Analysis Of Engine Fault Based On EEMD And SVM Joint Diagnosis

Engine has tough working condition and high failure rate as the power source of the automobile. Valve clearance fault and ignition system fault are the most common fault for engine in daily use. Valve clearance fault has influence on intake and exhaust of engine, dynamic performance, emission and other indicators; the ignition system will directly lead to engine inability to work. Vibration signal of cylinder head of engine surface can be measured conveniently; it contains a wealth of vibration, shock and noise signal information, which immediately reflect the working state of engine. From the vibration signal of engine cylinder head, aiming at two core part of fault diagnosis---feature extraction and fault condition recognition, analysis method of engine fault Based on EEMD and SVM joint diagnosis is discussed in valve clearance abnormal and ignition system abnormal conditions.The calculation process of correlation coefficient of Intrinsic Mode Function method is deduced from theory of signal analysis in this paper and effectiveness of this method is verified through signal feature vector extraction to simulation signal by this method. With the DA462 engine as the research object, engine signal acquisition and test system is set. Valve clearance, misfire fault and other 8 kinds of working condition are simulated artificially and No.1 and No.4 cylinder head signals of corresponding conditions are collected and divided into training set and testing set of samples. IMFs of each order is acquired by EEMD; feature vectors are obtained from correlation coefficients method based on No.1 and No.4 vibration signals. Fault identification models of BP neural network, the default parameters of the SVM, cross searching SVM and particle swarm optimization algorithm SVM are established with feature vectors of No.1 vibration signals regarded as fault feature vectors and testing set classification results are obtained. the above process is repeated based on No.1 and No.4 vibration signals; classification models are are established and testing set classification results are obtained.Combined classification results and classification time between two methods,the result shows that by combining acquirement of fault feature vectors by EEMD based on No.1 and No.4 vibration signals with fault analysis and recognition by SVM; this method has high classification stability, good adaptability to dynamic environment, strong adaptability for small sample and high recognition accuracy rate for engine fault artificially set.There is applied value for engine fault diagnosis research without disassembly for joint diagnosis of engine fault based on EEMD and SVM.