Seam Tracking Monitoring Based on Adaptive Kalman Filter Embedded Elman Neural Network During High-Power Fiber Laser Welding

This paper proposes a method of seam tracking monitoring during high-power fiber laser welding. A visual sensor system was employed to capture the infrared images of molten pools and the surroundings in the laser welding process. A weld seam position variable was extracted by the image difference and centroid algorithms. The state and measurement equations for weld seam position were established based on an eigenvector derived from the weld seam position variable. A Sage-Husa adaptive Kalman filter (AKF), as an estimator of the noise statistical characteristics, was applied in order to enhance the filtering precision. By embedding an Elman neural network into the AKF, an error estimator was used to compensate for the filtering errors. The results of the welding experiments have demonstrated the effectiveness of the proposed method to improve the accuracy of weld detection.