the PFC.
86 8
Mato Pavlovic et al. / Procedia Structural Integrity 4 (2017) 79–86 Mato Pavlovic / Structural Integrity Procedia 00 (2017) 000–000
5. Conclusions and outlook The investigation showed that several factors have an influence on the response amplitude of the ultrasonic inspection system. Furthermore, the change of amplitude will have a direct influence on the probability of detection of the crack. The analysis showed that crack position, crack orientation, crack depth extension, crack shape and geometry of the axle are all influencing factors. All these factors have to be included in the reliability analysis to properly estimate the capability of the inspection system to detect cracks. After the initial investigation with the ultrasonic simulation software, a series of experiments on realistic and real cracks will be performed. Finally, by using both the experimental and the simulation results, applying the multiparameter model, the POD as a function of all influencing parameters will be calculated. This should give a comprehensive view of the capability of the inspection system to detect cracks. The sensitivity of the inspection can be adjusted so that the false alarms are minimised without reduction of the probability of detection. References [1] Rummel, W.D., 1998. Probability of Detection As a Quantitative Measure of Nondestructive Testing End-To-End Process Capabilities, Materials Evaluation, 56. [2] Berens A.P., 1988. NDE Reliability Data Analysis, ASM Metals Handbook, Volume 17, 9th Edition: Nondestructive Evaluation and Quality Control. ASM International, Materials Park, Ohio, pp. 689-701. [3] Pavlovic M, K Takahashi and Müller, C., 2012. Probability of Detection as a Function of Multiple Influencing Parameters, Insight 54(11):606-611. [4] DIN 27201-7., 2014, State of railway vehicles – Basic principles and production technology – Part 7: Non-destructive testing, Beuth.