Minimizing energy usage in multi-point manufacturing: a dimensional transfer learning strategy

Abstract

In this study, a minimum energy consumption scheduling strategy is proposed tailored to address Multi-Point Manufacturing (MPF) problems, such as drilling and spot welding tasks. Firstly, the MPF problem is formulated as a Multi-dimensional Weighted Traveling Salesman Problem (TSP), subsequently employing a Graph Neural Network (GNN)-based neural method for resolution. To the best of our knowledge, the proposed model is the first to utilize neural methodologies in scheduling MPF problems. During the neural network training phase, Transfer Learning is leveraged to utilize existing planning experience. A low-dimensional representation is first approximated using Self-Adjusting Multi-modal Artificial Neural Networks (SAMANN), and the entire model undergoes fine-tuning to enhance planner performance. It is found that the new proposed approach surpasses the performance of the current state-of-the-art (SOTA) neural model in terms of both solution quality and training time across multi-dimensional TSPs and MPF problem domains.