Research On Intelligent Shifting Of DCT Based On Data Mining Method

Shifting of Dual clutch transmission(DCT)has always been a focus of research,which includes two parts: Gear decision-making and shifting process control.Gear decision-making of Dual clutch transmission proves the key factor to affect the economy and power performance as well as driving experience of automobile.Driving intentions and driving environment(e.g.,road slope)should be taken into account when determining the optimum gear.The torque coordination control of the two cluches is closely related to the impact slip friction work and shifting time,which plays a decisive role in the smoothness and dynamic response of the vehicle.Now the recognition of most driving intentions and driving environment take use of qualitative judgment,fuzzy reasoning membership functions and fuzzy rules seem very dependent on experience,and the identification accuracy is not high enough.Traditional gear decision-making method of DCT is mostly limited to the extraction and correction of shifting schedules,which failed to combine with the big data generated in the process of vehicle driving.It is difficult to obtain the optimal target torque trajectory of the clutches during the actual shifting process.In this paper,a 7-speed DCT provided by an automobile manufacturer is taken as the object to carry out the data-based intelligent gear decision-making research and intelligent control of shifting process.The main research contents are as follows:(1)Driving intention recognition based on data mining method.According to the collected vehicle driving data,the data of brake force,vehicle speed and throttle pedal opening after denoising are obtained by using wavelet analysis method.The dimensionality reduction features retaining most of the original information are obtained by principal factor analysis method.The driving intentions such as maintenance,parking,acceleration and deceleration are clustered based on bisecting K-means algorithm.The data including five types of driving intentions are used to establish the MGHMM model for real-time identification of driving intentions.(2)Gear optimization based on dynamic programming considering fuel economy.BP neural network optimized by genetic algorithm is used to predict the vehicle speed of the future 10 s under the driving condition NEDC UDDS and US06.Combined the results of vehicle speed prediction with dynamic programming algorithm,the DP gear decision-making based on vehicle speed prediction becomes realized.The real-time performance,fuel consumption and shifting times between the DP gear decision-making based on vehicle speed prediction and the DP gear decision –making based on the known global working condition is completed.Taking use of the analytic method,the optimal economic and dynamic shift schedules are extracted,meanwhile the shifting model of vehicle is also established,the economic and dynamic performance as well as shifting times of four gear decision-making methods are completely analyzed.(3)Intelligent gear decision-making of data-driven under different driving intentions and slopes.The separation boundary of each adjacent gear area was found by SVM algorithm and therefore the upshifting and downshifting lines of all gears were extracted.Based on the driving intentions identified by data mining method,the particle swarm optimization(PSO)algorithm is used to realize the optimization of gear sequence under different driving intentions,taking the dynamic and economic performance and shift frequency of the whole working condition as optimization objectives.The mean value and standard deviation of slope,total length of the driving condition as well as the length of the single slope are taken as the characteristic parameters to construct the slope curve which changes with time.Finally,making use of the particle swarm algorithm to optimize the gear sequence in slope conditions considering the driving intentions,with the mapping relationship between intention,slope,gear at the current moment,speed and the opening of the accelerator pedal with the optimum gear established.Therefore,the intelligent gear decision-making LSTM model is simultaneously established,compared with the BP neural network model,the advantages of the LSTM model has also been demonstrated.(4)Intelligent control of gear shifting process based on support vector machine algorithm(SVM).Based on the torque data of two clutches simulated by DCT dynamic model of gear shifting process,the fitting method with confidence interval of clutch torque data was studied.Aiming at the maximum jerk,friction work and shifting time,the optimization method of the torque data for two clutches in the shifting process was also carried out.The intelligent control method of the shifting process based on SVM model needs to be conducted by using the clutch torque data optimized,the speed difference between the main and slave ends of the two clutches and the rate of change of the speed difference.Based on the oil pressure of two clutches collected during the experiment of gear shifting process,the validation of the proposed intelligent prediction model for clutch torque could be verified.