Most materials applications require a long lifetime under service conditions. It is therefore essential to better understand the mechanisms and kinetics of damage in order to perform a reliable lifetime estimation. The diagnostic phase consists in detecting and identifying the various damage mechanisms. These methods exploit data measured by a network of AE sensors to determine the damage state, and then prognostics strategies can predict the remaining useful lifetime (RUL) of the structure. This approach based on AE data includes several steps:
1) Identification of the acoustic signature of damage mechanisms using acoustic emission and machine learning: diagnosis of health state. This provides an approach for identifying critical damage during service, with a view to controlling component lifetime.
2) Prediction of lifetime during fatigue tests in a PHM (Prognostic Health Management) approach. This approach is based on determining the energy released and identifying critical times in the energy release during mechanical tests. Thus, beyond this characteristic point, criticality can be modelled by a power-law to evaluate the time to failure.
3) Modelling of the acoustic emission from the physical mechanism to the AE signal: towards a quantitative analysis. In addition, modelling AE signals allows to expand the training database while avoiding the high costs involved in large-scale experimental campaigns.
These approaches will be presented, highlighting their advantages and limitations. The presentation will also attempt to discuss the scientific issues that need to be resolved to improve robustness and reliability.