Abstract:
In order to solve the online detection of laser welding depth of fusion and to solve the problem of optical coherence tomography (OCT) depth of fusion detection signal noise, a percentile filtering plus moving average algorithm was used to fit the weld depth curve. The weld depth of 304 stainless steel keyhole welding process was detected online by optical coherence photomicrographic imaging system, in the process of extracting the percentile depth curves from the original scatter data using percentile filtering; it was found that there were spikes interfering noises in the percentile depth melting curves extracted by this method, and the percentile depth curves were processed by low-pass filtering through the moving average algorithm, and the OCT weld depth curves were finally extracted and fitted from the original scatter data. The results show that, comparing the OCT weld depth curve with the actual weld depth curve on the longitudinal surface of the weld, the measurement accuracy of the OCT depth of fusion curve obtained by this method has been improved by 15% at most. The method is effective in extracting and fitting more accurate OCT weld depth curves from the original scatter data.