Research On Mechanical Fault Diagnosis Based On Neural Network

With the improvement of mechanical equipment complexity and automation level,the importance of mechanical equipment fault diagnosis is becoming more and more significant, It is very important to choose the appropriate diagnostic method for the diagnosis. In the study of intelligent fault diagnosis technology, wavelet analysis and neural network technology is a hot spot and frontiers in the research content.Firstly, the research content and significance of fault diagnosis technology are expounded, and the main methods and steps of fault diagnosis are introduced in this paper. Through the introduction of mechanical fault vibration signal,the signal in time domain and frequency domain is analyzed, then introduces the basic features and properties of the fault diagnosis technology based on neural network, and the main types of the neural network, and analyzes the advantages of neural network as a new technology. Secondly,the article describes the RBF network in detail.and the RBF neural network and BP network for comparison. Because the prediction accuracy of RBF network is larger than that of BP neural network, the training of RBF network is obviously faster than that of BP network,it shows a greater advantage in fault diagnosis. Through the analysis and comparison of various theoretical basis, it provides a theoretical basis for the following work.Due to the wavelet transform does not have time shift invariance, for the lack of wavelet analysis, this paper proposes a multi-resolution analysis and wavelet fixed time basis analysis. Taking marine air compressor as an example,the vibration signal in the operation of air compressor is studied,the reciprocating compressor valve vibration signal was collected,and the corresponding measurement data were obtained. The experimental data obtained from the Labview platform based on wavelet fixed time base are analyzed as input samples of neural networks. Finally,the fault diagnosis of air compressor is carried out by RBF neural network.Fixed time based wavelet analysis effectively tick in addition to the compressor valve fault signal redundancy,reducing the input dimensions of the neural network,improve the convergence performance of the network,thereby the training time of the network was reduced,and it avoid the network into a local minimum. Finally,the correctness and effectiveness of fault diagnosis based on wavelet fixed time base and neural network are proved by simulation experiments.