Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

When AI Bends Metal: AI-Assisted Optimization of Design Parameters in Sheet Metal Forming

Ahmad Tarraf, Koutaiba Kassem-Manthey, Seyed Ali Mohammadi, Philipp Martin, Lukas Moj, Semih Burak, Enju Park, Christian Terboven, Felix Wolf

TL;DR

The paper tackles the costly problem of optimizing sheet metal forming parameters by introducing an AI-driven workflow that combines a deep-learning initial guess with Bayesian optimization using a Gaussian Process Latent Variable Model to efficiently explore the design space. It further enhances the approach with a Mixture of Experts, a geometry-aware encoder, and an early-termination mechanism to handle new parts and reduce compute time and energy. The framework supports both fully automated optimization and human-guided active learning, showing substantial speedups, energy savings, and faster time-to-solution on OpenForm deep-drawing scenarios. These contributions enable more sustainable, scalable, and accessible design-space exploration for industrial sheet metal forming.

Abstract

Numerical simulations have revolutionized the industrial design process by reducing prototyping costs, design iterations, and enabling product engineers to explore the design space more efficiently. However, the growing scale of simulations demands substantial expert knowledge, computational resources, and time. A key challenge is identifying input parameters that yield optimal results, as iterative simulations are costly and can have a large environmental impact. This paper presents an AI-assisted workflow that reduces expert involvement in parameter optimization through the use of Bayesian optimization. Furthermore, we present an active learning variant of the approach, assisting the expert if desired. A deep learning model provides an initial parameter estimate, from which the optimization cycle iteratively refines the design until a termination condition (e.g., energy budget or iteration limit) is met. We demonstrate our approach, based on a sheet metal forming process, and show how it enables us to accelerate the exploration of the design space while reducing the need for expert involvement.

When AI Bends Metal: AI-Assisted Optimization of Design Parameters in Sheet Metal Forming

TL;DR

The paper tackles the costly problem of optimizing sheet metal forming parameters by introducing an AI-driven workflow that combines a deep-learning initial guess with Bayesian optimization using a Gaussian Process Latent Variable Model to efficiently explore the design space. It further enhances the approach with a Mixture of Experts, a geometry-aware encoder, and an early-termination mechanism to handle new parts and reduce compute time and energy. The framework supports both fully automated optimization and human-guided active learning, showing substantial speedups, energy savings, and faster time-to-solution on OpenForm deep-drawing scenarios. These contributions enable more sustainable, scalable, and accessible design-space exploration for industrial sheet metal forming.

Abstract

Numerical simulations have revolutionized the industrial design process by reducing prototyping costs, design iterations, and enabling product engineers to explore the design space more efficiently. However, the growing scale of simulations demands substantial expert knowledge, computational resources, and time. A key challenge is identifying input parameters that yield optimal results, as iterative simulations are costly and can have a large environmental impact. This paper presents an AI-assisted workflow that reduces expert involvement in parameter optimization through the use of Bayesian optimization. Furthermore, we present an active learning variant of the approach, assisting the expert if desired. A deep learning model provides an initial parameter estimate, from which the optimization cycle iteratively refines the design until a termination condition (e.g., energy budget or iteration limit) is met. We demonstrate our approach, based on a sheet metal forming process, and show how it enables us to accelerate the exploration of the design space while reducing the need for expert involvement.

Paper Structure

This paper contains 27 sections, 9 equations, 24 figures, 1 table.

Table of Contents

  1. Introduction
  2. Sheet Metal Forming
  3. Deep Drawing
  4. Control Parameters
  5. Simulation Software OpenForm
  6. Deep learning Component
  7. Deep Learning Parameters
  8. Deep Learning Model
  9. Optimization Loop of the Workflow
  10. Surrogate Model
  11. Input Space
  12. Selecting the Next Sample
  13. Parallel Samples
  14. Human Guided Active Learning
  15. Adapting the Design
  16. ...and 12 more sections

Figures (24)

  • Figure 1: Forming of a cylindrical cup.
  • Figure 2: Traditional Workflow extensively relying on an expert to specify promising input design configurations for the simulation to obtain the acceptable target parameters.
  • Figure 3: Improved workflow utilizing AI methods to reduce the expert involvement, opposed to Figure \ref{['fig:old_workflow']}.
  • Figure 4: Deep-drawing process (left) and tools (right).
  • Figure 5: Input parameters for the numerical simulation.
  • ...and 19 more figures