Intelligent diagnostic system design: A methodology based on Principal Component Analysis and Self-Organizing Maps

This paper presents the use of Principal Component Analysis (PCA) and Self-Organizing Maps (SOMs) for bearing faults diagnosis by vibration analysis. An acquisition system is developed to collect an exhaustive set of parameters in time and frequency domain. Indicators more confident with faults are selected by PCA, optimizing neural network pattern and acquisition device. SOMs are selected as intelligent technique for faults identification. A flexible procedure of training is proposed to better design the SOM. A logical architecture is provided for multi-fault condition management. The described methodology is successfully tested on case studies conducted by a test bench provided with healthy and faulty bearing kits. As result, an Intelligent Diagnostic System is realized and integrated in a Teleservice System in collaboration with an Automotive System Integrator provider. This work demonstrates the usefulness of the proposed methodology to design predictive maintenance systems, adapting them to the particular application and guaranteeing high performances.