Feature Extraction And Fault Diagnosis Research On The Diesel Engine

As the power machinery, diesel engine is the main equipment with itscomplicated structure. It tends to higher occurrence rate of fault functions and willaffect the running condition of the whole mechanic system, which leads to danger. Soit has been a hot topic about how to do fault diagnosis promptly and accurately inorder to prevent and avoid malfunctions. Recently, the technology about onlinemonitoring and fault diagnosis for diesel engine has been developing rapidly and ithas achieved a lot. However, the diesel engine fault diagnosis is still in the stage ofmethod study at present. Based on the study and summarizing of the existing methodsand technology basis, the technique realize the diesel fault diagnosis by vibrationsignals.In the process of the diesel fault diagnosis, feature extraction is an important stepwhich determines whether the diagnosis would success. The traditional featureextraction method contains time domain analysis, frequency domain analysis andtime-frequency analysis. Through the example contrast, a method based on theensemble empirical mode decomposition and singular value decomposition is putforward which contains the superiority in processing the non-stationary signals. Thesupport vector machine is chosen to make fault pattern recognition. It seeks for thecompromise between the specific training sample accuracy and sample recognitionability. It has been widely used in the nonlinear problem about characteristicsparameters.The technology which contains EEMD, SVD and SVM realizes the faultdiagnosis successfully. The main steps are as follows.1. The vibration signals aremeasured on the diesel under four fault states of working conditions, which includingvalve clearance fault, the injection advanced angle fault,the injection quantity faultand the injection pressure fault.2. The time domain, frequency domain andtime-frequency domain of vibration signal from diesel engine are processed in orderto extract parameters. The diesel engine state will be judged by contrasting thecharacteristics and parameters of the different fault conditions.3. Fault diagnosistechnology extracts the features, normalizes the feature data and recognizes the faultpattern so that the four fault state and normal state are identified. The technology isproved. The method which extracts fault symptom vectors from the vibration signals ondiesel engine cylinder head is simple, so it is easy to realize the fault diagnosis, evenfor diagnosis without disassemble. SVM can recognize the identification of faultdiagnosis. With the developing of its theory, method and kernel functions, SVM willgive greater promote on the technology of the diesel engine fault diagnosis.