二氧化碳短路过渡焊接电流神经模糊控制系统

Research of Fuzzy-Neural Network Control System for Carbon Dioxide Short Circuit Transition Arc Welding

研制了一种以 89C5 2单片机为核心器件 ,通过自软件方式实现稳定二氧化碳短路过渡焊焊接电流的神经网络自适应模糊控制系统 .系统中所嵌入的神经网络可实现对隶属函数的微调和模糊控制的调节 ,克服了常规模糊控制系统隶属函数非自适应性的缺点 .该神经模糊控制器可消除因电弧电压调节、网络电压波动、保护气纯度流量以及焊炬高度变化等随机因素引起的焊接电流偏差 .焊接工艺实验表明 ,在实验电流范围内 ,平均电流的最大偏差不超过 7A ,平均电流相对误差小于 5 % ,而控制前平均电流的偏差不小于 12A ,相对误差不小于 9% ,因此该控制系统响应速度快 ,超调小 。

A fuzzy-neural network controlled system with single chip 89C52 is developed to stabilize welding current of carbon dioxide short circuit transition welding. The system can self-adjust the memberships and fuzzy control rules through the embedded neural network algorithm, and the exiting fuzzy controller memberships do not possess the above function. The designed fuzzy-neural network controlled system can eliminate the welding current deviation which may be caused by the adjustment of the arc voltage, the fluctuation of the circuit net voltage, the variety of the flow and the purity of the shielding gas and the change of height of the welding torch. The welding experiment results show that in the range of experimental current the absolute deviation of welding current is less than 7 A and the relative error of welding current is less than 5%, but the current deviation is not less than 12 A and the relative error is not less than 9% under the existing fuzzy control alone.