AUTHOREA
Log in Sign Up Browse Preprints
LOG IN SIGN UP
Stephan Schumacher
Stephan Schumacher
Consultant
Stephan Schumacher received his Bachelor of Science (B.Sc) in Business Administration and Engineering with a focus subject on business informatics in 2022 at the Technische Hochschule Lübeck, Germany. He is currently pursuing his Master of Science (M.Sc) degree at the Nordakademie in Hamburg while working as a Consultant for AI and software development at CGI.

Public Documents 1
Reinforcement Learning based Optimization of SMD Production
Stephan Schumacher

Stephan Schumacher

October 08, 2024
Optimization of production processes is a critical focus in manufacturing, with machine and deep learning approaches increasingly applied across various domains. This paper introduces a novel application of Deep Reinforcement Learning (DRFL) for optimizing the sequential ordering of products in Surface-Mounted Device production. The approach addresses the overlap of components in assemblies, thereby minimizing changeover times. A Deep Neural Network (DNN) is employed to estimate production times, and a secondary DNN, integrated with DRFL, determines the optimal sequence of product assembly. This setup models the problem as a variant of the Traveling Salesman Problem, where products are represented as nodes in a graph with unique properties. The production time model, derived from both simulation and a Regression Deep Learning Network, feeds into the DRFL agent for optimization. Experimental results demonstrate that DRFL significantly outperforms heuristic methods such as Nearest Neighbor, City Swap, and Simulated Annealing, achieving approximately 5% time savings over heuristic approaches and 30% over manual or random product ordering. Furthermore, DRFL shows superior computational efficiency, making it a viable solution for real-time production environments. This paper underscores the potential of DRFL to enhance production efficiency, offering substantial cost savings in manufacturing operations.

| Powered by Authorea.com

  • Home