神经网络和遗传算法在Assel轧机辊型设计中的应用

来源期刊：冶金自动化2009年第1期

论文作者：陈希 肖晋为 王玉峰 钟锡弟

文章页码：1 - 5

关键词：Assel轧机； 轧辊特征参数； 神经网络； 遗传算法

Key words：Assel rolling mill； roll characteristic parameter； neural network； genetic algorithm

摘    要：为了在引进的Assel轧机上拓宽产品规格,满足市场需求,提出了一种Assel轧机辊型设计的方法。首先利用遗传神经网络建立模型,推理出轧制每种规格管材的喂入角和辗轧角,然后利用德国M eer公司辊型设计公式计算轧制每种规格管材的Assel轧辊特征参数,最后采用最小包容区域法和遗传算法进行参数寻优,确定一种Assel轧机辊型的特征参数。采用上述方法优化设计的Assel轧机辊型完成了6种管坯600种规格无缝钢管的轧制,比原引进设备的设计产品规格扩大了10余倍,且产品质量处于同类机组领先水平。

Abstract: In order to broaden product specifications and meet market demand at imported Assel rolling mill,a design method of roll profile for Assel rolling mill was presented.Firstly,the model was establisted using genetic artificial neural network(GANN),and feed angle and toe angle for each rolling gauge tube were obtained.Then,Assel roll characteristic parameters for each rolling gauge tube were calculated via roll profile design formulas provided by Meer Co.of Germany.Finally,with least-inclusive regional and genetic algorithm optimization,characteristic parameters of a roll profile of Assel rolling mill were determined.After the roll profile designed based on above mentioned optimization was used in Assel rolling mill,more than 600 kinds of seamless pipe including 6 kinds of pipe blank have been produced.Compared with imported equipment,product gauge has been expanded more than 10 times.Product quality reached advanced level among the similar type of mills.