The Need

Designing online monitoring (OLM) for safety‑critical systems is constrained by scarce early‑stage operational data and by quantitative models that are slow to build, brittle across configurations, and costly to iterate. This creates expensive sensor networks with blind spots, poor diagnosability/prognosability, and challenging field integration. Industry needs a data‑light, system‑agnostic way to choose sensor types, counts, and locations that jointly maximize diagnostic power and observability while minimizing cost and installation burden, even when only qualitative knowledge is available.

The Technology

OSU engineers have developed a novel technology that generates a qualitative model of the target system, simulates fault propagation to extract signal features, and uses multi‑objective optimization to produce Pareto‑optimal sensor deployment options. Each option specifies sensor types and placements that best meet user‑defined criteria (fault detection/discrimination, risk reduction/early warning, observability, sensor failure tolerance, functionality, integrability, and cost), under explicit constraints.

Commercial Applications

• Advanced nuclear and conventional power.
• Process industries (chemical, refining, gas).
• Transportation and aerospace platforms.
• Manufacturing, electronics, and environmental monitoring.

Benefits/Advantages

• Works when historical data are limited, leverages qualitative physics to anticipate fault signatures and transients.
• Simultaneously optimizes multiple objectives and constraints, returning Pareto‑optimal, ready‑to‑trade‑off deployment choices.
• Explicitly accounts for integrability (installation effort) and total sensor cost to reduce penetrations and project risk.
• Improves diagnosability/prognosability and supports sensor failure tolerance, enhancing resilience of OLM architectures.