Study On Simultaneous Monitoring Of Penetration And Seam Tracking During Laser Deep Penetration Welding

With the improvement of machining automation in the automotive industry, laser welding which has advantages of high speed, little deformation of workpiece, little spatter and easy control has become one of the main welding methods in automobile manufacturing industry. However, due to the small size of the laser spot and the welding deformation of the workpiece, laser spot is easy to deviate from the actual welding seam. Welding quality is difficult to maintain long-term stability. Monitoring the laser welding process by machine vision method had been attracting attention of many scholars.In order to realize simultaneous monitoring of penetration status and welding seam, a coaxial monitoring system was built for fiber laser welding. Keyhole and weld seam image were collected with 532 nm and 808 nm narrowband pass filter. Then, the 808 nm laser illuminant was used to increase the light intensity of the weld zone. Results indicate that detailed images of keyhole, full penetration hole and weld seam can be taken with 808 nm narrowband pass filter under 808 nm laser auxiliary illuminant with the same hard ware under strong plasma interference.In this study, coaxial images of keyhole, full penetration hole and weld seam were taken under different welding parameters. According to the image processing algorithm, the corresponding target information is extracted. Image processing of penetration status monitoring includes selected region of interest, image noise reduction, threshold segmentation and morphological image processing. After image processing, edge of keyhole and full penetration hole can be extracted exactly. The mechanical properties of the welds, the macro morphology of the welds and morphology of the welds section under different welding parameters were analyzed, so as to determine the appropriate penetration welding process parameters. Relationships between penetration status and area of full penetration hole were established, so that we can monitor the penetration status. Image processing of weld seam monitoring includes selected region of interest, image no ise reduction, threshold segmentation, morphological image processing and Hough transformation. After image processing, upper and lower edges of weld gap were extracted precisely. Median line of the upper and lower edges of the weld gap can act as line of the weld seam center. The deviation between the center of keyhole and weld seam center were calculated. Accordingly, weld gap monitoring and seam tracking monitoring were achieved.During image processing, keyhole and full penetration hole may be affected by the white spots whose image is formed by the reflected light. Thus it results in wrong extraction of keyhole and full penetration edge. So, we proposed tolerance mechanism. Limits of area of keyhole and the difference of slope of the weld seam center line is the criterion for judging the failure of extraction. If the extraction fails, all values of the last calculation will be assigned to this calculation. Then, this calculation will be interrupted and next calculation will be executed.Studies have shown that when the area of full penetration hole is between 4.09×10-4 mm2 and 1.52×10-2 mm2, the weldment is the appropriate penetration. In this study, weld seam tracking algorithm can recognize the weld gap as small as 0.02 mm, the seam tracking monitoring error is 0.043 mm. It takes about 35.5 ms to process a image, which can be applied to real-time practice. Therefore simultaneous monitoring of penetration status and welding seam were realized.The research results have significant theory and application value on real time monitoring and simulation of laser welding.