Reliability analysis of laser ultrasonics for train axle diagnostics based on model assisted POD curves

Abstract

High speed train axles are integrated for a lifetime and it is time and resource consuming to conduct in service inspection with high accuracy. Laser ultrasonics is a proposed solution as a subset of non-contact measuring methods effective also for hard to reach areas and even recently proved to be effective using Laser Doppler Vibrometer (LDV) or air-coupled probes in reception. A reliability analysis of laser ultrasonics for this specific application is here performed. The research is mainly based on numerical study of the effect of high energy laser pulses on the surface of a steel axle and of the behavior of the ultrasonic waves in detecting possible defects. Probability of Detection (POD) concept is used as an estimated reliability of the inspection method. In particular Model Assisted Probability of Detection (MAPOD), a modified form of POD where models are used to infer results for making a decisive statistical approach of POD curve, is here adopted. This paper implements this approach by taking the inputs from limited experiments conducted on a high speed train axle using laser ultrasonics (source pulsed Nd:Yag, reception by high-frequency LDV) to calibrate a multiphysics FE model and by using the calibrated model to generate data samples statistically representative of damaged train axles. The simulated flaws are in accordance with the real defects present on the axle. A set of flaws of different depth has been modeled in order to assess the laser ultrasonics POD for this specific application.