| Due to the influence of accidental factors,there may be many problems such as the sudden change of weld backside width or the lack of penetration at the back during the welding process,which will seriously affect the welding quality.The subsequent repair will also reduce the work efficiency.Therefore,in the process of welding,according to the characteristics of welding parameters and the front size of the pool,the real-time prediction of the backside weld width and the dynamic adjustment of the welding process will play an important role in improving the welding quality and realizing the intelligent welding.In this paper,the prediction and control of back weld width are studied from the aspects of weld pool image information processing and feature extraction,back weld width prediction,weld pool fuzzy controller,etc.the research contents and results are as follows:(1)Based on the image processing algorithm,the molten pool feature extraction system is designed.The system includes: noise filtering,image enhancement,threshold segmentation,mathematical morphology,edge detection,edge curve fitting,minimum circumscribed rectangle,feature definition,scale conversion and other functions,which can greatly reduce image noise,increase the contrast of molten pool boundary area,detect clear and complete molten pool shape and quickly extract the basic features of molten pool,which has the certain application value for project.(2)SVM(support vector machine)and GBDT(gradient boosting decision tree)algorithm are used to model the weld pool and predict the backside weld width.Model training data acquisition includes experiment method,ANSYS simulation method.The establishment of welding database to achieve the rapid retrieval of welding data.It is verified that the integrated learning model GBDT converges stably in the training stage,has better fitting effect,and has smaller prediction error in the unknown data set,which can effectively predict the backside weld width of the molten pool.(3)The fuzzy controller is designed to control the welding process parameters.Taking the backside melt width error,error change as the input and voltage adjustment as the output,the fuzzy design,fuzzy control rule table design and fuzzy control design are carried out respectively,which can realize the real-time adjustment of welding process parameters according to the melt width error. |