Authored by Mahyar Asadi, Mohammad Mohseni, Majid Tanbakeui Kashani, Michael Fernandez, and Mathew Smith; and presented at the NAFEMS NORDIC Virtual Seminar.
Physics-based digital twins offer a critical core-competency that enables the smarting through limited data. The current data-driven digital twins that use machine learning take a significant initial data set to mature for making decisions. In most manufacturing processes, there is no such data set to draw on. The majority of SMEs need cyber-manufacturing systems that work with limited data and physics based digital twins. However, physics-based digital twins that are entirely built on deploying simulation tools such as finite element analysis (FEA) cannot be responsive enough for real-world applications. A solution is to use machine learning (ML) algorithms that emulate the time-consuming FEA-solver behaviour at a much shorter time. We build a hybrid physics-based digital-twin that takes advantage of data-driven digital twins for quick response while gaining fidelity through adaptive learning with FEA simulation tools. We use our hybrid digital-twin to explore various weld sequences in real-time to form a platform for smart welding fabrication. This tool enables engineers to analyze and compare different patterns to assess fabrication scenarios without computational time delay.
Read the complete paper here.