Research On Subjective And Objective Evaluation Methods

With the development of the domestic auto industry,consumers pay more and more attention to the driving experience of vehicles,and various OEMs continue to strive to create models with good driving experience to enhance market competitiveness.The improvement of drivability needs to be verified in the stage of product development and finalization.Accurate drivability evaluation will help car companies to clarify the direction of improvement in drivability optimization and provide a reference for drivability improvement.This thesis takes the model of a domestic OEM as the research object,starts with the driving experience survey,analyzes the influencing factors and principles of drivability,stipulates the scoring rules and objective evaluation index definitions for the subjective evaluation of drivability,and conducts the evaluation according to the content of the subjective and objective evaluation.Real car test.According to the test data,a BP neural network and extreme learning machine(ELM)are used to establish a driving subjective score prediction model.The specific research results are as follows:(1)Starting from the results of the drivability investigation,the factors and influence laws that have a greater impact on drivability,such as engine speed,accelerator calibration scheme,transmission speed ratio design,transmission type and shift logic,etc.,are analyzed.Elements that should be possessed by the evaluation method of performance: accuracy,stability,test repeatability.(2)According to the driving and riding usage scenarios of the vehicle,the test is divided into engine starting,idling,starting,accelerating,decelerating,shifting,and pressing and releasing the accelerator quickly.For each working condition,corresponding subjective evaluation items and scoring rules are specified.In terms of objective evaluation,the objective evaluation indicators under twelve sub-conditions are defined,as well as the driving process data content that the corresponding indicators need to obtain.(3)For the subjective evaluation items and the data required for the objective evaluation,carry out a real vehicle test,and use the AVL-Drive software and its supporting hardware system to collect data in the process of the subjective evaluation driver’s driving and scoring.Noise reduction processing is performed on the acceleration signal obtained from the test.Using wavelet threshold noise reduction and empirical mode decomposition(EMD)processing respectively,the signal-to-noise ratio of the signal after EMD noise reduction is 4% higher than that of wavelet denoising,and the root mean square error is 8.9% lower than that of wavelet denoising.Therefore,EMD has a better noise reduction effect on the acceleration signal in this thesis.(4)Based on the data collected from the real vehicle test,the drivability evaluation index value is obtained by processing it,and at the same time,the consistency between the subjective evaluation and the objective evaluation is studied in combination with the scores given by the subjective car appraisers for the corresponding process.BP neural network and ELM were used for nonlinear regression analysis respectively,and a subjective score prediction model was established.The trained model is validated on the test set.The prediction results show that the maximum prediction error of the BP neural network model is 0.1999,and the maximum prediction error of the ELM model is 0.4129,The prediction errors are all less than 0.5.ELM is more practical when the amount of data is large due to its low modeling difficulty and high learning rate.