Diesel Engine Fault Diagnosis Based On Morphological Filter And Grey Theory

Diesel is the most common reciprocating power machinery equipment, it was used aspower source in automobiles, ships, locomotives and aircraft equipment, so it occupies a veryimportant position in the entire mechanical system. However, due to the complex structure ofthe diesel engine and the harsh conditions caused a failure of its multiple and diverse, how totimely detective and rapid, effective diagnosis and troubleshooting are the main contents ofscholars at home and abroad.This paper presents a method of diesel engine fault diagnosis based on morphologicalfiltering and gray theory. First, summaried the rules of using morphological filteringtechnology structure elements, then designed an adaptive generalized morphological filter toreduce noise in diesel engine vibration signal, adaptively adjust the weights by gradientmethod allows optimal noise reduction. From the time-domain and power spectrum of thecylinder blasting shock features six kinds of conditions were studied separately, illustrates theinherent characteristics of the diesel engine vibration, and realize to diagnosis thesingle-cylinder off oil failure and double-cylinder oil off failure at first. After denoised thevibration signal by the morphological filter, according to the time and frequency domain needto extract the eigenvalues and wavelet packet energy value as a feature vector, respectively,through the Tangs correlation and improved correlation analysis to calculate the correlationbetween detection feature vectors and standard mode vectors. The results show that theimproved correlation analysis have improved the accuracy of fault diagnosis.This paper integrated of the gray system theory and neural network model, and usingMATLAB software based on neural network toolbox custom to designe a gray neural networkmodel to achieve complementary advantages and disadvantages of both. Gray neural network model designed in this paper does not require network training, so as to avoid the existence oftraining in network training in a long time, network convergence and other issues, it’s the coreidea is the use of parallel processing computing ability of neural network to achieve grayrelational computing degrees. Fault diagnosis based on gray neural network model, with asimple calculation, strong parallel processing capabilities, high diagnostic accuracy and so on.The results show that the gray neural network model can achieve the desired diagnosticcapabilities.