High-power CO2 laser welding has been widely used in the industry because of its high productivity and excellent weld quality. In order to tap the potential of this process completely, it is important to have on-line weld quality inspection methods to improve the process productivity and reliability by achieving 100 percent weld inspection. Weld penetration is one of the most important factors critical to the quality of a laser weld. However, it is very difficult to directly measure the extent of penetration without sectioning the workpiece. In this paper a model-based penetration depth estimation technique suitable for the production environment is developed. The proposed model relates the temperature measured on the bottom surface of the workpiece, weld bead width, laser beam power and welding speed to penetration depth. The closed-loop depth estimator combines the model and a model-error compensator to compensate for the uncertainty in the measurement of the laser power and absorptivity. Other effects considered are the averaging due to the finite size of the sensor, delay based on the sensor location and the process and sensor dynamics. Several bead-on-plate and butt welds were made on low carbon steel plates to validate the static process models and the depth estimation scheme. Temperatures on the bottom surface of the workpiece during welding were measured using infrared thermocouples. The welds were sectioned longitudinally to obtain the penetration profile. The penetration profiles estimated by the depth estimator matched satisfactorily with the measured penetration profiles. The results validate the capability of the proposed depth estimator to estimate penetration depth and its ability to trace the dynamic changes in penetration depth.
