基于神经网络的压块机生产率预测研究

在对生产率影响因子分析的基础上，提出了应用正交试验法确定影响生产率的主要因子的快速方法。通过对主要影响因子与生产率的正交试验数据进行分析，得到用于BP神经网络预测模型的输入、输出变量以及训练神经网络所需的数据样本，多次试取隐含层和各隐含单元，并选取trainhn作为最优训练函数，建立了压块机生产率预测的人工神经网络系统。在试验结果中随机选取6组试验样本，进行连续5次挠度预测，预测值和试验实测值最大相对误差为0．14mm，解析结果表明：压块机预测结果与实验值吻合的较好，建立的人工神经网络预测模型具有较高的预测精度。

Productivity Prediction for Cube Formation Machine Based on BP Neural Network

To accurately predict the productivity of loaded cube formation machine, orthogonal experiment is used to ascertain the main impact of influence productivity. On the basis of this, the BP-artificial neural network is used to analyze productivity prediction, A BP artifi- cial network model is established by using plenty of experimental statisties as training specimens, trying to access all kinds of crytic layers and elements, choosing trainlm as optimal function. Five predictions are done continuously aiming at every group after six groups of speci- mens are selected from experimental results.The relative error between the predicted result and the experiment result is 0.14 The results show that the predicted result fits experiment result well, and show that the productivity prediction model of cube formation machine of neural network can predict accurately and rapidly.