Detection Of Molten Pool Infrared Characteristics And Welding State Detection During Laser Welding Of304Austenite Stainless Steel

Bead width prediction and seam deviation detection are key technologies of welding automation. In the laser tight butt welding, welding quality would be enhanced effectively by the feedback of actul bead width prediction and laser power output and welding velocity are controlled. At the same time, the laser beam must be followed the tracks of seam accurately, but it’s difficult to detect the seam deviation of micro-gap. At present, the technology of precise target detection using machine vision is maturing increasingly, while laser welding micro-gap detection based on machine vision technology is an advanced automatic welding technology. Therefore, image procesing method and mathematical modeling for actul bead width prediction and seam deviation detection were adopted in the paper, in addition, the author did the preliminary research on the mechanism of actual bead width and attempped to detect the seam deviation with metal spatter interference using machine vision technology.The weldment in the experiment is Type304stainless steel, the laser output power is10kW, high-speed camera is used for capturing the molten pool infared images. The author developed and designed the laser welding visual detection software of image processing for actual bead width detection and seam deviation detection based on Visual C++2010platform and OpenCV2.3.1library, multi-features of molten pool infared image were extracted in the experiment, feature data were ultimately processed, analysed and modeled associating with the laser welding future knowledge.Firstly, actual bead width and seam deviation were detected by image processing. Bilateral filter image processing was adopted in bead width prediction, it could preserve the edge information of molten pool width as far as possible and do FIR filter on measured value, process the mesured valeu map. The result was consistent with actual bead width vary trend, we could achieve bead width online prediction, it’s of great significance for laser welding quality control. Sub-pixel corner detection was used for testing the molten pool infared image corners distribution, seam deviation was calculated by the relationship between corners distribution and seam center position, it provides a new image processing method for seam deviation detection in high-power laser welding.Secondly, molten pool multi-features were extracted through molten pool infared image processing. Then, we filtered and normalized the feature data, analysed the correlation between the features, according to the laser welding future knowledge, we searched the factors affected actual bead width and features contained seam deviation information and metal spatter interfere information. The results showed that the centroid of the keyhole changes relative to the position of laser beam in the welding direction has a close relationship with the bead width, essentialy, it reflected laser energy concentration area changes in the welding direction, it provides a new theoretical basis for forming principle of bead width and its online prediction.Finally, In order to improve the accuracy of bead width prediction and seam deviation detection. On the basis of image processing directly predicted the bead width and detected seam deviation, extraction and analysis of the molten pool features according to the knowledge of laser welding and the correlation between the features analysis to select the features, Use these features to establish the binary regression model of bead width prediction and seam deviation detection. The accuracy of the model has been further improved compared to the image processing method.The research showed that the image processing method can predict the bead width and detect the seam deviation in real-time. But the accuracy only overall meet the requirements of laser welding automation because the interference of metal spatter and other noise. In order to improve the accuracy of bead width prediction and seam deviation detection. According the knowledge of laser welding to analysis the correlation between molten pool features, look for the key features of which affect the bead width and contains the information of metal spatter or seam deviation. Analysis the affection of keyhole centroid position changes in the direction of welding to the bead width in theory. The results showed that, essentially, the changes of concentrated area of the laser energy affect the bead width; and make a theoretical analysis to seam deviation detection with the interference of metal spatter. Fusion key features to establish regression model of the bead width prediction and seam deviation detection based on the image processing method, the accuracy has been further improved. Especially, the accuracy of deviation detection with the interference of metal spatter has been significantly improved. It provided a new idea to improve the accuracy of seam deviation with metal spatter by vision. Finally repeated experimental verification bead width prediction and deviation detection model, The results show that the model has a high accuracy in the same laser welding conditions.