基于高斯过程回归模型的发动机性能预测

Engine Performance Prediction Base on Gaussian Process Regression Model

基于1.5L汽油增压发动机的VVT标定试验,对进排气VVT角度采用拉丁超立方抽样建立试验样本,并构建高斯过程回归模型,通过试验数据分别训练扭矩回归模型和油耗回归模型,使用训练后的回归模型预测发动机扭矩、比油耗。通过与发动机万有特性测试数据对比,结果表明:在发动机扭矩大于25N.m的区域,油耗回归模型的预测值偏差小于5%;在发动机小负荷区域,因测量误差、发动机燃烧等因素影响,油耗回归模型的预测值偏差较高。总而言之,高斯过程回归模型对发动机万有特性的预测具有较高的精度及普适性,为发动机参数的优化工作提供参考。

Based on VVT calibration test result of 1.5L gasoline turbocharged engine, test plan for VVT Angle of intake and exhaust was sampled via Latin hyper-cube method, torque and fuel consumption Gaussian process regression(GPR) models were constructed. Torque regression model and fuel consumption regression model were trained respectively by test data. Compared the engine universal characteristics with test data, the results show that the deviation of the predicted value of the fuel consumption regression model is less than 5% in the region where the engine torque is greater than 25N.m. Due to the influence of measurement error, engine combustion etc.,the predicted value of fuel consumption regression model has high deviation in small engine load area. In a word, Gaussian process regression model has high accuracy and generalization for the prediction of engine universal characteristics, which can provide reference for the optimization of engine parameters.