作　　者： ; ; ;

机构地区： 燕山大学机械工程学院

出　　处： 《塑性工程学报》 2009年第2期10-14,共5页

摘　　要： 智能化拉深过程中材料性能参数实时识别是关键技术之一。盒形件拉深难以用精确的力学模型来描述,文章引入基于LM算法的神经网络模型对材料参数进行识别,并仔细研究了在盒形件拉深过程中的适用性。针对盒形件提出了拉深初期采用恒定压边识别的方案,并采用平均值和去除奇异数据的方法大幅度地减小了识别误差,在该文的样本数据范围内,4种材料性能参数的最大识别误差在2%以内,为实现整个拉深成形过程的智能化控制奠定了基础。 Parameters identification in intelligent deep drawing is a key technology. In this paper, neural network model based on LM algorithm is established to identify parameters of material properties because it is difficult to find the exact mechanical model. For rectangular box deep drawing how can neural network do better in identification is also studied. As a result for this case identification should be at the period of begin time. How to reduce the identification error is also studied and at last the max identification error of four kinds of material mentioned in this paper is within 2 %. These measures will ensure the intellectualized control for the whole process of deep drawing.

关 键 词： 智能化拉深 参数识别 算法 神经网络

领　　域： [金属学及工艺]