Research On Intelligent Control Strategy Of Dual Clutch Transmissions Based On Driving Behavior Characteristics

Due to its technical characteristics,Dual Clutch Transmission(DCT)is especially suitable for the current technical foundation and manufacturing conditions in China.Therefore,many domestic auto manufacturers have carried out independent research and development of DCT.In recent years,with the rapid development of automotive intelligent network and the maturity of machine learning algorithms,it is necessary to further study the DCT intelligent control strategy considering driving behavior characteristics(driving style and starting driving intention).In order to improve the intelligent control of the starting process of the car,this paper studies the intelligent control strategy of the dual clutch automatic transmission based on the driving behavior characteristics.The main research contents are as follows:(1)Research on classification and recognition of driving style based on feature engineering.First,a road test was designed considering the influencing factors,in which the driving data is collected,and the driver driving style is subjectively evaluated.Subsequently,a discretization method based on information entropy is applied to discretize the velocity and accelerator pedal degree,where 44 feature quantities are extracted to characterize the driving style.Taking into account the strong correlation and redundancy between the constructed feature quantities,principal component analysis is used to reduce the dimension,and fuzzy C-means clustering is used to classify the driving style.Finally,a parameter optimization-based support vector machines(SVM)algorithm is proposed to identify the classified driving style.The classification and recognition method of driving style lays a foundation for the research of vehicle starting intelligent control strategy.(2)Modeling and identification modeling of starting driving intents with integrated driving style.Firstly,the influencing factors of the driver's starting intention,the difference of the starting intentions of different styles of drivers and the importance of predicting the throttle opening are analyzed.Then,due to the ability of BP neural network to reconstruct arbitrarily complex nonlinear continuous functions,it is used to predict the accelerator opening during the driver's starting process.Finally,the fuzzy C-means clustering is used to classify the predicted throttle opening and its first derivative.Cluster analysis is used to extract and formulate the fuzzy rules of the starting driving intention recognition system from the objective data level.Bsed on this,the design of the starting driving intent recognition system based on fuzzy clustering is completed,which lays a foundation for the vehicle starting intelligent control strategy to better adapt to the starting driving intention.(3)Intelligent control of DCT starting process based on fuzzy neural network.By analyzing the influencing factors of the clutch combination at the start,the starting driving intention,the clutch master-slave disc speed difference,the actual engine speed and the target speed difference are finally taken as the control variables in the clutch coupling process.Then,in order to better learn the control rules of the original simulation control strategy,combined with the advantages of neural network and fuzzy control,an adaptive fuzzy neural network starting intelligent control system is designed.The objectivity of fuzzy rule acquisition and the self-learning and self-adjustment functions of fuzzy rules and membership functions are realized.Finally,the Simulink simulation model of the starting process is established,and the proposed control strategy is simulated and verified.(4)Research on clutch combined displacement optimization in DCT starting process based on multi-objective particle swarm optimization algorithm.Firstly,the evaluation indicators in the process of vehicle start control are analyzed.Considering that the vehicle starting process is a dynamic problem,the clutch combined displacement is selected as the optimization object,and the clutch combined displacement is fitted into a polynomial function by using neural network and least squares method to obtain the polynomial parameters that need to be optimized.Then the objective constraints and starting intentions in the starting process are taken as the constraint conditions,the impact degree and the sliding work are taken as the objective functions,and the clutch displacement is used as the optimization object.Based on this,the clutch combined displacement optimization algorithm based on multi-objective particle swarm optimization algorithm for DCT starting process is designed.Thus,a set of optimal clutch engagement displacement curves in consideration of the starting intention is obtained.Finally,using fuzzy set theory,the optimal individual solution is found from the optimal solution set,and the optimization of clutch combined displacement in the starting process is realized.