Robotic Welding Path Identification Using FPGA-Based Image Processing

Abstract

The robotic welding process is widely used in many industry sectors, and its use in production lines is becoming more common day by day. Obtaining a smooth weld seam in robot welding depends on the geometric structure of the welding path and the stability of the control loop. However, the weld path and the weld gap are usually not fixed, and their change negatively affects the automatic control. Programming the complex welding path by the operator may take more time than executing the task for some welding jobs. In addition, the variable weld gap negatively affects the weld quality in the constant control loop. This study proposes a system that provides a real-time definition of the weld path and its geometry on the embedded system to address this issue. The weld path image is captured using a camera, and the weld path is determined by image processing techniques using the embedded Linux operating system running on system-on-chip (SoC) hardware. The images captured through the Hard Processor System (HPS) unit are stored in memory, processed in the FPGA unit, and output by the HPS unit. Unprocessed SoC images and measurement images of weld pieces are presented with their values. When the values obtained from the processed weld path image are compared to manually measured path values, it is seen that the proposed system produces successful results.