Study On Model-Free Adaptive Control Of Weld Pool Dynamic Process In Pulsed GTAW

The fundamental difficulty of realizing pulsed GTAW automation and intellectualization is to detect and control the weld pool's dynamic behavior, such as weld pool's geometry parameters, weld penetration and weld beam forming. Arc welding is a complex process which involves interactions of materials, metallurgy and physical chemistry. Weld quality (weld beam forming, microstructures and properties of joint) is related to the multiple parameters of welding technology, these parameters are independent, coupled in the dynamic process and overlapped in the static process. Because welding process is a multi-variable, non-linear, time-varying process which also contains many uncertain factors and constraint conditions, it is very difficult to get an efficient control model based on the accurate mathematic modeling method, and this makes it very difficult to control the weld pool's dynamic changes, and to control the weld pool's width, penetration and weld beam forming. To solve the problem by using of classic and modern control theory will face the challenges both on the theory and application, and this makes it difficult to get an ideal result. This paper takes the pulsed GTAW as the research object, and does research on the control of welding process based on the passive visual sensor of weld pool, this paper focuses on the combination of the model-free adaptive control method and practical application of welding process, and applies the model-free adaptive control method in the penetration and weld beam forming control of pulsed GTAW. This method only needs the input and output data of GTAW process, it not only overcomes the difficulty of obtaining an accurate mathematic model of welding process, but also overcomes many uncertain factors of the welding process because it has stronger adaptive ability.In order to search the dynamic relation between pool characteristic parameters and welding parameters, the conventional step experiment were firstly carried out and the one-order inertia models of welding peak current, wire feeding speed and welding speed to geometry parameters of weld pool are obtained. The results show that arc welding is characterized as multi-variable, strong coupling, nonlinear, time varying. In order to predict the backside width and topside height of the weld pool and meet the requirement of control simulation, the BP (Back Propagation) neural network model and ARX (Auto-Regressive Exogeneous) model were built in a random design. These models can be used in control simulation and actual welding experiments.It's difficult to model welding process accuratly. To overcome this problem, SISO (Single Input Single Output) model-free adaptive controller of GTAW was designed, and the control variable was weld peak current, the controlled variable was backside width of weld pool. The closed loop control system simulation was carried out based on ARX model and BP neural network prediction model to prove the feasibility of this method. To verify the validity of this controller, three kinds of worpieces were designed to represent different heat emission conditions, which were trapezia-shaped workpiece, graded dumbbell-shaped workpiece and mutant dumbbell-shaped workpiece. The actual welding experiments of three different shaped workpieces were carried out, the results show that SISO model-free adaptive control can get better controlled performance.Enlightened by the intelligent behavior of welders, this paper combines the fuzzy control and model-free adaptive control algorithm base on the general form of model-free adaptive control, merges the welder's experience knowledge into the design of controller, and designs the fuzzy logic model-free adaptive controller. This controller can reflect the welder's operating experiences to some extent. Compared with the simulation of the basic model-free adaptive control method, it can be shown that the overshoot of the model-free adaptive control method with fuzzy adjusting function is smaller and the response curve is more smoothing. The welding experiments on the three work-pieces of different shape prove the effectiveness of this method.Besides influencing of uncontrolled variables and external, GTAW process is directly affected by many controlled welding parameters. In order to improve weld quality and reach high efficient automatic welding technology, MISO (Multi Input Multi Output) model-free adaptive control of GTAW were designed by analyzing the problem of variables selection. Simulation and welding experiments show that the welding shaping is uniform and can meet the needs of welding process. Compared with single variable model-free adaptive control of welding process, multi variables model-free adaptive control not only receive better control performance but also make the variation range of control variable smaller.Based on the theory of MISO model-free adaptive control, MIMO (Many Input Many Output) model-free adaptive control theory, design methods and implementation steps are presented. Then MIMO model-free adaptive control of GTAW was designed, in which the control variables were welding peak current and the wire feeding speed, the controlled variables were backside width of weld pool and topside height of weld pool. Simulation and experiments results show that when the heat emission condition of worpieces changes, control variables can make controlled variables to arrive at desire values fastly. And because of choosing topside height of weld pool as another controlled variable, not only fusion penetration can be ensured but also topside appearance of weld quality can be ensured, which is important in practical productions.