基于Kriging理论的发动机万有特性研究  

Study on the engine’s universal characteristics based on the Kriging theory

作　　者：何伟 徐继东 陈翠翠 HE Wei;XU Ji-dong;CHEN Cui-cui(Weichai Power Co.,Ltd.,Weifang 261021)

机构地区：[1]潍柴动力股份有限公司,山东潍坊261021

出　　处：《机械设计》2020年第8期102-106,共5页Journal of Machine Design

摘　　要：发动机万有特性是衡量发动机燃油经济性并进行动力总成匹配的重要依据。为提高万有特性建模的精度,以某系列柴油机试验数据为例,基于Kriging理论并采用遗传算法优化其初始值θ0建立万有特性预测模型。在此基础上,对比分析优化初始值θ0后的Kriging理论、BP神经网络及最小二乘法所建立的万有特性预测模型精度和鲁棒性。结果表明:采用遗传算法优化后的Kriging理论建立的万有特性预测模型精度和鲁棒性均高于BP神经网络和最小二乘法。The engine’s universal characteristics serve as an important basis to measure the engine’s fuel economy and match the power assembly.In order to ensure a high standard of accuracy,with the test data on a series of diesel engines as the example,the model of predicting the universal characteristics is worked out based on the Kriging theory,and the genetic algorithm is adopted to optimize the Kriging initial valueθ0.Based on this,the contrast analysis is conducted on the accuracy and robustness of the prediction model which is set up by the Kriging theory with the optimized valueθ0,the BP neural network,and the least square method.The results show that the accuracy and robustness of the prediction model which is set up based on the Kriging theory and the genetic algorithm are more desirable than those of the models which are set up by the BP neural network and the least square method.

关 键 词：发动机 万有特性 Kriging理论 BP神经网络 

分 类 号：TH122[机械工程—机械设计及理论]