Exploring the parametric design space to manage computational weld mechanics analyses using design of experiment

Abstract

Development of a computational weld mechanics (CWM) framework that automates multiple setups and evaluations is required to practically explore a design space by given design of experiment (DOE) matrices. Saving an expert-user's time to prepare several analyses and allocating CPUs to be utilized efficiently make this framework cost effective and time effective to manage designer-driven optimization and control application of CWM. A validation analysis is conducted in this framework to identify the CWM control vector that minimizes the difference between the computed and experimental data. Actual CWM problems with continuous and/or discontinuous parametric design spaces are solved in this framework to minimize weld distortion using derivative-free optimization algorithms and DOE matrices that become attractive in this framework.

Welding Distortion Can be Mitigated if Welding Current and Traveling Speed Vary Optimized Along a Weld Path

“…They have collected weld optimization applications including Artificial Neural Networks (ANNs), Genetic Algorithm (GA), Response Surface Method (RSM), and Design of Experiment (DOE) to develop a mathematical relationship between the welding process parameters and the objective function(s) of the weld joint ranging from weld-bead geometry to mechanical properties. Success in this category significantly depends on having a reliable predictive model that feasibly and quickly explores large set of design parameters for implementing the modern algorithms of decision-making [12].…”