What is MOSOF?
A plain-English explainer for the Multi-Objective Sensor Optimisation Framework. The technical paper is in Sensors; this page is for the reader who just wants to know what MOSOF is, why it exists, and whether it applies to their problem.
TL;DR
MOSOF is a framework for choosing what sensors to put on a complex engineering asset by treating sensor selection as a multi-objective optimisation problem. Instead of asking "is this sensor good?" — a question with no defensible answer — it asks "given everything we care about, which sensor configurations are non-dominated?" The output is a Pareto front, not a single number, so stakeholders pick a configuration with the trade-offs visible rather than hidden.
The acronym
MOSOF stands for Multi-Objective Sensor Optimisation Framework. It was introduced in the doctoral programme of Dr Burak Suslu at Cranfield University's IVHM Centre, and developed across two peer-reviewed publications in Sensors (MDPI) — a 2023 foundation paper and a 2025 paper integrating the Normalised Diagnostic Contribution Index (NDCI).
What problem it solves
On most complex engineering assets — aircraft systems, industrial plants, vehicles — sensor networks accumulate rather than get designed. Each subsystem owner adds the sensors they need; nobody owns the whole picture. The result is networks that are simultaneously over-instrumented in some places and blind in others.
MOSOF replaces "ask each subsystem and add up" with "treat sensor selection as a constrained search over the asset's information surface, with multiple objectives explicitly traded off". That sounds abstract; the concrete output is a small set of recommended sensor configurations and the trade-offs between them.
How it works (high level)
- Encode the asset. What are the candidate sensor positions, the failure modes that matter, and the objectives the team actually argues about? Cost, weight, reliability, diagnostic coverage, latency.
- Encode the constraints. Physical, regulatory, supply-chain, and schedule constraints that no candidate configuration is allowed to violate.
- Search the candidate space. A multi-objective evolutionary algorithm (typically NSGA-II family) searches the space of possible sensor configurations.
- Score with NDCI. When the candidates are sensors, MOSOF uses the Normalised Diagnostic Contribution Index to score how much each sensor contributes to system-level diagnostic capability — fixing a long-standing ambiguity in sensor-design literature about what "good sensor" actually means.
- Deliver the front. The team gets the Pareto front (every configuration on it is non-dominated) and a knee-point recommendation, with the trade-offs visible.
Where it's been validated
MOSOF was validated on the Boeing 737-800 Environmental Control System through Cranfield's SESAC simulation platform — an established testbed in IVHM research. The validated approach consistently identified smaller, lighter, cheaper sensor configurations that matched or exceeded the diagnostic capability of larger reference designs.
When MOSOF is the right tool
- You're scoping a sensor network on a complex asset (aerospace, industrial, automotive).
- The decision genuinely involves multiple objectives that pull against each other.
- Stakeholders disagree about what "right" looks like — and a Pareto front is the honest answer.
- The cost of getting the call wrong is large enough to justify the analysis.
It's not the right tool when the trade-off space collapses to a single dominant objective, or when the candidate space is small enough that brute-force enumeration is faster.
Read the source
If you want the technical depth — the math, the encoding, the experimental setup — the two peer-reviewed papers are the right starting point.
Understanding the Role of Sensor Optimisation in Complex Systems
NDCI Integration to Multi-Objective Sensor Optimisation Framework — An ECS Case
Got a sensor-network decision?
If you're scoping or auditing a sensor network on a complex asset, the Sensor Optimisation Audit applies MOSOF directly to your problem in three weeks. The full consulting page has the details.
See the consulting page →