Diagnosis of Rotor Bearing System in Noisy Conditions using Artificial Neural Networks and Genetic Algorithm

摘要

The classification of machinery faults is a very important aspect in diagnosing the machinery malfunctions. In this paper, classification of the faults from vibration data of Rotating Machinery is done by combining Genetic Algorithms (GA) and Artificial Neural Networks (ANN). ANN is combined with GA to improve fault detection even with acceptable noise levels in the vibration data. It is discovered that the noise addition to input vector of a training pattern may improve the performance of generalization. It is also discovered that selection of most significant feature set from the available large set of potential features improves the classification accuracy. In this paper Genetic Algorithm is used to optimize the derived feature. Noise is injected to the input vector with the noise optimized by Genetic Algorithm to improve the performance of generalization with noisy data. The concept of Feature Selection and Noise Injection is combined to improve the performance of the Neural Network. The performance of Neural Network is taken as the fitness function. The vibration signals were derived by synthesizing different sinusoids composed of possible frequencies and amplitudes of sinusoids normally observed in machine vibrations and faults such as unbalance, misalignment and defects in anti friction bearings. The test results show that the proposed scheme improves the performance of diagnosis from 96.8 to 100% by feature selection concept, with the absence of noise and it also improves the performance from 85.6 to 97.6 by combining the concepts of feature selection and noise injection even with as high as 25% random noise in the input signals.