Research On Prediction Of Ultrasonic Assisted Laser Welding Forming Based On Characteristics Of Molten Pool And Plasma

Laser welding is an efficient welding method widely used in industrial manufacturing.In the laser welding process,metal vapor plasma,molten pool and keyhole are important features to describe the welding process,which can reflect the stability of welding at each moment and the welding quality status,and can be used for welding quality monitoring and control.The introduction of ultrasonic vibration can improve the porosity,refine the grain and increase the melting width.However,due to the instability of the welding process,there are differences in the coupling effect between the various parameters,and the welding process is Each feature is closely related.By analyzing the laser welding seam forming rule,dynamic characteristics and prediction of the weld seam under ultrasonic vibration,it can lay the foundation for guiding the ultrasonic assisted laser welding process and achieve the characteristics of the welding process.The quantity changes to provide a theoretical basis for monitoring welding quality.Therefore,for the 6mm thick AH36 marine steel,the image processing and spectrum diagnosis are used to extract the characteristic quantities of the molten pool,keyhole and metal vapor plasma.The BP neural network is used to establish the characteristic quantity as the input and the weld cross section size as the output prediction model to achieve the goal of predicting the weld cross-sectional size.Firstly,by building a high-speed camera monitoring system,for different subjects,choose the appropriate wavelength band filter and corresponding auxiliary light source,and adjust the high-speed camera different shooting angles to obtain the clearest image.At the same time,a spectrum acquisition system was built to collect the spectral characteristics of the metal vapor plasma.The system carried out an experimental study of ultrasonic assisted laser welding.Secondly,The effects of different process parameters on the molten pool,keyhole,plasma characteristics,weld penetration and weld width under ultrasonic conditions were studied.Then,through the relationship between the single process parameters and the weld size,the multiple regression equations of weld penetration and width and welding process parameters were established respectively,and the regression equations of ultrasonic assisted AH36 laser welding penetration and welding process parameters were obtained:H=0.14306+0.002097(F+0.5791)~2+0.77592P-0.37251v,the correlation coefficient reaches 0.86785 The empirical formula of melt width and various parameters is W=0.476+0.058F+0.548P-0.197v,and the correlation coefficient is 0.908.The results show that the regression equation is highly significant.Can be used to predict the cross-sectional size of the weld.Finally,considering the high coupling between the feature quantity and the weld cross-sectional dimension during the welding process and the nonlinear mapping law,the length of molten pool L,width of molten pool W,area of keyhole S1,area of plasma S2,height of plasma H And the average grayscale of the plasma G,these six image feature parameters,the spectral intensity Imn of the metal vapor plasma spectrum,the electron temperature T,and the electron density Ne.The three spectral feature parameters are used as input neurons of the neural network,and the weld cross section as the output of the neural network,the penetration and width of the size are established,and a BP neural network with a topology of 9-19-2 is established.And use the global search ability of genetic algorithm to optimize the neural network to improve its prediction accuracy and generalization ability.