The aim of the study is to develop an automatic bearing fault diagnostic method using artificial intelligence (AI) techniques based on multiple sensor signals.
Traditionally, the on-line signals from sensors are plotted (mean of RMS vs time) to monitor the trend and detect bearing failures. This has yielded a degree of success, but it would be better to have an intelligent system that integrates the fault detection, diagnosis and reasoning process and enables automatic condition monitoring for bearings. This research, which began October 2005, has been exploring the use of statistical based AI intelligence techniques including clustering, Gaussian Mixture Model (GMM), T-squared statistics, and log-likelihood scores to extract further information from the signals that help in bearing fault detection, failure prediction and fault diagnosis. So far, significant success has been achieved with experimental data mainly based on the software developed by GE Aviation. For example, adapted GMM is used to deal with the outliers within the signals; a Principle Component Analysis (PCA) based method is used to locate the faulty bearing components; a Bayesian Belief Network (BBN) is applied as an inference platform for reasoning and cause rooting.
Conferences and events associated with this project:
