AbqPilot: AI-Driven Abaqus Simulation Automation Agent

This article introduces the BrunelXian/AbqPilot project, a closed-loop CAE-ODB agent that converts user goals into executable, diagnosable, and repairable Abaqus simulation workflows. It aims to address pain points in traditional engineering simulations and achieve automation and intelligence in the simulation process. The project is maintained by BrunelXian and was released on the GitHub platform on June 16, 2026 (link: https://github.com/BrunelXian/AbqPilot).

In aerospace, automotive, civil engineering, and other fields, finite element analysis (FEA) is a key design verification method. However, software like Abaqus has issues such as a steep learning curve and cumbersome processes:

• Tedious model preparation: Requires a lot of manual operations from CAD to finite element models
• Complex parameter settings: Needs professional knowledge for material properties, boundary conditions, etc.
• Difficult error troubleshooting: Time-consuming to diagnose the root cause when simulation fails
• Low iteration efficiency: Need to repeat preprocessing steps after design changes AbqPilot attempts to turn these pain points into automation opportunities through AI agent technology.

Core Architecture: Closed-Loop CAE-ODB

AbqPilot adopts a closed-loop design with the following process: Input user goals and basic models → Compile into Abaqus commands → Execute simulation → Diagnose ODB results and logs → Automatically repair and adjust, then iterate until the goal is achieved.

Agent Capabilities

• Intent understanding and task decomposition: Break down natural language goals into atomic operations
• Model rationality check: Verify geometry, mesh, materials, boundary conditions, etc.
• Fault diagnosis and repair: Analyze error logs (e.g., contact convergence, element distortion) and adjust automatically

Technical Implementation

• Based on Abaqus Python API: Operate Session, Model, Job objects and parse ODB
• LLM role: Natural language to code conversion, error analysis, result summary generation (using RAG, chain of thought, code verification to improve reliability)
• Workflow orchestration: State machine management, persistent storage, version control, parallel exploration

Application scenarios of AbqPilot include:

• Design optimization automation: Set goals (e.g., minimize weight) and constraints, automatically explore optimal solutions
• Batch simulation processing: Automatically generate variant models and perform parallel computing for parameter scanning, sensitivity analysis, etc.
• Knowledge inheritance and training: Encode the experience of senior engineers to help new employees learn quickly
• Cloud-native simulation: Combine containerization and cloud computing to achieve elastic resource scheduling and cost optimization

AbqPilot represents the direction of integration between AI and CAE, and can be extended to platforms like ANSYS and COMSOL to form a general simulation agent ecosystem. In the long run, it will bring:

1. Simulation democratization: Lower technical barriers to benefit non-professional users
2. Knowledge automation: Convert experts' tacit knowledge into reusable algorithms
3. Real-time design feedback: Integrate with CAD environments to achieve a closed loop of design-simulation-optimization

Domain Knowledge Accuracy

• Limit operation scope (whitelist)
• Manual confirmation for key operations
• Establish result credibility scoring

Computing Resource Management

• Estimate costs to avoid waste
• Monitor status to terminate abnormal tasks
• Queue management to avoid resource contention

Error Recovery Strategy

• Distinguish between automatically repairable and manually intervenable issues
• Provide diagnostic reports and repair suggestions
• Support human-machine collaboration mode