摘要
利用神经网络预测车削表面的粗糙度有利于改进车削过程的自动化程度,但神经网络输入数据的误差和网络自身的缺陷不可避免地给预测带来了误差。采用了一种基于T-S网络的技术,对原神经网络的输出进行了修正,能有效地减少预测误差。相关的试验不但证明了其有效性,而且还对网络结构和有关参数提出了建设性的建议,其结果对实践有重要的指导意义。
The improvement of automation in turning process is achieved by the surface roughness forecast with network, while the error in data as well as the defect of network itself will lead to inevitable error in the prediction. The forecast error can be reduced effectively with a modification technique for the output of original network, based on T-S network. The experiment involved not only improves the effect, but also indicates the constructive opinions for the structure of the T-S network and the parameters concerned, which is significant for practices.
出处
《机电工程》
CAS
2008年第7期15-16,33,共3页
Journal of Mechanical & Electrical Engineering
基金
宁波市自然科学基金资助项目(2006A610035)
关键词
表面粗糙度
T-S网络
预测
修正
surface finish
T-S network
forecast
modification
