Studies on quantitative acoustic emission with applications to material fatigue testing

The thesis develops a quantitative waveform-based AE technique for studying mode-I fatigue of PH13-8 stainless steel. There are two major challenges in successful application of the technique: the characterization of wave propagation, and the correlation of detected AE phenomena with the underlying fatigue mechanisms.; This work studies elastic wave propagation in both 2D and 3D plates using finite element method (FEM) and laser-ultrasonic techniques. Comparison of the results from numerical simulation and experiments confirms that FEM is capable of simulating and revealing the features of AE pulse propagation. Using the developed FEM model, the effect of rise-time of AE source and the finite sensor size on the resulting AE waveforms are studied. Furthermore, the moment tensor representation for internal crack source is incorporated with the FEM to compute the responses generated from a mode-I crack.; A series of high-cycle mode-I fatigue tests using single-edge-notched specimens are conducted via on-line AE monitoring. An experimental procedure is developed which enables the repeatability of AE test results. The acoustic activities observed during the fatigue process consist of crack AE events and larger amount of grip noise, as well as other noise events, and are found to be reproducible. The explicit crack AE activities identified in post-processing, disclose two stages of fatigue crack growth: the surface crack growth, and the through-thickness crack growth. A transition between these two stages is identified acoustically. The AE activities and the associated crack growth behavior in these two stages are found to follow different relations. An empirical relation for estimate the length of a surface crack using AE events detected is developed. The results obtained show good agreement with the actual crack size.; This study also presents an algorithm for automatic separation and clustering of AE events from the noise-mixed data using principal component analysis (PCA), a feature-based statistical signal processing technique. It is shown using this algorithm the noise content can be reduced by over 40%. Applications of the PCA technique to the frequency domain data and design of a “noise filter” for tracing the evolution of signals generated during fatigue process are also demonstrated.