摘要
神经网络具有良好的记忆、归纳和学习能力 ,对难以用数学方法建立精确模型的信息、工艺等能够进行有效地预测建模。该文通过对BP神经网络的分析和研究 ,针对传统BP算法的不足 ,采用Levenberg -Marquardt(LM)优化算法的建立一个基于BP神经网络预测建模系统。在介绍了系统的主要功能之后 ,给出了用MATLAB软件实现该系统主要模块的具体程序。最后采用该系统对一个制造过程中刀具磨损量的进行了预测建模 ,实验仿真结果表明 :系统具有良好的预测效果 ,刀具实际磨损量与预测磨损量的误差基本上在 10 %以下。
The neural network has the abilities of memory, induction & study ,it can effectively build forecasting model for information & technics which are difficult to be built into an exact math model.A BP neural network for forecasting model based on Levenberg-Marquardt(LM)algorithm has been introduced in this article, through the analysis & research of BP neural network and contraposing its shortage. Also the author provides the detailed program of MATLAB after introducing the main function of the forecasting model system. Finally an example of tool wear during machining is forecasted with the system. The simulation results show that the forecasting model system is effective and most of all the forecasting error between the actual tool wear and forecasting tool wear are below ten percent .
出处
《计算机仿真》
CSCD
2004年第9期48-50,89,共4页
Computer Simulation
基金
广西教育基金项目资助 (项目编号 :桂教科研 [2 0 0 2 ] 3 16号 )