Research On Automobile ABS Fault Diagnosis Based On Neural Network Ensemble

Nowadays,automobile industry and the products has experienced a rapid development,the automobile has become an important way for the urban residents to travel,thus,people pay much attention and more stringent on the safety of the car.Automobile’s Anti-Lock Braking System(ABS)as the core of automobile brake system is an important component of active safety device to ensure vehicle’s safety,so,to make the fault diagnosis more accurate have the important meaning.Since neural network has the advantages such as adaptability,good fault tolerance,parallel computing,this paper applies it to the fault diagnosis of ABS problem when the regulator and sensor gets a fault.The three layer BP neural network model is built to detect and diagnose the regulator and sensor samples.At the same time,the BP algorithm is optimized by the artificial intelligence algorithm of genetic algorithm to overcome the defects of slow convergence speed and easy to fall into local minimum.Genetic algorithm is used to optimize the initial weights and thresholds of BP neural network,then the method is applied to the regulator and sensor’s fault diagnosis of ABS.Based on the MATLAB platform,using the neural network toolbox and programming genetic neural network algorithm(GA_BP),training and simulation the regulator and sensor samples respectively,the result shows that after the genetic algorithm’s optimization,the efficiency and reliability of ABS fault diagnosis has all be improved.Finally,in view of the single neural network’s disadvantage of high complexity,small fault space and small sample size,an integrated computing method is introduced,and the method of ABS fault diagnosis based on neural network ensemble is proposed.The method can make use of different information,divide the fault space and the sample space into subspace,and construct the sub neural network to calculate the simulation separately,then ensemble the outputs together.The regulator’s samples of ABS simulation results show that the neural network ensemble can effectively improve the efficiency and accuracy of fault diagnosis,and enhance the generalization performance of the network.