A Componential Coding Neural Network based Signal Modelling for Condition Monitoring

Many condition monitoring (CM) techniques have been investigated for early fault detection and diagnosis in order to avoid unexpected breakdowns due to machinery failures. However, manual techniques require well-skilled labours which will increase the cost of the monitoring process and may not always be available at the site. One of the most promising approaches is to automate the monitoring process using artificial intelligence (AI) techniques. However, the majority of AI-based techniques have been developed in CM for the post-processing stage, whereas the critical tasks including feature extraction and selection are still manually processed. This study focuses on the extending AI techniques in all phases of CM process by using a Componential Coding Neural Network (CCNN) which has been found to have unique properties of being trained through unsupervised learning, capable of dealing with raw data sets, translation invariance and high computational efficiency. These advantages of CCNN make it particularly suitable for automated analysis of the vibration data arisen from typical machine components such as the rolling element bearings which exhibit periodic phenomena with high non-stationarity and strong noise contamination. The CCNN was evaluated using both simulated and experimental data collected from a healthy and two defective tapered roller bearings under different operating conditions. Both of the results showed the capability of CCNN in detecting the initial anomalies of roller element bearings.