Research On Fault Diagnosis Of Diesel Engine Based On Acoustic Signal

Diesel engine as an important power machinery,its state performance will directly affect smooth progress of the task.Fault diagnosis technology is an important guarantee for safe and stable operation of mechanical equipment.It is very important to study new method and new technology in the field of diesel engine fault diagnosis.The mechanical fault diagnosis method based on acoustic signal analysis has the advantage of non-contact,without disassembly,high efficiency and convenience.It is very suitable for the diagnosis of mechanical equipment in harsh environment and has been widely used.In this paper,diesel engine as research object,aiming at fault diagnosis,taking the method based on acoustic signal analysis,mainly to study signal denoising technology,feature extraction technology and pattern recognition technology of diesel engine acoustic signal.Through the construction of acoustic signal acquisition experimental platform of 6135D type diesel engine,get signal samples of different fault types,design corresponding diagnostic methods respectively,and the proposed method is experimentally verified one by one.The main work of this paper includes following aspects:1.Aiming at non-stationary and non-linear characteristic of diesel engine acoustic signal,a denoising method based on improved wavelet threshold and empirical mode decomposition is designed.This method utilizes the advantage of wavelet threshold denoising and empirical mode decomposition denoising.The correlation coefficient method is used to find the demarcation point of signal-dominated and noise-dominated intrinsic mode function components,and the high frequency noise dominated intrinsic mode functions are denoised by improved wavelet threshold function,then reconstruct signal,thereby get the signal after denoising.The experimental result of simulation signal and measured signal shows that the proposed method has better denoising performance.2.Variational mode decomposition is introduced into fault diagnosis,a typical fault diagnosis method of diesel engine combined with variational mode decomposition and laplacian score is studied.Firstly,an acoustic signal is broken down by variational mode decomposition,and characteristic parameters are extracted from various mode functions obtained from the decomposition.Then features are sorted by improved laplacian score algorithm.Finally,support vector machine is used to diagnose fault,and determine the optimal dimension.Experiment showes that this method can effectively identify typical fault of diesel engine,with a higher diagnostic accuracy.3.In order to solve the problem that compound fault is difficult to be identified in fault diagnosis of diesel engine,a fault diagnosis method based on variational mode decomposition,optimized neighborhood rough set and community hierarchical clustering is designed.Considering the influence of redundant features,we use optimized neighborhood rough set to select features in order to achieve the purpose of attribute reduction.The fault diagnosis network is established by using community structure in complex network.Community structure is found by designing community criterion function,and fault diagnosis classification is realized at the same time.The validity and superiority of the proposed method is fully confirmed in the experimental result.