Modeling and strategies for arc welding penetration control algorithm: a review

In automated arc welding, real-time control of welding parameters is crucial for achieving consistent weld penetration and width, which are key to maintaining weld quality. However, the welding process is inherently complex due to its nonlinear dynamics, time delays, thermal inertia, and instability. Effective real-time control typically relies on system identification methods, including the use of step inputs and pseudo-random signals to excite the process and capture system behavior. In recent years, a wide range of modeling and control approaches have been proposed to address the challenges of weld penetration control, including linear models, nonlinear adaptive systems, intelligent algorithms, and hybrid strategies. This review systematically classifies and evaluates modeling techniques such as MA/ARMA, NARMAX, Hammerstein, rough-fuzzy, and neural network-based models. Furthermore, we survey both conventional and advanced control strategies—including PID, fuzzy logic, model predictive control, active disturbance rejection control, iterative learning control, and model-free adaptive control (MFAC)—highlighting their applicability and effectiveness in different welding scenarios. For real-time penetration depth control, MFAC is particularly recommended due to its unique capability to handle time-varying dynamics without requiring precise mathematical models, enabling robust performance in complex welding environments with minimal computational overhead. This paper aims to provide a structured overview of the state-of-the-art in weld penetration control, offering insights for future research and industrial applications.