The Research On Regression Prediction Of Engine Economy And Emission Performance Based On Machine Learning Algorithm

The use of machine learning algorithms in the development of automotive engine performance for regression analysis of the engine’s economic and emissions performance is different from traditional technical means.The use of mechanism-based numerical prediction models can achieve the best prediction accuracy,but it is difficult to achieve the desired effect.The regression prediction method based on machine learning algorithm can r educe the technical difficulty and provide a new technical attempt.Based on various machine learning algorithms,this paper conducts regression prediction research on the BSFC,NO _x,HC,CO and CO₂parameters of the engine,and uses genetic algorithms to a djust the hyperparameters of the regression prediction model to optimize the prediction performance of the model.The machine learning regression algorithms used include linear regression algorithm,K nearest neighbor regression algorithm,support vector m achine algorithm,decision tree algorithm,multi-layer perceptron algorithm and integrated learning.The main conclusions of the paper are as follows:（1）In the regression prediction of the economic indicator BSFC,after the algorithm optimization,the training R² value of the regression tree,multi-layer perceptron algorithm and integrated algorithm s are all above 0.99,and the test R²value are above 0.94,which are algorithm choices.（2）In the regression prediction of the emission index,after the algorithm optimization,the Gradient Boost algorithm’s NO _x training and test R² values are 1 and0.85,respectively,which is the best algorithm choice;for HC,there is a certain overfitting of each regression prediction model Suppression,the Gradient Boost algorithm performs better,and the test R² value is 0.69221,which is a better choice;for CO,the Gradient Boost algorithm test R² value can reach the highest 0.93654,which is the best choice;for CO₂,the random forest,Ada Boost and Gradient Boost algorithms have the highest prediction accuracy compared to other regression algorithms.（3）On the training set,the genetic algorithm greatly improves the prediction accuracy of the K-nearest neighbor regression algorithm,and has little effect on the integrated algorithm.On the test set,the optimization algorithm greatly suppresses the overfitting of polynomial linear regression,and has a significant effect on the prediction accuracy of the Ada Boost and Gradient Boost algorithms.The use of machine learning regression algorithms and optimization algorithms for regression prediction can help in the calculation of economic and emission indicators,and is of great significance to the optimization of the engine’s economy and emission performance.