Abstract:This study introduces a generative imputation model leveraging graph attention networks and tabular diffusion models for completing missing parametric data in engineering designs. This model functions as an AI design co-pilot, providing multiple design options for incomplete designs, which we demonstrate using the bicycle design CAD dataset. Through comparative evaluations, we demonstrate that our model significantly outperforms existing classical methods, such as MissForest, hotDeck, PPCA, and tabular generative method TabCSDI in both the accuracy and diversity of imputation options. Generative modeling also enables a broader exploration of design possibilities, thereby enhancing design decision-making by allowing engineers to explore a variety of design completions. The graph model combines GNNs with the structural information contained in assembly graphs, enabling the model to understand and predict the complex interdependencies between different design parameters. The graph model helps accurately capture and impute complex parametric interdependencies from an assembly graph, which is key for design problems. By learning from an existing dataset of designs, the imputation capability allows the model to act as an intelligent assistant that autocompletes CAD designs based on user-defined partial parametric design, effectively bridging the gap between ideation and realization. The proposed work provides a pathway to not only facilitate informed design decisions but also promote creative exploration in design.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of missing parameter data in engineering design. Specifically, the author introduced a new generative imputation model, combining Graph Attention Networks (GATs) and Tabular Diffusion Models, to complete the missing parts of parameter data in engineering design. #### Main problem description: 1. **Limitations of existing methods**: - Traditional imputation methods (such as MissForest, hotDeck, PPCA, etc.) perform poorly when dealing with complex dependencies and multivariate interactions. - These methods usually oversimplify the complexity of design parameters and cannot fully capture the complex dependencies between different design parameters. 2. **Challenges in engineering design**: - In engineering design, the situation of partial parameter missing often occurs, which seriously affects the effectiveness and integrity of the design. - The missing data will prevent engineers from generating effective design schemes, especially when complex dependencies between multiple components need to be considered. 3. **Innovation points**: - **Combining Graph Neural Networks (GNNs) and diffusion models**: By using the structural information provided by assembly graphs, GNNs can better understand and predict the complex dependencies between different design parameters. - **Generative imputation task as a design recommendation system**: Re - envision the generative imputation task as a design recommendation system, where users can obtain multiple design completion suggestions based on incomplete design inputs. - **Enhancing design decision - making and exploration**: This model not only fills in the missing information, but also proposes design improvement suggestions based on a comprehensive understanding of the design space, thus promoting creative exploration in the design process. #### Solution overview: - **Model architecture**: An innovative diffusion model architecture was introduced, which explicitly combines the hierarchical structure and relationship complexity in the assembly graph, thereby filling in the missing parameter data more accurately. - **Performance evaluation**: By comparing and evaluating with classical imputation methods and other state - of - the - art diffusion models, the significant advantages of this model in terms of accuracy and diversity were demonstrated. - **Application example**: Taking the bicycle CAD design data set as an example, it was demonstrated how the model can act as an AI design co - pilot to automatically generate complete design schemes according to the partial parameter information defined by users. ### Summary This paper proposes a new generative imputation model by combining Graph Neural Networks and diffusion models, which solves the problem of missing parameter data in engineering design. This model not only improves the accuracy of the imputation task, but also provides diverse design suggestions, enhancing the flexibility and creativity of design decision - making.

Bridging Design Gaps: A Parametric Data Completion Approach With Graph Guided Diffusion Models

Generative Structural Design Integrating BIM and Diffusion Model

DesignFusion: Integrating Generative Models for Conceptual Design Enrichment

A Hybrid Prototype Method Combining Physical Models and Generative Artificial Intelligence to Support Creativity in Conceptual Design

Parametric-ControlNet: Multimodal Control in Foundation Models for Precise Engineering Design Synthesis

Towards Goal, Feasibility, and Diversity-Oriented Deep Generative Models in Design

Latent Diffusion Models for Structural Component Design

VehicleSDF: A 3D generative model for constrained engineering design via surrogate modeling

Vitruvion: A Generative Model of Parametric CAD Sketches

Data-Driven Design Space Exploration and Exploitation for Design for Additive Manufacturing

Bridging Model-Based Optimization and Generative Modeling via Conservative Fine-Tuning of Diffusion Models

A parametric BIM approach to foster bridge project design and analysis

Parametric: Empowering In-Situ Parametric Modeling in Augment Reality for Personal Fabrication

Design Space Exploration and Explanation via Conditional Variational Autoencoders in Meta-Model-Based Conceptual Design of Pedestrian Bridges

Interactive design space exploration and optimization for CAD models

Parametrische Modellierung und generatives tiefes Lernen für den Brückenentwurf

A comparative analysis of CAD modeling approaches for design solution space exploration

DiffDesign: Controllable Diffusion with Meta Prior for Efficient Interior Design Generation

Fairness- and uncertainty-aware data generation for data-driven design

Efficient Web-Based Data Imputation With Graph Model

Automated geometric digital twinning of bridges from segmented point clouds by parametric prototype models