基于组合式神经网络的柴油机性能评估预测模型  

Study on the appraisal analysis and forecasting model of diesel engine performance based on modular neural network

作　　者：李增芳[1] 陶雪梅[1] 何勇[1] 

机构地区：[1]浙江大学生物系统工程系

出　　处：《浙江大学学报（农业与生命科学版）》2005年第2期229-231,共3页Journal of Zhejiang University:Agriculture and Life Sciences

基　　金：高等学校优秀青年教师教学科研奖励计划资助项目(02411);浙江省自然科学基金人才基金资助项目(R02067);浙江省教育厅资助项目(JK20020316).

摘　　要：在分析发动机结构参数和运转参数对发动机性能影响的基础上,提出了一种基于组合式神经网络的柴油机性能状态评估预测模型.该模型首先运用动态聚类法将大样本分成若干小组,然后分别用于子网络训练.性能评估时,运用模糊识别法选择相关的子网络进行评估分析.实例验证表明,这种模型能有效解决大样本下神经网络训练速度慢和难以收敛的问题,提高柴油机性能评估预测精度.By analyzing the relationship between technical states and performance parameters of diesel engine, an appraisal model based on modular neural networks is proposed. Firstly, the original samples are divided into several groups by applying the dynamic clustering method. Then each group is used to train a separate seed neural networks. During the appraisal phase, pattern recognition is employed to activate the corresponding seed neural networks for appraisal analysis. Experiment indicates that the model can solve the problem of low learning speed and difficult convergence and improve the accuracy of prediction.

关 键 词：柴油机 神经网络 动态聚类法 评估预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TK418[自动化与计算机技术—控制科学与工程]