Abstract: In order to improve the stability of magnetic-control arc weld automatic tracking system and the tracking precision, the Latin hypercube sampling test design method and the technology of Kriging surrogate model was used to establish the prediction model of the magnetic-control arc sensor parameters (excitation frequency, excitation current, the magnetic gap, coil number of turns), analyze the main parameters of magnetic-control arc sensor on the influence law of seam tracking signal and its stability in welding process,then select the optimized sensor parameters as designed parameters. The reliability of the optimized magnetic-control arc sensor parameters was verified by experiments. Finally,the results testified that the waveform of the optimized seam tracking sensor signal was obvious and less disturbance.The improved noise ratio of the signal fundamentally improved the stability and the accuracy of seam tracking.