摘要
系统地介绍了工程车辆模糊神经网络 (FNN)换挡控制的后续研究。针对原系统神经网络训练收敛速度慢、挡位决策时间长的问题 ,在仔细分析原控制系统缺陷的基础上 ,对系统中的挡位决策信息进行了研究。主要的改进工作包括训练样本的评价准则、按工况分离方法、输入与排序、表示方法等 ,最后改进了控制程序 ,并且给出了改进前、后系统性能对比试验的结果。
In the paper, aiming at the problems existed on our previous developed FNN shifting control of construction machinery, long training time of FNN, some improvements were made. The improvements on the evaluation, separation, input and seriation of sample were made, and the control program was re-compiled with the multi-thread technology. Four worker threads were employed to process 4 working states of vehicle respectively, which were fore-going with up-gears, fore-going with down-gears, back-going with up-gears and back-going with down-gears. The samples standing for each state were stored in the memories of themselves, and the input order of samples was modified according to a practical vehicle working. Finally, a comparative test between the improved system and the original one was carried out.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2004年第2期1-4,共4页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目 (项目编号 :5 0 0 75 0 3 3)
