A Framework to Optimize Laser Welding Process by Machine Learning in a SME Environment

Abstract

The Metallurgical industry, and in particular the make-to-order industry, are extremely dependent on the composition of the raw material. As a subcontractor, the company is limited by customer constraints and must be able to adapt the process as fast as possible. Therefore, machine programs need to be adapted to control material reactions (i.e., deformations, vibrations, burrs, porosity, etc.) following the laser cutting and welding. Material reactions depend on the gap between theoretical and intrinsic mechanical characteristics of the materials, on the geometry and on the tools used to maintain the parts. The variability of this gap leads to scrap, machine delay and productivity loss. To improve our knowledge, Industry 4.0 provides several technologies enabling the automation of complex tasks. For instance, Artificial Intelligence (IA), supervised or unsupervised, provides opportunities for optimizing and customizing machine programs considering material reactions. The case study presented in this paper proposes to use machine learning (ML) to define a decision support system. Two ways are explored: quality and productivity optimization. The objective is to free up resources (operators, machines) time for other strategic projects for the company.