TIG Welding Skill Extraction using a Machine Learning Algorithm
Date
2017-12
Authors
Huff, Shelby A.
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Tungsten Inert Gas (TIG) welding is the superior arc welding process used in the manufacturing industry for high quality welds. Skilled welders are capable of monitoring a weld bead and dynamically adjusting the welding parameters (current, voltage, speed, etc.) to produce a desired weld bead quality (width, height, depth). A shortage of skilled workers and motivation for industrial automation has increased research in welding process control and optimization. We propose a Machine Learning algorithm to model the TIG welding process and extract human skill. First, an automated TIG welding system uses an industrial robot to conduct aluminum welding experiments. Controlled process parameters and resultant weld bead quality measurements are used to form a welding process dataset. A Gaussian Process Regression (GPR) algorithm is applied to model the relationship in the dataset inputs variables and output variables. For a desired weld bead thickness, the required adjustment in welding current, or welding skill, can be predicted to robustly control the process. The addition of artificial intelligence to industrial robots can solve many automation solutions in the manufacturing industry dealing with complex processes.
Description
Keywords
TIG welding, Artificial intelligence, Machine learning, Gaussian process regression, Industrial robots, Manufacturing engineering
Citation
Huff, S. (2017). <i>TIG welding skill extraction using a machine learning algorithm</i> (Unpublished thesis). Texas State University, San Marcos, Texas.