摘要
以北京现代伊兰特G4GD发动机为试验台,将电控系统故障作为实验变量,测得规定时间内双传感器组合发生故障时的发动机怠速,并选原车ECU较难控制的6种组合怠速故障进行分析。基于量子粒子群算法(QPSO)对长短时记忆神经网络(LSTM)隐含层节点、训练次数与学习率进行寻优预测,将预测结果与多种神经网络进行对比,并通过均方根误差(RMSE)评价指标进行判断。使用Origin数据拟合将预测输出结果进行数值拟合,之后输入Matlab中使用Simulink搭建控制单元模型,由模糊常量-积分-微分(FPID)控制器对输出结果进行怠速控制。结果表明:基于量子粒子群算法改进的长短时记忆神经网络预测效果最好;模糊常量-积分-微分控制器对怠速的控制可有效缩短电子控制单元(ECU)的控制时间,无超调,且可有效调节至规定怠速。
Taking Beijing Hyundai Elantra G4GD engine as the testing bench,and the electronic control system fault as the experimental variable,the engine idle speed when the dual sensor combination fails within the specified time is measured,and six combined idle speed faults that are difficult to control by the original vehicle ECU are selected for the analysis.Based on the quantum particle swarm optimization(QPSO)algorithm,the hidden layer nodes,training times and learning rate of the long-term and short-term memory neural network(LSTM)are optimized and predicted.The prediction results are compared with the results of various neural networks,and judged by means of the evaluation indicators such as root mean square error(RMSE).The predicted output results are numerically fitted by means of Origin data fitting and input into Matlab Simulink to build the control unit model.The fuzzy constant-integral-differential(FUZZYPID,referred to as FPID)controller is used to control the idle speed of the output results.The results show that the improved LSTM based on QPSO has the best prediction effect.The FPID controller can effectively shorten the control time of the electronic control unit(ECU)for idle speed control,without overshoot,and can be effectively adjusted to the specified idle speed.
作者
赵晴
潘江如
董恒祥
郭鸿鑫
ZHAO Qing;PAN Jiangru;DONG Hengxiang;GUO Hongxin(College of Transportation and Logistics Engineering,Xinjiang Agricultural University,Urumqi 830052,China;College of Control Engineering,Xinjiang Institute of Technology,Urumqi 830023,China)
出处
《现代电子技术》
北大核心
2024年第8期75-82,共8页
Modern Electronics Technique
关键词
发动机怠速
量子粒子群优化算法
长短时记忆神经网络
模糊PID控制
故障分析
时间序列预测
fuel engine idle
quantum particle swarm optimization algorithm
long short-term memory neural network
fuzzy PID control
fault analysis
time series prediction
