神经网络在铝合金穿孔型等离子弧立焊中的应用

铝合金穿孔型等离子弧立焊是一种高能密度焊 ,焊接过程是一个复杂的、多参数耦合的高度非线性过程 ,存在许多难以量化、不确定的知识 ,无法实现解耦 ,在生产中难以进行有效的控制 神经网络具有非线性映射和自组织、自学习的特点 ,使其在焊接领域中的应用具有一定的优势 本文基于MATLAB6.1的神经网络工具箱 ,利用 3层BP人工神经网络建立铝合金穿孔型等离子弧立焊的输入 -输出的网络模型 ,通过训练该网络 ,能够根据输入节点的各个焊接工艺参数值 ,预测焊缝形状的输出结果 焊缝的形状参数与网络模型预测结果的误差在 8%以内 仿真试验结果表明 ,这个方案是可行的

Application of Neural Network in Aluminum Keyhole Plasma Welding in Vertical Position

Aluminum keyhole plasma welding in vertical position is a kind of welding of high density energy. It is a highly nonlinear process with complicated multi parameters coupling and many uncertain factors.So it is difficult to control the process effectively. Neural network has certain advantages in welding control because of its nonlinear mapping,self study and self orgnization. In this paper, Based on MATLAB6.1 neural network toolbox, a 3 layer BP neural network model about vertical position PAW input output is established.According to each welding parameter's value of input nodes,the output of weldment formation can be predicted by testing this network. The errors between parameters of welding formation and predicted results are within 8%. The results of simulation show the method to be practicable.