Research On The Fault Diagnosis Of Diesel Engine Based On Bp Neural Network Optimized By Harmony Search Algorithm

Diesel engine is a reciprocating machine with the high power in the field of the national industry and civil engineering,while it is also the source of power for a great deal of important equipment.However,due to its complicated structures and harsh working environment which contribute to the machine failure characterized by multiple,diversity and lag,so far,how to present an efficient fault diagnosis method has been a focus of the study for scholars at home and abroad.The BP neural network has many advantages such as the strong ability of parallel information processing,high fault tolerance,self-learning and self-adaptability,while it is easy to fall into the local minimum and the slow convergence rate.By contrast,the biggest feature of the harmony search algorithm is that the global optimization ability is outstanding,and the algorithm is simple.Therefore,in this paper,the harmony search algorithm is combined with BP neural network in order to highlight their respective advantages,especially improving the defects of BP neural network,the combination where fault diagnosis will be performed more efficiently and precisely.In this paper,the diesel engine of the R6105 AZLD is regarded as the experimental object.First of all,the common fault type of the diesel engine and its setting method are obtained through analysing and summarizing a series of cases.Based on this condition,through researching the sensitivity of the experimental points,the positions of the experimental points are obtained and the experimental data of the diesel engine are collected.By comparing the results of analysing signal-to-noise ratio of each wavelet basis function,the wavelet basis function will be adopted for signal-noise reduction.Secondly,through analysing the time-domain and frequency-domain characteristics when the noise signals are reduced and the energy characteristics of the wavelet packet,it is pointed out that the data on the eigenvalue of the wavelet packet energy are more suitable for this paper.After testing the standardbenchmark test function,the shortcomings of the BP neural network the neural have been proved again.Meanwhile,as regards the fault diagnosis of the diesel engine,the experimental results on the basis of the BP neural network is used as the control group for that of the follow-up experiment using the improved BP neural network.Then,the new BP neural network is optimized by the harmony search algorithm and this approach is proved to be effective by testing the standard benchmark test function.Finally,the results show that the optimized algorithm can really improve the accuracy and efficiency of the fault diagnosis.At the same time,compared with the original BP neural network diagnosis results,it is proved again that the BP neural network optimized by the harmony search algorithm have more superiority.