Combination Diagnosis Based on Genetic Algorithm for Rotating Machinery

Abstract

This paper presents the combination diagnosis method based on genetic algorithm for rotating machinery according to the limitation that any single fault feature or any single diagnosis method can not achieve the accurate diagnosis result in whole diagnosis state space. This method can effectively use all kinds of different characteristic fault features and diagnosis methods and then bring into play their advantage, so that the accurate rate is improved. This paper combines neural network diagnosis method with artificial immune diagnosis method using genetic algorithm according to different features. Then each diagnosis method displays its advantage in its optimal space. Wavelet Packet "energy" feature and Bispectrum feature are used for training two diagnosis methods. Genetic algorithm is adopted to optimize diagnosis combination weight matrix. The instance diagnosis result of rotating machinery shows that this combination diagnosis method can effectively improve the accurate rate of fault diagnosis and diagnosis system robustness. Moreover, this method can be applied in fault diagnosis for other machinery.