[DIVE 2024] Discovery Bot

LLM-based Generative Engineering Design Suggestion using Ansys Discovery

A chatbot-powered system that integrates LLMs (GPT-4) to assist users in understanding, analyzing, and improving engineering designs within Ansys Discovery.

Keywords — LLM-based Chatbot · Generative Design

Background

In the field of engineering design, many students and beginners face barriers when using simulation tools like Ansys Discovery. They often struggle to select appropriate materials, set boundary conditions, and interpret simulation results without expert guidance.

Despite the growing importance of simulation-based design, the steep learning curve and lack of intuitive support limit broader adoption. To address this, we propose a chatbot-powered system that integrates LLMs (such as GPT-4) to assist users in understanding, analyzing, and improving their designs.

Prototype

The prototype focused on intuitiveness and guidance.

We designed the chatbot interface to provide step-by-step assistance, especially in material selection and simulation interpretation, where users often feel lost. The UI was kept minimal and context-aware, helping users focus on decision-making without overwhelming technical details.

Key Functions & Demo

The core functionalities provided by Discovery bot are threefold. First, the chatbot initiates the conversation and offers multiple assistance options. Second, it analyzes simulation results and provides actionable feedback. Finally, it suggests alternative design options based on user feedback and performance requirements.

Function 1 — Starting the Conversation with Ansys Chatbot

The chatbot initiates the conversation by offering multiple assistance options, such as material suggestions, simulation guidance, and design review support.

Function 2 — Simulation Feedback Support

The chatbot analyzes simulation results and provides design feedback, such as material alternatives or structural improvements.

Function 3 — AI-Guided Design Alternatives

The chatbot suggests alternative design options by analyzing the user's feedback and performance requirements. Visual candidates are presented to help the user explore more ergonomic or aesthetically diverse forms.

System Flow

The system combines a GPT-4-powered analysis chatbot with a GAN-based generative design pipeline.

Part 1 — Analysis Support (GPT-4 Chatbot)

1. Request clarification on analysis results

2. Collect specific requirements

3. Identify areas for improvement and suggest follow-up questions

Part 2 — Generative Design Pipeline (GAN)

• STEP 1 — Modeling

• STEP 2 — Feasibility Check

• STEP 3 — Evaluation

• STEP 4 — Decision Making

Future Agenda

The system can be extended to cover the full design cycle — from AI-assisted shape generation and material selection to structural evaluation and topology optimization within Ansys Discovery.

Roadmap

1. Design Generation

2. Design Evaluation

3. Design Optimization

4. Design Recommendation

Links

• 💻 GitHub: HACK-DIVE

• 📺 YouTube: Demo Video

• 📰 News: 국제뉴스 기사