Analysis Of Droplet Behavior In Ultrasonic MIG Welding Based On Vision Sensing

Firstly,according to the characteristics of visual signal in the process of ultrasonic MIG hybrid welding,the accompanying white smoke and spatter will affect the acquisition of weld laser stripe and droplet contour.The intelligent welding system is built independently by using high-speed camera and visual sensor,and the welding droplet transition images under different welding process parameters are collected clearly and in real time.Then the welding droplet image is preprocessed combined with Open CV function library.Firstly,the droplet image is processed by Gaussian smoothing filter,and then the image is inverted.The optimal threshold can be determined automatically by Otsu threshold segmentation.Finally,Canny edge detection operator is used to obtain the contour of droplet image,so as to lay a good foundation for droplet feature extraction.The correlation mechanism betweeAt present,with the popularization and rapid development of industrial automation,it has driven the vigorous development of domestic manufacturing industry.Among the welding methods,the arc welding process is relatively stable,the weld forming effect is good,the microstructure and performance is excellent,the arc welding efficiency is high,the cost of welding equipment is low,and the automatic and intelligent process can be highly realized in the manufacturing industry,so it has been widely used.Taking the ultrasonic MIG composite welding process of Q235 galvanized steel sheet as the research object,this thesis analyzes and processes the droplet images collected in real time under different welding process parameters,preprocesses the collected images with Open CV,combines the visual feature information of typical welding defects,and designs a feasible extraction algorithm to extract the relevant features online,The correlation mechanism between image feature information and droplet transfer mode and typical welding defects is analyzed;Then,the visual features extracted from the welding droplet image are preprocessed,and the droplet counting algorithm model is established to realize the on-line monitoring of welding quality.nwelding droplet transfer mode,welding joint quality and image feature information is analyzed.Through the experimental verification and comparison of the actual welding results,it is found that under the welding low-power parameters,the droplet transfer mode is mainly short-circuit transfer.At this time,ultrasonic will hinder the droplet transfer,reduce the droplet contour area and increase the height from the molten pool.When the welding power increases,the droplet transfer modes under the ultrasonic assistance are mainly large droplet transfer and jet transfer.The geometric size of the droplet decreases,the height from the molten pool decreases,and the transition frequency to the molten pool increases.Finally,the initial acquisition image is defined as the background frame,and then the welding droplet image is obtained by background subtraction and image processing.Then a droplet counting algorithm is proposed,which effectively counts the number of droplets of galvanized steel plate under different welding methods,and obtains the relationship between droplet transition frequency and welding parameters.The experimental results show that the droplet transfer cycle can be up to 37 times in 1000 frames,which ensures the stability of quality in the welding process and accelerates the development of welding intelligence.