Fault diagnosis of manufacturing systems using data mining techniques

Abstract

Fault is one of the main causes of failure, and the accurate diagnosis is one of the most significant steps in fault treatment. This paper considers the diagnosis system to solve some maintenance optimization problems in manufacturing systems. The proposed architecture deals primarily with three modules, namely, the detection module, the diagnosis module, and the decision making module. In this case, the fault needs to be detected and diagnosed as early as possible after its occurrence. Data mining techniques can support repairmen in diagnosis decision-making process. To be successful, we suggest new classification approach based on hybrid neural network technique focusing this industrial application for developing a diagnosis system. Two models of neural networks: Gradient Descent and Momentum & Adaptive LR and Levenberg-Marquardt are investigated. Classifier system was used in order to construct accurate system for fault classification based on regression technique. The performance of the approach is evaluated using mean square error and classification accuracy. Case study and experimental results are given and discussed. Results achieved in this paper have potential to open new opportunities in industrial diagnosis of probable faults.