用于SPC控制图识别的神经网络训练并行算法研究

Research on Neural Network Parallel Algorithm for SPC Control Charts' Identification

在研究统计过程控制基础上建立控制图识别的神经网络模型。通过对前向型神经网络分析,设计了粒子群优化算法与阻尼最小二乘法相结合的复合训练算法,基于并行运算原理对上述复合算法进行优化。实验结果表明,该复合算法经过并行优化后训练速度明显提高,具有较大的实用价值。

Design a neural network model for control chart identification basic on the research of statistical process control. After the forward neural network analysis, design a particle swarm optimization algorithm with Levenberg- Marquardt method of complex training algorithm. Use parallel computing implementation of the optimization of complex algorithms based on the principle of parallel computing. Experimental results show that the compound through the parallel optimization algorithm greatly enhance the training speed.