High-frequency resistance straight seam welding quality state online detection method

Abstract

The invention provides a high-frequency resistance straight seam welding quality state online detection method. The method mainly comprises the steps of welding image collection, main component analysis image dimension reduction, radial basis function neural network training and prediction outputting. According to the invention, welding spot fusion phenomenon state images collected at a high speed are utilized for the detection of the welding quality, the collected image samples are preprocessed, then main component analysis is utilized for dimension reduction of the image data, training is carried out in such manner that the image data after the dimension reduction is used as input of a radial basis function neural network and the wielding quality states corresponding to the image samples are used as output of the network, and finally the trained network is used for prediction. By adopting the method, the high-frequency resistance straight seam welding quality state can be effectively detected on line.

Description

technical field [0001] The invention relates to an on-line detection method, in particular to an on-line detection of the quality state of high-frequency resistance straight seam welding. Background technique [0002] In recent years, artificial neural network has received more and more attention in fault detection due to its characteristics of distributed parallel processing, high fault tolerance, and strong generalization, and has opened up a new way for fault detection. Due to its local approximation characteristics, the radial basis neural network has a fast convergence speed in the learning process, and has the best approximation characteristics. There is no local minimum problem, and it can achieve good results for classification problems. [0003] In the past two years, according to the influence of electrode pressure, welding current and welding time on welding defects, metallographic analysis and tensile shear tests have been used to detect welding quality condition...

AI Technical Summary

Problems solved by technology

These methods can only be detected offline after welding, or the image processing algorithm is complex, which affects the real-time performance of the detection, or the detection recognition rate cannot meet the welding quality requirements of high-quality oil and gas pipelines. Therefore, it is urgent to propose a The online real-time detection algorithm of welding quality status can take into account the requirements of online linearity, real-time and high recognition rate, and meet the requirements of on-site production of straight seam high-frequency welded pipelines

Images