Research On The Consistency Of Subjective And Objective Evaluation For DCT Shift Quality

The shift performance of automatic transmission is generally reflected by the shift quality,and shift quality of DCT decides the performance of the vehicle power transmission system,affecting driving feeling of driver.The evaluation methodology of DCT shift performance has become the focus during development process and scientific evaluation method is an important point to improve the shift quality of automatic transmission.The research on the evaluation method of shift quality has subjective and objective aspects,in which the objective index representing shift quality is related to the subjective score,and the mapping relationship between the subjective and objective evaluation results is more complicated.Improving the product performance of automatic transmission based on the scientific and accurate evaluation result.The subjective and objective evaluation results should be considered synthetically during the process of shift quality evaluation,and primary content of this paper is the research about the consistency of subjective and objective evaluation of DCT shift quality,the essence of the research is to establish the relationship between objective indices and subjective evaluation results during shift process.First of all,analyzing the subjective evaluation methods of shift quality at home and abroad,in order to establish the subjective evaluation system of shift quality and design the subjective scoring table.At the same time,defining the objective evaluation index based on the change of performance parameters during shift process.Considering two kinds of shift conditions and different groups of people,the test collects the vehicle bus data during the shift process which is used for extracting objective evaluation index by pre-processing and model calculation.The test results of shift quality evaluation include the objective evaluation index of shift and the subjective evaluation value of ten about drivers with different driving styles.It is found that the individual characteristics of drivers will affect the subjective evaluation results with analyzing the changing of the data.The research considers the influence on the evaluation process of shift quality because of different driving styles.Secondly,drivers of different driving styles have different concerns about DCT shift performance.Before obtaining the accurate consistency analysis results,drivers must be classified according to driving style.Building gray variable weight clustering model through gray whitening weight function,whose input data are throttle opening,throttle opening rate and maximum acceleration,the experimenters are divided into three categories: radical,moderate and conservative drivers according to the principle of maximum membership grade.The driving style identification model is built at the same time.Afterwards,analyzing the relativity of subjective and objective evaluation results to judge linear correlation relation.On the one hand,calculating the correlation ratio of each objective evaluation index to get rid of the repeated index to simplify the data;analyzing the consistency of subjective and objective evaluation results relying on three kinds of data with different driving styles.On the other hand,the neural network technology is used for the experimental data mining,the subjective and objective evaluation results of drivers with different driving styles are used as machine learning samples.Then the weights and deviation parameters of the network are obtained by model training in order to establish the nonlinear mapping between objective index of DCT shift process and subjective evaluation results.Finally,the consistency model of subjective and objective evaluation about DCT shift quality is completed in Matlab/Simulink.The model is a multiple input-single output system whose input parameters are the objective evaluation indices during the shift process,and the model can evaluate the DCT shift process instead of the driver.The result about this research shows that the error about the output results of the model is low,which would be beneficial to the personalized and intelligent calibration of DCT.