Remaining Useful Life Estimations Using Acoustic Emissions

Abstract

The potential of the Acoustic Emission (AE) method in real-time predictions of remaining useful life (RUL) has been a topic of interest for several decades now. The expectation in nondestructive and structural health monitoring applications of AE is that since it is a passive approach it has the advantage of being capable of providing at the very least trends of data that could be related to other information sources, for example mechanical loading and other monitoring devices. The challenge in actually extracting practical information related to RUL is associated often with the dual need of quantifying the progressive damage in the material/component. Hence, the issue of RUL estimations using AE has often being posed through the corresponding problem of material/structural state description. Technical issues that prevent the application of such a predictive scheme based on AE are related to the inherent characteristics of the acquired signals that typically include various forms of signal conditioning and are definitely influenced by extraneous factors, typically characterized as noise. To overcome such challenges several acquisition and post-processing schemes have been developed that include novel sensors and sensor arrangements, location algorithms that enable real time and post mortem filtering of recorded signals, as well as elaborate deterministic and statistical/probabilistic treatments of acquired AE datasets to extract reliable information on the developing damage, especially under dynamic loading conditions. Despite of this state-of-the-art, the applications in which AE has been successfully implemented for RUL estimations or predictions is rather limited or is driven by commercial applications and therefore lacks generality as well as robustness. In this context, this article describes a route for material agnostic RUL estimations based on fatigue tests in various stress amplitudes and in several classes of materials. Efforts to identify AE feature trends that are repeated among different tests and are supported by mechanical and other nondestructive information, primarily using full field optical methods, are presented. The success in computing RUL estimates is quantified by observing the 25% and 50% of the total life of the specimens tested. doi: 10.12783/SHM2015/300