An order statistics treatment of acoustic emission signals for fatigue life prediction in real time

Abstract

In fatigue, both the crack propagation rates and the cumulative acoustic emission activity are known to be related to the applied stress intensity range. By considering the energy balance during crack propagation and the relation of strain energy release to the acoustic emission characteristics, a formal relation between acoustic emission amplitudes and initial fatigue crack propagation rates has been derived. Continuous monitoring of acoustic emission during low cycle (tension-tension) fatigue tests has been conducted on aluminum 2024-T3 and 7075-T6 alloys, until fracture. Initial crack sizes and orientations in the fatigue specimens were randomly distributed. Every few hundred cycles, the acoustic signal having the highest peak amplitude was recorded as the extreme acoustic emission event for the elapsed period. The extreme peak amplitudes, related to extreme crack propagation rates, were shown, by an Order Statistics treatment, to be extremally distributed. Statistical approach to fatigue considers that only extreme crack propagation rates are vital to fatigue lives. Knowledge of the distribution function of propagation rates is therefore essential in design for fatigue. Such knowledge can now be obtained in a non-destructive manner, during service in real time, by analyzing the distribution of amplitudes of acoustic emission signals emitted during cyclic stressing. The statistical treatment enables the prediction of the number of cycles left until failure. Predictions performed a-posteriori, based on results gained early in each fatigue test, were in good agreement with actual fatigue lives. The amplitude distribution analysis of the acoustic signals emitted during fatigue tests has been proven to be a feasible non-destructive method for predicting fatigue life of components and structures under cyclic stress.