Identification of welding torch deviation with rotating arc sensor

Identification of welding torch deviation is necessary to accomplish welding seam tracking.In consideration of the fact that welding current signals are often disturbed by outside noises, soft threshold wavelet filtering method is applied to process welding current signals, which makes the welding current shape obviously smooth and the signal-to-noise ratio much be improved.For more exactly acquiring the distance of welding torch depart from welding seam, fuzzy clustering and artificial neural network deviation identifying methods were proposed.Left-right integral method, character harmonic method, fuzzy clustering method and artificial neural network method were used to identify the welding torch deviation, respectively.In the last, the deviations acquired by the four methods were fused and the last deviation was obtained.The experiment results show that the fusion method can greatly improve the precision and reliability of identifying welding torch deviation.