改进粒子群优化-Elman算法在发动机曲轴脉宽预测中的应用  

Applications of Advanced PSO-Elman in Engine Crankshaft Pulse Width Predictions

作　　者：孟蓉歌 张春化[1] 梁继超 MENG Rongge;ZHANG Chunhua;LIANG Jichao(School of Automobile,Chang'an University,Xi'an,710064;Shaanxi Heavy-duty Automobile Co.,Ltd.,Xi'an,710200)

机构地区：[1]长安大学汽车学院,西安710064 [2]陕西重型汽车有限公司,西安710200

出　　处：《中国机械工程》2018年第7期766-770,共5页China Mechanical Engineering

基　　金：陕西省工业科技攻关资助项目(2016GY-002)

摘　　要：针对发动机曲轴脉宽难以预测的问题,提出了改进粒子群(PSO)优化Elman神经网络预测的方法。采用Elman神经网络建立脉宽预测模型,根据网络陷入局部最优的代数与迭代次数动态更新网络惯性权重使PSO算法得到改进,利用改进的PSO算法对Elman神经网络的权值和阈值进行优化。对YC6G270-30型增压中冷柴油机曲轴信号脉宽的预测结果表明,改进的PSO-Elman算法比最小二乘、Elman、PSO-Elman算法具有更高的预测精度,收敛速度更快,验证了所提出方法的有效性与实用性。Aimed at the unpredictability of the engine crankshaft pulse widths,advanced PSO-Elman predictive method was put forward.The model of pulse width predictions was built by Elman neural network,according to the generation of network trapped into the local optimums and the iterations,the inertia weight were updated and the PSO was improved.The Elman weight and threshold were optimized by advanced PSO.Compared with the least squares,Elman and PSO-Elman by predicting the YC6G270-30 crankshaft pulse widths,the advanced PSO has simple structures and fast convergences.At the same time,the validity and practicability of the proposed method were verified.

关 键 词：曲轴脉宽 ELMAN神经网络 粒子群优化算法 惯性权重 

分 类 号：TK40[动力工程及工程热物理—动力机械及工程] TP368[自动化与计算机技术—计算机系统结构]