Chapter 11 - Automatic statistical analysis of acoustic emission data sets

Abstract

Acoustic emission testing is a structural health monitoring technique with a wide range of applications. Several structural components in various renewable energy systems, for example. Wind turbine blades made of fibre reinforced plastics, towers, foundation, tidal turbine blades, wave energy harvesting systems, pressure vessels in concentrated solar power plants, and many others, can be monitored using acoustic emission. Acoustic emission measurements can produce datasets which contain thousands to hundreds of thousands of logged signals. Particularly for large components, such as a wind turbine blade, where several piezoelectric acoustic emission sensors are required to monitor the load bearing parts of the structure over a prolonged time period, the size of the dataset generated may be exceedingly large. Although manual analysis of large acoustic emission datasets is possible, it is not trivial, particularly when complex damage mechanics are involved. Therefore, it is desirable to use methods to automatically analyse complex acoustic emission datasets without the need for manual intervention. Effective automatic analysis can also help establish appropriate filtering methodologies which can help improve data acquisition parameters so as to minimise the logging of unwanted signals from noise sources such as friction and echoes of primary damage events. The automatic processing of large acoustic emission datasets can be based on statistical analysis using different algorithms. This chapter summarises the various algorithms that can be used for the unsupervised clustering of complex acoustic emission datasets.