摘要
基于遗传神经网络建立PQF轧制工艺模型.根据工艺要求利用最大最小值归一法对数据进行预处理.利用遗传算法的全局搜索优化BP网络的初始权重,避免BP算法陷入局部收敛,改善收敛速度.实验表明,将建立的模型应用于PQF轧制工艺设计中,能预置PQF自动化控制参数,达到良好的预测效果.
A PQF rolling process model based on genetic neural network was established. According to the process require- ments, maximum and minimum normalization method was used to preprocess data. Using genetic algorithm global search to optimize BP network initial weights, BP algorithm was prevented from getting into local convergence and the convergence speed was improved. Experiments show that using the established model in PQF rolling process design can preset PQF automation control parameters and achieve better prediction.
出处
《天津科技大学学报》
CAS
2012年第5期74-78,共5页
Journal of Tianjin University of Science & Technology
关键词
BP神经网络
遗传算法
最大最小值归一法
PQF轧制工艺
BP neural network
genetic algorithm
maximum and minimum normalization method
PQF rolling process