Research On Diesel Engine Fault Diagnosis Based On Support Vector Machine Optimized By Artificial Bee Colony Algorithm

With the development of modern industry, diesel engine as the power device has been widely applied to social production for various industries, such as transportation, mining machinery and other fields, and the working position becomes increasingly important. However, due to the complexity of the equipment structure, mutual influence in different parts, and poor working environment and complex working conditions, the failure rate is high, and once the fault occurs, the impact will be incalculable. Therefore, in order to avoid unnecessary losses and ensure the normal operation of the system, the real-time state monitoring and fault diagnosis of diesel engine are necessary to be able to take effective measures timely for the abnormal situations.Through the clarification and reflection on advantages and disadvantages of diesel engine fault diagnosis methods and in view of the current experimental conditions, this paper chooses vibration signal analysis method for fault diagnosis of R6105 AZLD diesel engine, and sums up the whole research process into three parts: signal acquisition experiment, signal processing, fault identification, the specific work of each part is as follows:(1)Signal acquisition experimentAccording to the structural arrangement and working principle of diesel engine, explore rational measuring points in order to be able to obtain the vibration signal data containing the enrich work information of diesel engine and analyze the main failure form and the incidence of diesel engine faults to set faults purposefully to increase the availability and fully meet the requirements of sampling frequency for signal collecting so as to reduce the complexity of calculation under the premise of complete signal information.(2)Signal processingThe diesel engine vibration signal not only has non-stationary, and because of the complexity of input and output path and many sources of the vibration, the signal information collected is mixed seriously. For the sake of selecting more suitable signal data of measurement point, effectively purifying the working condition information and reducing distortion to the most extent, in the paper, based on wavelet threshold denoising methods, it makes the signal to noise ratio, the root mean square error and the smooth degree as the evaluation index to seek the best measurement point and more suitable denoising rule. Then, in order to further clarify vibration signal feature information of diesel engine under different conditions and ensure that the information useful, the paper applies wavelet packet analysis method to extract the characteristic values, and to reduce the final workload for classification, the larger parts of the relative energy in the feature values are selected as the diagnostic sample data.(3)Fault identificationTo solve the problem that the support vector machine(SVM) can effectively solve the small sample classification in fault identification but the selection of key parameters will directly influence the quality of the final classification performance and that basic artificial bee colony algorithm is less controllable parameters,easy to understand and implement but easily falls into the local optimal solution and slow convergence speed in the later stage, the paper finally chooses the research method which takes advantage of the artificial bee colony algorithm which is improved in search method to optimize the parameters of support vector machine to identify fault, and through the comparison classification accuracy with support vector machine without optimization and support vector machine optimized by the basic artificial bee colony algorithm, it further shows that the improved method is more practical.