摘要
文章以滚压速度、进给量、滚压力、滚压次数为输入参数,建立了对表面粗糙度进行预测的径向基函数神经网络模型,利用该模型对高硅铝合金基复合材料的已加工表面粗糙度进行了预测。结果表明,预测值可达到满意的精度要求,对7组样本进行预测时最大相对误差不超过12%,且表面粗糙度值越大,模型的预测效果越明显;模型的学习速度和精度均优于传统的BP神经网络。此外,利用所建立的模型对滚压工艺参数进行了优化,得出了工艺参数的最佳范围。
A radius basis function (RBF) neural network was introduced to predict surface roughness of burnished high silicon aluminum alloy based composites using input vectors such as burnishing speed, feed rate, burnishing force and number of passes. It was found that the model based on RBF was superior to the model by conventional BP neural network, which had faster convergence speed and better accuracy. The maximum relative error of prediction was no more than 12%. The larger the surface roughness, the higher the prediction accuracy was. Meanwhile, the optimization of four main process parameters was conducted using the established model.
出处
《组合机床与自动化加工技术》
2005年第4期8-10,共3页
Modular Machine Tool & Automatic Manufacturing Technique
基金
国家自然科学基金重点项目(50135020)
