Automating Structural Analysis Across Multiple Software Platforms Using Large Language Models

Researchers developed a multi-agent LLM system that automates structural analysis workflows across multiple finite element analysis (FEA) platforms including ETABS, SAP2000, and OpenSees. Using a two-stage architecture that interprets engineering specifications and translates them into platform-specific code, the system achieved over 90% accuracy in 20 representative frame problems, addressing a critical gap in practical AI-assisted engineering deployment.

This research represents a meaningful advancement in applying large language models to domain-specific engineering tasks. While prior work demonstrated LLMs could operate individual FEA platforms, real-world structural engineers depend on multiple tools based on project requirements and organizational constraints. The two-stage multi-agent approach elegantly solves this multi-platform challenge: Stage 1 agents collaboratively parse natural language input and extract geometric, material, boundary, and load parameters into a unified JSON schema, while Stage 2 translation agents convert this intermediate representation into executable syntax for target platforms.

The architecture reflects growing sophistication in prompt engineering and agent design. By isolating domain reasoning from code generation and enabling parallel translation agents, the system becomes extensible to additional FEA platforms without retraining. The 90%+ accuracy threshold across all tested cases demonstrates consistent reliability, a critical requirement for engineering applications where modeling errors propagate through analysis results.

For structural engineering firms, this capability potentially accelerates project workflows by reducing manual modeling time and eliminating transcription errors between platforms. It enables smaller firms to leverage multiple tools without hiring platform-specific expertise. However, the 20-problem test set remains limited; real-world validation across complex geometries, nonlinear behavior, and dynamic analysis cases would strengthen confidence.

The research establishes a template for similar multi-platform automation in related engineering disciplines. Future work likely includes handling edge cases, integrating validation checks against building codes, and expanding to more specialized FEA tools used in mechanical or civil engineering.

• →Two-stage multi-agent LLM system automates structural analysis across ETABS, SAP2000, and OpenSees platforms simultaneously.
• →Unified JSON intermediate representation enables platform-agnostic engineering specification interpretation and parallel code translation.
• →System achieved 90%+ accuracy across 20 frame problems with consistent performance in repeated trials.
• →Architecture addresses practical engineering workflow constraints where multiple FEA tools are standard practice.
• →Extensible design pattern could accelerate adoption of LLM-assisted automation across other engineering disciplines.