Study On Ensemble Learning Diagnosis Method For Diesel Engine Fault

Diesel engine as a driving mechanism, has a wide application in the field of vehicle, ship, power plant, engineering machinery and agriculture machinery etc.. The probability of diesel engine fault is relatively high because of its characteristics of multiple components, complex motion, poor working environment, so fault diagnosis research has very important significance in ensuring the normal operation of the equipment. With the development of the intelligent technology, fault diagnosis methods are also increasingly perfect.The paper is based on the existing methods and techniques of learning and summarizing, conduct in-depth study on diesel engine fault diagnosis method using ensemble learning theory. Ensemble learning using support vector machine(SVM) as learner is applied in fault pattern recognition of diesel engine, as compared to its application, integrated with BP neural network, genetic algorithm and single SVM conduct systematic research on pattern recognition, the key link of diesel engine fault diagnosis. According to the vibration signal feature of the normal state and six kinds of fault state of diesel engine, extraction the sub-band energy feature signal and the time domain signal, combine the respective advantages of different methods in fault pattern recognition, the thesis first focuses on the research of single support vector machine and support vector machine ensemble learning algorithm for fault diagnosis technology, also study the application of BP neural network and genetic algorithm in fault diagnosis of diesel engine. Establish the support vector machine fault diagnosis model, and optimize the best parameters with Cross-validation; establish support vector machine ensemble fault diagnosis model based on Boosting algorithm; analysis and study BP neural network and BP neural network based on genetic algorithm fault diagnosis model, conduct comparative analysis with the diagnosis model of the strong learner based on ensemble learning.The results show that the diagnosis correct rate of support vector machine ensemble fault diagnosis model is higher than that of single SVM and SVM after parameter optimization, it is significantly higher than the application results in diesel engine fault diagnosis of BP neural network and BP neural network based on genetic algorithm. The idea of establishing strong learner based on ensemble learning is successfully verified, that it is feasible of application of ensemble learning to diesel engine fault diagnosis.