Increasing the Sensitivity of Ultrasonic Phased Array Wheel Set Axle Inspection by Using Signal Processing

Transcription

1 ESIS TC24 Workshop: Integrity of Railway Structures Increasing the Sensitivity of Ultrasonic Phased Array Wheel Set Axle Inspection by Using Signal Processing Thomas HECKEL 1, Rainer BOEHM 1, Wolfgang SPRUCH 2, Sebastian JACOB 2 1 Bundesanstalt für Materialforschung und -prüfung (BAM), Berlin, Germany 2 Büro für Technische Diagnostik GmbH & Co. KG BTD, Brandenburg, Germany Contact thomas.heckel@bam.de Abstract The geometry and the surface condition of the shaft influence the signal to noise ratio of ultrasonic inspection of wheel set axles significantly. Signal processing algorithms may be applied on the recorded data of in-service inspections to decrease sensitivity to geometry changes by minimizing echos generated by indications of the seats. Using signal processing methods also enable the reduction of the influence of the individual condition of the wheel set on the sensitivity of the inspection. The main challenges to overcome by signal processing are on the one hand difficult coupling conditions of the probes attached to the outer surface of the axle due to the complex geometries of the shaft. On the other hand coupling quality can be decreased by a mixture of dust, mud and grease on the shafts. Therefore signal processing algorithms applied have to be stable against deviations in geometry and as well have to compensate variations in signal amplitude caused by altering coupling conditions. Different off-line algorithms have been developed and tested against each other on a given number of measured data sets by BAM and BTD at laboratory scale. Solutions for use in the field will be presented. License: 1

2 INCREASING THE SENSITIVITY OF ULTRASONIC PHASED ARRAY WHEEL SET AXLE INSPECTION BY USING SIGNAL PROCESSING Thomas Heckel 1, Rainer Boehm 1, Wolfgang Spruch 2, Sebastian Jacob 2 1 BAM Bundesanstalt für Materialforschung und prüfung, Berlin, Germany 2 BTD Büro für Technische Diagnostik, Brandenburg, Germany ESIS TC24 Wittenberge 2017, Talk ESIS TC-24 Wittenberge 2017, Talk 1.4 1

3 In-Service Inspection of Wheelsets Boogie SBB EC Waggon SBB Cargo Wheelset Bilder:Copyrighted free use, Files moved from de.wikipedia, Kategorie:Datei:Mit OTRS-Freigabe to Commons ESIS TC-24 Wittenberge 2017, Talk 1.4 2

4 In-Service Inspection Using Angle Beam Probes Pk Pk seat for brake disc I ESIS TC-24 Wittenberge 2017, Talk 1.4 3

5 Optimization of probe design - Change coupling conditions to local immersion testing - sensor size increased to gain sensititvity - use of an acoustical lens to optimize soundfield - angular scan area extended 4

6 Change Coupling Conditions to Local Immersion Testing Technique Plexiglas plexiglas Plexiglas plexiglas Stahl steel R Stahl steel steel Wasser water R local immersion technique 5

7 Increase of Sensor Size to Gain Sensititvity plexiglas water plexiglas plexiglas water plexiglas steel steel steell R steel R echo height for point type reflector The increase of the sensor area leads to defocussing of the sound beam on curved surfaces. 6

8 Use of an Acoustical Lens to Optimize Soundfield plexiglas water R B m s a 0 a plexiglas steel R water plexiglas lens steel R R 0 R 0 steel steel echo height for point type reflector compensation of defocussing by use of acoustical lens 7

9 Acoustical Lens Performance Test Verstärkung in db Einschallwinkel in ohne mit Wasserlinse Gain of sensitivity by acoustical lens on a 2 mm saw cut is +6 db to +12 db Probe in immersion setup with a waterpath of 2 mm 8

10 Extend of Angular Scanning Area -3dB -12dB -3dB -12dB -6dB -6dB 16 elements, width of element 1.4 mm 32 elements, width of element 0.9 mm 9

11 Signal Processing Optimization of scans/images for evaluation - suppression of echos induced by geometry - suppression of noise - separation of spurious signals 10

12 Signal Processing Suppression of Echos Induced by Geometry (1) Probe: 3 MHz, 16 elements α 0 = 45, α = Re A - Scan Reflektor 6 TD - Scan rotation Selected A-scan used for reference TD-Bild original soundpath Subtracted TD-Scan 11

13 Signal Processing Suppression of Echos Induced by Geometry (2) A - scan Identification of circumferential indications by means of statistic methods and image processing Calculation of TGC curves to reduce circumferential indications rotation TD - scan soundpath 12

14 Signal Processing The New Approach rotation Filtering without modification of echo height GOAL: display of relevant indications only SOLUTION: suppression of unwanted noise position on axis Axle with arteficial flaws, software by BTD, overlay of TD-scans for angles

15 Signal Processing No Modification in Amplitude is Allowed raw data decision matrix processed image Quelle: DB Quelle: BMVI statistic based signal processing classification 14

16 Signal Processing Statistical Evaluation of Raw Data Set Formanzeigenverdacht bei AW < 1 Fehleranzeigenverdacht bei AW2 > SNR min Auswahlgröße Criterion AW Indication by flaw? Auswahlgröße AW 2 Criterion AW Indication 0.5 Indication by geometry Anteil der A-Bilder mit Echosignal in % Number of A-Scans with significant signal in % Umfangsposition Rotation of axle in % Criterion AW 1 : AW 1 < 1 Recognition of geometry caused circumferential indications Criterion AW 2 : AW 2 > SNR min Recognition of indication caused by flaws Distinguish between noise, flaws and spourious signals on form and amplitude 15

17 Signal Processing Statistic Evaluation on Test Data Set echo height echo height echo height raw data set decision matrix processed data set Based on the raw data set the algorithm computes a decision matrix. Where relevant information has been detected, the information from the raw data is copied to the processed data set. 16

18 Signal Processing on Test Wheel Set Significant Decrease of Noise Axle without arteficial flaws, overlay of TD-scans for angles

19 Signal Processing on Test Wheel Set Axle with arteficial flaws, overlay of TD-scans for angles

20 Examination Result on Axle with Arteficial Flaws 0 0 back 90 back 90 bottom 180 bottom 180 front 270 front 270 top Grad 360 top Grad

21 Conclusion - Improvements Fachbereich 8.4 Local immersion technique Increase of sensitivity by use of lager transducer Optimization of sound field by use of a lens Increase in scan area by use of smaller elemtens Identification and suppression of geometry caused indications Suppression of noise Suppression of spourios signals 20

22 Conclusion What is left to do? Fachbereich 8.4 Test algorithms in the field collect data sets from different types of axles Make thresholds adaptive to signal quality Gefördert vom im Rahmen eines MNPQ Projekts Messen, Normen, Prüfen, Qualitätssicherung 21