论文部分内容阅读
研究了基于神经网络的丝杠螺纹磨削过程的智能预测与控制问题.基于误差反向传播的机制,针对连续制造过程的预测与控制,提出多层神经网络的逐个样本学习算法.对逐个样本学习算法和目前广泛采用的B-P算法进行了比较和讨论,实验结果表明,逐个样本学习算法比B-P算法具有更好的收敛性.最后,介绍了多层神经网络模型在丝杠螺纹磨削过程的预测与控制中的应用.
The problem of intelligent predictive and control of screw thread grinding process based on neural network is studied.Based on the mechanism of error backpropagation, a sample-by-sample learning algorithm of multi-layer neural network is proposed for the prediction and control of continuous manufacturing process. Learning algorithm and the widely used BP algorithm are compared and discussed, the experimental results show that the sample-by-sample learning algorithm has better convergence than the BP algorithm.Finally, the multi-layer neural network model is introduced in the process of screw thread grinding Prediction and Control Applications.