Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Generative Model for Mechanical System Configuration Design

Yasaman Etesam, Hyunmin Cheong, Mohammadmehdi Ataei, Pradeep Kumar Jayaraman

TL;DR

This work tackles mechanical configuration design, specifically gear-train synthesis, by introducing GearFormer, a Transformer-based generator that produces valid component-and-interface sequences constrained by a domain-specific language and evaluated via a physics simulator. The approach blends end-to-end generation with hybrid search methods (EDA-Transformer, MCTS-Transformer) to improve solution quality and speed, and it introduces the GearFormer dataset and a DSL with a formal grammar and lexicon. Key contributions include a scalable data-driven framework for rapid, constraint-satisfying designs, a differentiable weight objective via Gumbel-Softmax, and demonstration of interactive workflows enabling engineers to explore multiple design iterations in real time. The methodology has practical impact for accelerating engineering design blocks, with potential extension to other mechanical domains through analogous DSLs and physics simulators.

Abstract

Generative AI has made remarkable progress in addressing various design challenges. One prominent area where generative AI could bring significant value is in engineering design. In particular, selecting an optimal set of components and their interfaces to create a mechanical system that meets design requirements is one of the most challenging and time-consuming tasks for engineers. This configuration design task is inherently challenging due to its categorical nature, multiple design requirements a solution must satisfy, and the reliance on physics simulations for evaluating potential solutions. These characteristics entail solving a combinatorial optimization problem with multiple constraints involving black-box functions. To address this challenge, we propose a deep generative model to predict the optimal combination of components and interfaces for a given design problem. To demonstrate our approach, we solve a gear train synthesis problem by first creating a synthetic dataset using a grammar, a parts catalogue, and a physics simulator. We then train a Transformer using this dataset, named GearFormer, which can not only generate quality solutions on its own, but also augment search methods such as an evolutionary algorithm and Monte Carlo tree search. We show that GearFormer outperforms such search methods on their own in terms of satisfying the specified design requirements with orders of magnitude faster generation time. Additionally, we showcase the benefit of hybrid methods that leverage both GearFormer and search methods, which further improve the quality of the solutions.

Deep Generative Model for Mechanical System Configuration Design

TL;DR

This work tackles mechanical configuration design, specifically gear-train synthesis, by introducing GearFormer, a Transformer-based generator that produces valid component-and-interface sequences constrained by a domain-specific language and evaluated via a physics simulator. The approach blends end-to-end generation with hybrid search methods (EDA-Transformer, MCTS-Transformer) to improve solution quality and speed, and it introduces the GearFormer dataset and a DSL with a formal grammar and lexicon. Key contributions include a scalable data-driven framework for rapid, constraint-satisfying designs, a differentiable weight objective via Gumbel-Softmax, and demonstration of interactive workflows enabling engineers to explore multiple design iterations in real time. The methodology has practical impact for accelerating engineering design blocks, with potential extension to other mechanical domains through analogous DSLs and physics simulators.

Abstract

Generative AI has made remarkable progress in addressing various design challenges. One prominent area where generative AI could bring significant value is in engineering design. In particular, selecting an optimal set of components and their interfaces to create a mechanical system that meets design requirements is one of the most challenging and time-consuming tasks for engineers. This configuration design task is inherently challenging due to its categorical nature, multiple design requirements a solution must satisfy, and the reliance on physics simulations for evaluating potential solutions. These characteristics entail solving a combinatorial optimization problem with multiple constraints involving black-box functions. To address this challenge, we propose a deep generative model to predict the optimal combination of components and interfaces for a given design problem. To demonstrate our approach, we solve a gear train synthesis problem by first creating a synthetic dataset using a grammar, a parts catalogue, and a physics simulator. We then train a Transformer using this dataset, named GearFormer, which can not only generate quality solutions on its own, but also augment search methods such as an evolutionary algorithm and Monte Carlo tree search. We show that GearFormer outperforms such search methods on their own in terms of satisfying the specified design requirements with orders of magnitude faster generation time. Additionally, we showcase the benefit of hybrid methods that leverage both GearFormer and search methods, which further improve the quality of the solutions.
Paper Structure (32 sections, 7 equations, 5 figures, 3 tables)

This paper contains 32 sections, 7 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem
  4. Configuration Design of Mechanical Systems
  5. Gear Train Synthesis Problem
  6. Proposed Approach
  7. Transformer-based Model
  8. Hybrid Methods
  9. EDA+Transformer
  10. MCTS+Transformer
  11. Domain-Specific Language
  12. Grammar
  13. Lexicon
  14. Dataset Generation
  15. Generate variable sequences
  16. ...and 17 more sections

Figures (5)

  • Figure 1: GearFormer is based on the Transformer architecture. It includes an encoder module that processes the input design requirements that have been embedded by a multi-layer perceptron (MLP). These input embeddings are consumed by the Transformer decoder module via cross-attention to generate the gear train sequence that satisfies the requirements.
  • Figure 2: A hybrid method that combines MCTS to explore the first few critical tokens in a sequence and a Transformer-based model to complete the sequence for evaluation.
  • Figure 3: Gear train designs generated with GearFormer
  • Figure 4: Examples of gear pair placements for each meshing token. (a) Spur gears. (b) Miter/Bevel gears. (c) Rack-and-pinion and worm gears. (d) Hypoid gears.
  • Figure 5: As we increase $w_1$ to prioritize the weight loss term, the average weight of the generated sequences drops at the expense of producing fewer valid and feasible sequences.