基于模糊神经网络的弧焊熔池几何参数调节  

Neural Networks Based Self-organized Fuzzy Logic Control in Arc Welding Pool Geometry Regulation

作　　者：李迪[1] 宋永伦[1] 井上胜境[2] 

机构地区：[1]华南理工大学 [2]日本国大阪大学

出　　处：《焊接学报》1999年第S1期135-141,共7页Transactions of The China Welding Institution

基　　金：广东省自然科学基金! (94 0 5 0 4;990 5 5 0 );国家自然科学基金!(5 9785 0 0 4 );教育部留学回国人员科研启动基金

摘　　要：模糊控制(FLC)在焊接过程控制中是一个非常有前景的控制方法。但是对于一般的FLC模糊规则依赖于专家(如高级焊工)的经验。另外,模糊集的隶属度函数是非自适应的,即一旦由经验或其它方法确定后则保持不变。对焊接这样一个时变、强干扰系统,固定的隶属度函数不能保证所需的性能特性,而应采用自适应隶属度函数改善系统的性能。因此,模糊规则集的自动生成及隶属度函数的在线调整对于现代焊接过程控制来说是迫切需要的。本文将FLC和神经网络相结合,研究实现上述目的的可行性。隶属度函数的自适应及模糊规则的自组织通过神经网络的自学习和竞争来实现。并以GTAW过程为对象验证所提出的方法。计算机仿真表明系统的性能得到提高。Fuzzy Logic Control (FLC) is a promising control strategy in welding process control. However, in basic FLC, the fuzzy rule relies heavily on the experts' (e.g. advanced welders') experience. In addition to this, the membership function for fuzzy set is non-adaptive, i.e. it remains unchanged as long as it is determined by experience or other means. For welding process, which is time-variable systems and strong disturbance exists in it, fixed membership function may not guarantee the required system performance, and attempts should be made to improve the system performance by adopting adaptive membership function. Therefore, the automatic determination of the fuzzy rule and in-process adaptation of membership function are required for the advanced welding process control. This paper discussed the possibility by using the combination between FLC and neural network (NN) to realize the above purpose. The adaptation of membership function as well as the self-organizing of fuzzy rule is realized by the self-learning and competitiveness of the NN. Taking GTAW process weld bead width regulating system as the controlled plant, the proposed algorithm was testified for such a process. Computer simulation showed the improvement of the system characteristics.

关 键 词：焊接 模糊控制 神经网络 

分 类 号：TG444[金属学及工艺—焊接]