Research On The Matching Of Function And Mass Of Vehicle Body In White Based On Machine Learning

With the increase of people’s requirements on all aspects of the car,the weight and cost of the car are also increasing,and the impact on the environment is also increasing.Therefore,automotive lightweight design has been a hot issue in academia and industry.However,the realization of lightweight design always needs a lot of time,so the lightweight potential design method can help people find the most promising aspects to achieve weight loss and improve the efficiency of lightweight design.At present,although there are many researches on lightweight methods,there are few researches on the potential of lightweight.And the matching of function and mass is an important part of the research on the potential of lightweight.Based on the target weighing approach,this paper firstly analyzes the functions of the automobile body-in-white system based on the collected relevant data by establishing a function structure.And then matches the functions and mass of vehicles body-in-white to estimate the mass of each function thinking about comparing functions and analyzing the data from people.Then,according to the estimation results,the method of machine learning is adopted to predict the functional mass,which can simplify the determination process of the functional mass.The cross-validation method was used to compare the performance measurement results of multiple learning algorithms,such as linear regression,Bayesian ridge regression,elastic network regression,support vector regression,gradient boosting regression and XGBoost model.And it was found that XGBoost model and gradient boosting regression had better prediction effect on the functional mass.Finally,based on the idea of SMOTE algorithm,the original data volume was expanded,and the influence of the expanded data volume on the prediction results of functional mass and the performance of each machine learning model were analyzed.It is found that after the increase of data volume,the better performance evaluation is still the XGBoost model.In the prediction of some functional mass,the performance measures of linear regression,Bayesian ridge regression,elastic network regression and support vector regression are improved.