Investigating the feasibility of a machine learning-based framework for welding design****

点击翻译按钮获取中文摘要

SKC Engineering is a renowned Canadian provider for welding engineering services with over 250 clients throughout Western Canada. For many clients, the company has observed that the weld distortion is a common problem. A usual approach is to either optimize welding by relying on the experience of welding engineers, which is difficult to reproduce, or use simulation algorithms to capture and couple physics of welds based on 3D structural and thermal stress simulations. Examining all possible simulation scenarios is currently practically infeasible, as this requires choosing from potentially thousands of permutations. **In this project, in collaboration with Dr. Abolmaesumi at the University of British Columbia, we will examine the feasibility of using a machine learning framework to learn from simulations and create case-specific real-time models of real-world welding processes. Knowledge gained in this project will be potentially a breakthrough in the use of machine learning in accelerating finite element modelling and prediction of simulation outcomes.**