Multi-Agent AI Architecture for Risk-Aware Manufacturing Decision Support
The newly developed MAKA (multi-agent knowledge analysis) architecture focuses on achieving high-precision CNC machining for aerospace parts, including Ti-6Al-4V rotor blades. This system combines intent routing, quantitative analysis, knowledge graph retrieval, and critic-based verification to ensure physical plausibility, safety limits, and complete provenance prior to human approval. Testing was conducted on a machining testbed, employing virtual-machining path-tracking error fields, simulations of cutting force and deflection, as well as scan-based 3D inspection deviation maps from 16 blades. This architecture overcomes the shortcomings of standard LLMs in managing risk-constrained multi-step numerical processes and in delivering auditable provenance for critical decision-making.
Key facts
- MAKA is a multi-agent decision-support architecture for manufacturing.
- It separates intent routing, tools-only quantitative analysis, knowledge graph retrieval, and critic-based verification.
- The system enforces physical plausibility, safety bounds, and provenance completeness.
- It was instantiated on a Ti-6Al-4V rotor blade machining testbed.
- The testbed used virtual-machining path-tracking error fields, cutting-force and deflection simulations, and scan-based 3D inspection deviation maps from 16 blades.
- The architecture is designed for high-precision CNC machining of free-form aerospace components.
- It requires bounded compensations informed by inspection, simulation, and process knowledge.
- The system is human-in-the-loop, requiring human approval for recommendations.
Entities
—