基于机器学习的车用发动机性能预测及优化方法研究  

Study on Vehicle Engine Performance Prediction and Optimization Method Based on Machine Learning

作　　者：王威 刘吉绪 吴春玲 韩松 李国田 郝婧 WANG Wei;LIU Jixu;WU Chunling;HAN Song;LI Guotian;HAO Jing(China Automotive Technology and Research Center Co.,Ltd.,Tianjin 300300,China)

机构地区：[1]中汽研汽车检验中心(天津)有限公司,天津300300

出　　处：《内燃机》2023年第5期28-34,共7页Internal Combustion Engines

基　　金：国家重点研发项目(2022YFC3703600)。

摘　　要：随着人工智能技术的不断进步,基于机器学习的研究方法逐渐被应用于解决车用发动机性能优化问题。本文提出了一种基于机器学习的车用发动机性能预测及优化方法,并进行了案例研究:通过利用台架试验数据,建立了遗传算法-反向传播神经网络(GA-BPNN)预测模型,对发动机功率和有效燃油消耗率(BSFC)实现了较为准确的预测,误差率仅分别为1.58%和1.72%。此外,采用交叉遗传-粒子群(CMPSO)算法对功率和BSFC进行了多目标优化,将最优控制参数输入到台架试验中,得到的功率和BSFC的实际运行值与优化值基本一致。研究结果证明了本文提出的方法的有效性。该方法在保证一定精度的前提下,大幅减少了时间和经济成本的投入,为发动机性能优化研究提供了一种新的工作思路。As artificial intelligence technology continues to advance,research methods based on machine learning are gradually being applied to address automotive engine performance optimization problems.This paper proposes a machine learning-based method for predicting and optimizing automotive engine performance and presents a case study:a GA-BPNN prediction model is established using chassis test data to achieve reasonably accurate predictions of engine power and BSFC with errors of only 1.58%and 1.72%,respectively.Additionally,a CMPSO algorithm is applied to perform multi-objective optimization of power and BSFC.The optimal control parameters are then implemented in the chassis test,resulting in closely matching actual operational values with the optimized values.The research findings demonstrate the effectiveness of the proposed method.This approach,while ensuring a certain level of accuracy,significantly reduces time and economic costs,providing a new avenue for research in engine performance optimization.

关 键 词：发动机功率 有效燃油消耗率 机器学习 反向传播神经网络 优化 粒子群优化算法 

分 类 号：U464[机械工程—车辆工程]