Physics-Grounded Multi-Agent Architecture for Traceable, Risk-Aware Human-AI Decision Support in Manufacturing

A team led by Danny Hoang and including researchers from the US Army DEVCOM Ground Vehicle Systems Center and the University of Michigan has introduced MAKA (Multi-Agent Knowledge Analysis), a novel human-in-the-loop decision-support architecture for high-precision CNC machining of free-form aerospace components. The system addresses a critical gap in off-the-shelf LLM assistants: they cannot reliably execute risk-constrained multi-step numerical workflows or provide auditable provenance for high-stakes manufacturing decisions. MAKA separates intent routing, tools-only quantitative analysis, knowledge graph retrieval, and critic-based verification that enforces physical plausibility, safety bounds, and provenance completeness before recommendations reach human operators.

MAKA was instantiated on a Ti-6Al-4V rotor blade machining testbed, fusing virtual-machining path-tracking error fields, cutting-force and deflection simulations, and scan-based 3D inspection deviation maps from 16 blades. The analysis decomposes deviation into evidence-linked pathing, drift-based wear proxy, residual systematic compliance, and variability proxy for instability-aware escalation. In a three-level tool-orchestration benchmark (single-step through ≥3-step stateful sequences), MAKA improved successful tool execution by up to 87.5 percentage points relative to an unstructured single-model interaction pattern. Digital twin what-if studies showed MAKA can coordinate traceable compensation candidates that reduce predicted surface deviation from order 10⁻²in to approximately ±10⁻³in over most of the blade, providing a pre-deployment verification signal for risk-aware human decision-making.