Fault Diagnosis for a Rolling Bearing Used in a Reciprocating Machine by Adaptive Filtering Technique and Fuzzy Neural Network

This paper presents a method of fault diagnosis for a rolling bearing used in a reciprocating machine by the adaptive filtering technique and a fuzzy neural network. The adaptive filtering is used for noise cancelling and feature extraction from vibration signal measured for the diagnosis. A fuzzy neural network is used to automatically distinguish the fault types of a bearing by time domain features. Using the signals processed by adaptive filtering, the neural network can quickly converge when learning, and can quickly distinguish fault types when diagnosing. The spectrum analysis of an enveloped time signal is also used for the fault diagnosis. Practical examples of diagnosis for a rice husking machine are shown in order to verify the efficiency of the method. All diagnosis results of the spectrum analysis and the fuzzy neural network show that the method proposed in this paper is very effective even for cancelling highly correlated noise, and for automatically discriminating the fault types with a high accuracy. Key-Words: fault diagnosis, adaptive filtering, fuzzy neural network, rolling bearing, Reciprocating Machine, signal processing