Intelligent Methods for Condition Diagnosis of Plant Machinery

Abstract

In the case of condition diagnosis of the plant machinery, particularly rotating machinery, the utilization of vibration signals is effective in the detection of faults and the discrimination of fault type, because the signals carry dynamic information about the machine state. Condition diagnosis depends largely on the feature analysis of vibration signals, so it is important that the feature of the signal can be sensitively extracted at the state change of a machine (Lin et al, 2000) (Liu et al, 1999) (Matuyama, 1991) (Wang et al, 2007a). However, feature extraction for fault diagnosis is difficult because the vibration signals measured at any point of the machine often contain a strong noise. Intelligent systems such as neural networks (NN) and support vector machine (SVM) have potential applications in pattern recognition and fault diagnosis. Many studies have been carried out to investigate the use of neural networks for automatic diagnosis of machinery, and most of these methods have been proposed to deal with discrimination of fault types collectively. However, the conventional neural network cannot reflect the possibility of ambiguous diagnosis problems, and will never converge when the first layer symptom parameters have the same values in different states. Furthermore, diagnostic knowledge is ambiguous because definite relationships between symptoms and fault types cannot be easily identified. In addition, due to the complexity of plant machinery conditions, and the number of fault states to be identified is enormous, it is very hard to find one or several symptom parameters that can identify all of those faults perfectly, simultaneously. Particularly, it is difficult to judge the relationship between fault states and the symptom parameters by a theoretical approach (Pusey, 2000) (Mitoma et al, 2008) (Wang et al, 2008a). For the above reasons, in order to process the uncertain relationship between symptom parameters and machinery conditions, and improve the efficiency and accuracy of fault diagnosis at an early stage, the authors reviewed their recent researches on intelligent diagnosis methods for rotating machinery based on artificial intelligence methods and feature extraction of vibration signals. That is: the diagnosis method based on wavelet transform, rough sets and neural network; the diagnosis method based on sequential fuzzy inference; diagnosis approach by possibility theory and certainty factor model; the diagnosis method on the basis of adaptive filtering technique; feature extraction method based on