| Observing and measuring the weld pool geometry in the gas metal arc welding (GMAW) process is the key to understanding and control of the complex process, and to improving and validating welding process models. Although different techniques have been applied to monitor the weld pool profile in GMAW, the harsh environment and highly dynamic characteristic of the weld pool surface makes the observation extremely challenging.;In this dissertation, a novel machine vision system to measure the highly dynamic three-dimensional GMAW weld pool surface geometry is proposed. A five-line-by-five-line grid pattern is projected onto the weld pool surface. The projected laser specularly reflects from the weld pool surface. The reflection is intercepted by an imaging plane, and results in a deformed grid pattern. A high-speed camera fitted with a narrow bandpass filter is adopted to capture the reflected pattern. The weld pool surface geometry is reconstructed based on the captured reflected pattern.;The weld pool surface geometry in GMAW is highly dynamic due to periodic droplet impingement into the existing molten metal, and strong arc force applied onto the weld pool surface. Consequently, the reflected pattern is not of regular shape at all times. To reconstruct the weld pool surface, an image processing algorithm is developed to first evaluate the captured images, segment and identify the reflection pattern from series of captured images, and establish line-to-line correspondence. A surface normal field-based reconstruction method is presented to calculate the three-dimensional weld pool geometry based on established corresponding relationships. The experiments conducted using simulation data and test target with known dimension proved the effectiveness of the measurement system.;The achievement of this dissertation provides a fundamental understanding for the dynamic of the weld pool surface in GMAW processes.;KEYWORDS: Weld Pool Surface, Machine Vision, Image Processing, Measurement, Welding Sensor. |
