Reliability Analysis of LASER ULTRASONICS as an NDT method for surface defect detection in High Speed Hollow Train Axles using Experimental and Numerical Methods

The objective of the research is to find a methodology to test the train axle for possible surface defects using non-contact laser ultrasonic technique and then studying the reliability of this non-destructive test (NDT) method by considering a large set of defects with varying sizes to build Probability of Detection (POD) curve. Generally, this NDT technique has three parts: laser beam generation and impingement on a point over the test object surface, ultrasound propagation within the material of the test object and reception of the ultrasound wave propagated at a certain distance from the laser source position. The wave propagation will be influenced by the path where it propagates through and specifically it will be modified in presence of a defect. The part of methodology covered in this research is more concentrated on the laser beam incidence and ultrasound generation within the material and its behavior as it moves through the train axle under study. Laser ultrasonics is being considered a replacement of conventional ultrasonic testing method as it eliminates many limitations of conventional method and has advantages of its own. In case of train axles preventive maintenance is a challenge using conventional method as once mounted it is very difficult to reach all areas of axle and dismounting is never a good idea with respect to time and economy. Though rotating probe method needs only small elements to be removed to reach the central bore in case of hollow axles but there was still an urge to use a non-contact method which is feasible for hard to reach areas and does not require dismounting of the axle completely. Laser ultrasonics if found feasible in future not only kills this issue but it is also quicker and if used properly can detect wider range of surface defects. As a starting point it is necessary to know the behavior of ultrasonic waves in the material which is steel in our case. The steel axle under research is being used in high speed train ETR 460. It becomes harder to see the behavior of the waves when there is a hollow axle with various abrupt variations in cross section along the length. Surface acoustic waves (SAW) which are precisely Rayleigh surface waves in our case are the most important for detecting surface defects. These waves are almost half in speed to longitudinal waves and so are with smaller wavelengths for a given frequency to become more sensitive to any hurdles in their way. It also leads to many reflections in the detected signals due to central hole and fillets present on each variation of the cross section. A section of the axle with two fillets is selected initially to study all of the above behavior. 2D model of this section is numerically analyzed using COMSOL multi-physics code. A series of experiments is also carried on selected section using the lab equipment available to produce Nd: YAG (neodymium-doped yttrium aluminum garnet) laser in the infrared region with a short pulse of about 16 Megawatt (maximum power) for 12 nanoseconds duration. It theoretically causes about 5 nanometer of penetration in the steel to cause thermo-elastic regime. This value also varies on the basis of surface quality and frequency of laser beam used. Due to this phenomenon, a transient increase in temperature in a small volume of material causes a thermal stress which in turn propagates three types of waves: longitudinal waves, transverse waves and Rayleigh surface waves. The reception signal was read by using Laser Doppler Vibrometer (LDV). Another option may also be considered by using air-coupled ultrasonic transducer making the method a hybrid Laser Ultrasonic system. Behavior of the stated waves studied numerically and experimentally by measuring vertical displacement of a point at different locations by simulating defect of varying size and position is what can help in finalizing the objective. Along with this sensitivity and efficiency of the NDT method against different range of defect sizes is also important. It can be achieved by measuring POD curve of the system. This curve tells us about the largest probable size of defect which can be missed by the technique under consideration. Parametric studies run for different defect sizes give a curve against RMS of the signal and the amplitude of the Rayleigh wave. For example to study change in RMS of the vertical displacement of a point when measured after defect of a constant width and varying depth from no crack to a maximum crack depth showing a variable behavior with a small incremental step each time gives a probability measure of the detection of the defect. The Rayleigh Surface wave attenuation was one of the main changing factors considered due to the inclusion of the defects. The important information is extracted using POD Curve to look for the minimum size of defect to be seen by the method with required level of probability. It also helps to define the inspection interval of a train axle under service. First of all a range of crack sizes is to be defined in which the inspection system under question cannot determine if the object is defected or not. Then a large number of specimen are required in this range of defects to do further analysis. Linear regression analysis of these results is used to get the conventional POD curve. This type of POD curve built with inputs from numerical models is termed as Model Assisted POD curve or MAPOD curve. MAPOD curve avoids the expensive real building of test specimen for different defect sizes. A large number of trial experiments were done on the real axle without defect to incorporate natural noise of the inspection method in the numerical readings. The model to be used was refined multiple times to get a fully qualified model fulfilling all the real requirements and avoiding as much as possible the stochastic behavior of numerical methods involved. A generalized alpha time stepping algorithm was used in the time domain study of all the experiments. Direct Solver used was the famous MUMPS solver and Newton method was used for the iterative non-linear solution. At the end MAPOD curve for the specific settings used showed that the Laser-Ultrasonics is fairly suitable to detect the typical crack sizes and if a suitable configuration is found it is quite efficient to serve the purpose of in service inspection of train axles without using any conventional contact techniques