Research On Model-based Parameter Optimization Method Of Clutch Starting Control

In the calibration field of automatic transmission control system,the artificial calibration method is usually adopted both at home and abroad,to complete the calibration of control parameters by lots of calibration tests of road conditions under different driving environments,considering the relationship among the various conditions and control parameters,and depending on the experience of engineers.Because of the complex working conditions and fuzzy control targets,it is difficult to get the optimal solution of control parameters,which results in many problems of too large calibration workload,time-consuming,high cost,and calibration results with empirical subjective willingness;and in addition,it also takes high cost and long time to cultivate a mature calibration engineer.Therefore,it is the constant target pursuit by domestic and foreign automobile manufacturers to seek a kind of calibration technology with high efficiency,low cost,and small dependence on technological experience.Model-based parameter optimization method is one of the hot topics.The paper is the subject of the National Natural Science Fund Project "Real-time gear optimization and online decision-making technology of automatic transmission based on power demand",and with the reference to the virtual calibration method widely used in the field of engine calibration,the model-based parameter optimization idea for clutch starting control that integrates system modeling in the simulation environment,design of experiment,analysis of performance prediction model and multi-objective optimization algorithm was proposed,so as to realize the clutch control performance prediction with various control parameters under the virtual calibration environment,determine the control parameters and automatically find the optimal solution,and achieve the target of improving the quality of clutch control performance and calibration efficiency,and reducing the development cycle and development costs.The main research content of this paper is in the following aspects.(1)The real-time simulation platform based on HIL was built.The dynamic analysis of the vehicle system and clutch starting process was carried out,the physical model and starting control model were respectively established based on AMESIM and MATLAB;then using x PC as the target machine,the semi physical simulation platform was set up with TCU in HIL,which laid the foundation for the later starting control quality evaluation and optimization.(2)The response model of performance prediction was established.Only for flat starting conditions of vehicles with AMT,the appropriate evaluation index were selected from the vehicle control model in demo,to design the starting control quality evaluation system.Based on the semi physical simulation platform built in the last chapter,the data acquisition was obtained by using Latin square experiment design,and sensitivity between the input clutch control parameters and output evaluation index was analyzed by the mathematical statistics method,to establish the performance prediction response model of input and output based on BP Neural Network,which provided the optimization target for AMT clutch control.(3)Parameter optimization of starting control was completed.Taking the highest level of integrated starting quality evaluation as the objective function,the clutch starting control parameters in the control strategy that influence the starting quality were optimized by NSGA-? multi-objective optimization algorithm,and the optimal solution set was obtained,which was then verified by the system simulation.Then the whole vehicle test was designed based on the real car,and the optimization results were further verified,with the better starting performance.