基于自适应共振人工神经网络的焊缝类型自动识别方法

Research on Recognition of the Seam Type Based on Adaptive Resonance Theory Network

介绍一种用自适应共振人工神经网络自动识别焊缝类型的方法。该方法首先根据不同类型的焊缝对电弧光和激光光带的影响， 将弧焊焊缝划分为 ４ 种类型， 在此基础上， 确定焊缝的特征参数， 并组成训练样本数据库， 由此抽取焊缝图像特征。制做 Ａ Ｒ Ｔ２ 人工神经网络分类器， 将 ４ 类焊缝的权值保存在其长期记忆层， 在实际分类时， 用 Ｃ Ｃ Ｄ 摄像机将检测到的焊缝特征参数输入， 处理后得到焊缝类型。对实时焊接过程中的焊缝进行跟踪实验， 跟踪精度在±０５ ｍ ｍ 之内。

It is very important for the automation of seam tracking system to identify seam type automatically, most of the method used is complex and has low reliability. A method used to recognize automatically the seam type based on adaptive resonance theory neural network is discussed in this paper. It is very useful for the seam line tracking. It can recognise the type of seam very fast and very reliablely. Some key technologies used in this paper are introduced. Tracking experiment is done, and tracking error is less than ±0.5mm.