There is needs varying heat input during welding for improving the weld quality. But it is not guarantee without maintaining stable are during variation of heat input. Current researches remain in monitoring the arc state, and arc stability control method is not developed up to now.
In this study, a learning controller using gradient method is proposed to maintain stable arc state in spite of heat input variation. The factors which affect the arc stability are investigated by a series of experiments, then the learning controller is designed based upon the experimental results. After design parameters of the proposed controller are experimentally determined for a constant heat input. Then, the learning controller with the parameters is implimented for a varying heat input. The results of the learning control experiments show that arc stability is remarkably enhanced under variation of the heat input during arc welding.