Research On Misfire Fault Diagnosis Algorithm Of Automobile Engine

Engine misfire fault diagnosis is an important issue in on-board diagnosis(OBD)system,which has significant impacts on emissions,fuel consumption and engine damage during vehicle movement.At present,most methods of automobile engine misfire fault diagnosis have the problems of low diagnosis accuracy,poor real-time performance,difficult multi-cylinder misfire fault diagnosis,etc.Additionally,there is a lack of the research on engine misfire fault diagnosis under transient working condition.In this paper,the method of automobile engine multi-cylinder misfire fault diagnosis under different working condition is mainly studied.The main work in this paper includes:1.The classification of automobile engine misfire judgment basis and the analysis of methods about misfire fault diagnosisAccording to the study on domestic and foreign literature of automobile engine misfire fault diagnosis,misfire judgment basis is divided into two types in this paper including post-cylinder factor and in-cylinder factor.Based on the previous research,the methods of misfire fault diagnosis in recent years are systematical y summarized and the problems and key points in the process of misfire fault diagnosis is analyzed in this study.Further,engine combustion torque is chosen to be judgment basis of misfire fault diagnosis.The method of indirectly estimating engine combustion torque by designing observer is proposed in this study,which lays the foundation of later research.2.Automobile engine combustion torque estimation based on sliding mode observerGiving strong robustness and high estimated accuracy of sliding mode observer,crankshaft dynamics model with the target parameter about engine combustion torque is established in this study.The sliding mode observer is designed and its stability is proved by using Lyapunov stability theory.The simulation results show that sliding mode observer can estimate engine combustion torque well in steady-state working condition.The estimated error is-25 ~ 20 N m.However,sliding mode observer becomes invalid in transient working condition.3.Automobile engine combustion torque estimation based on improved Luenberger sliding mode observerAiming to the problem of engine combustion torque estimation failure in transient working condition,Luenberger item is introduced into sliding mode observer in this study.The method effectively estimates engine combustion torque in transient working condition.In order to weaken chattering problems existing in the system,the strategy using saturation function to replace sign function of sliding mode observer is proposed in this study.The simulation results show that improved Luenberger sliding mode observer can weaken system chattering and improve estimation accuracy of engine combustion torque.The estimated error in steady-state condition can reach-5 ~ 12 N m and the estimated error in transient condition is 15-~ 60 N m.4.Automobile engine misfire fault diagnosis based on Elman neural networkBased on estimated engine combustion torque using improved Luenberger sliding mode observer,the method of multi-cylinder misfire fault diagnosis using Elman neural network in idle condition,acceleration condition and deceleration condition respectively is presented in this study.Firstly,engine combustion torque is segmental y disposed.Secondly,the proposed data are input into Elman neural network and used to train and test.Finally,back propagation neural network,support vector machine and multi layered perceptron network are introduced into experimental simulation and comparison.The simulation results show that Elman neural network can steadily converge to the global optimal value and effectively diagnose engine multi-cylinder misfire fault diagnosis in different working condition.