Intelligent Control Of Backside Weld Pool Width During Pulse GMAW Using Active And Passive Composite Sensing Technology

Automatic welding of thin plate is a difficult point in automatic welding,and the key is to control the penetration.Existing automatic welding of thin plate mainly uses gas tungsten arc welding(GTAW)or pulse gas metal arc welding(P-GMAW).Compared with GTAW,P-GMAW has the advantages of high efficiency,strong adaptability and controllable heat input.So P-GMAW has more extensive application and more research significance in automatic welding of thin plate.The existing welding process control researches mainly focus on GTAW while relatively few researches on P-GMAW.The main reason is that the welding process of P-GMAW is more complex than that of GTAW,and the sensing and control difficulty is much greater than that of GTAW.In this paper,automatic welding of thin plate with P-GMAW is taken as the research object.In order to realize control of backside weld pool width,the sensing technology of front weld pool,influencing factors of backside weld pool width,prediction modeling of backside weld pool width and intelligent controller design of backside weld pool width are studied in depth.Firstly,a set of active and passive composite visual sensing system for welding process is designed by considering the characteristics of arc light intensity and high temperature blackbody radiation,which realized the real-time online acquisition of weld pool and weld seam images.In order to realize intelligent control of P-GMAW backside weld pool width,a welding source with its control system and a robot with its control system are established respectively.In order to study the effect of welding process parameters on weld forming,the penetration s under different welding voltages,welding currents and welding speeds are analyzed,and the optimal welding process parameters are determined.On this basis,the linear models between welding process parameters and weld bead sizes are established through quadratic regression analysis,which lays a foundation for the control of welding process.In order to extract geometrical information of weld pool and weld bead from the obtained images real-time accurately,image processing techniques such as Otsu adaptive threshold segmentation,gray analysis and edge detection are used to develop image processing software through the analysis of weld pool and weld bead characteristics in P-GMAW process.Then real-time online detection of weld pool and weld bead geometrical information are realized.The relationship between image coordinate and real spatial coordinate is established through the three-dimensional calibration of the active and passive composite visual sensing system.A multistep prediction model of weld bead height is established by BP neural network algorithm,and the height of weld pool tail is calculated.By studying and analyzing the shape characteristics of solidified weld pool,a method of space surface fitting is proposed to fuse the geometric information of weld pool and weld bead to estimate and reconstruct the three-dimensional surface of weld pool.In order to solve the difficulty of backside weld pool width real-time online detection,influence factors of backside weld pool width are analyzed according to the conservation of mass deposited metal and the force balance characteristics of quasi-steady weld pool.On this basis,classical linear model and neural network model are used to establish the prediction model of the backside weld pool width,which provides a reliable basis for the real-time online control of the backside weld pool width.Finally,an actor-critic reinforcement learning adaptive controller based on RBF neural network is proposed to control the backside weld pool width.In order to verify the reliability of the controller,simulation experiments with nonlinear model,GTAW model and P-GMAW model are carried out.In order to verify the advantages of the actor-critic reinforcement learning adaptive controller,the classical PID controller and the single-neuron self-learning PSD controller are used as the comparison to control the front and backside weld pool width during the welding process.In order to verify that the designed actor-critic reinforcement learning adaptive controller’s control effect,resistance to welding parameters interference and ability to adapt to changing heat dissipation conditions in P-GMAW backside weld pool width control,backside weld pool width and welding speed are considered as controlled variable and control variable respectively,experiment with target value steps,welding speed disturbs,variable heat dissipation and dummy bell workpieces butt welding are carried out.The experimental results show that the controller has a good control effect on the backside weld pool width and meets the actual welding requirements well.