Sensor Optimisation for Aircraft Health Management Systems.
Doctoral thesis · Cranfield University · IVHM Centre · defended November 2025. A plain-English walkthrough of the work, the framework it produced (MOSOF), the score it introduced (NDCI), and what was actually validated on aircraft subsystems.
Title
Sensor Optimisation for Aircraft Health Management Systems
Institution & programme
Cranfield University · School of Aerospace, Transport & Manufacturing · IVHM Centre · PhD Transport Systems · academic years 2022–2025
Supervision
Supervisor: Prof. Ian K. Jennions · Associate Supervisor: Dr Fakhre Ali
Validation
Cross-subsystem evaluation across Engine, Fuel, EPS, and ECS via Cranfield's SESAC platform; repeated nested cross-validation
Key contributions
MOSOF · Normalised Diagnostic Contribution Index (NDCI)
Headline result
Pareto-front knee at ~12 sensors (Engine 5 · Fuel 2 · EPS 2 · ECS 3) at ≈0.69 normalised diagnostic performance
Sensor networks are made wrong.
If you walk through the design history of any complex engineering asset — an aircraft, a power plant, a factory floor — the sensor network is rarely the result of a deliberate optimisation. It's the result of accumulation. Each subsystem owner adds the sensors they need. Nobody owns the whole picture.
The consequence is predictable: simultaneously over-instrumented in some places (redundant sensors that contribute nothing new to a diagnosis) and blind in others (failure modes that no current sensor can detect).
MOSOF: a framework.
The Multi-Objective Sensor Optimisation Framework (MOSOF) treats sensor selection as a constrained search over an asset's information surface. Instead of asking "is this sensor good?" — a question with no answer — it asks "given everything we care about (cost, weight, reliability, diagnostic coverage, and more), which combinations of sensors are non-dominated?"
The output isn't a single answer. It's the Pareto front: every configuration on it is the best you can do on at least one axis without giving up something on another. From there, a stakeholder picks — explicitly, with the trade-offs visible — instead of arguing about which sensor "feels" right.
NDCI: diagnostic value, quantified.
MOSOF answers the framing question. The Normalised Diagnostic Contribution Index (NDCI) — the second contribution — answers the scoring question: how much does this particular sensor contribute to system-level diagnostic capability?
NDCI gives every sensor a comparable, normalised score for its share of diagnostic coverage. That's important because two sensors that look similar on paper can contribute very differently to the diagnosis of real failures, and the difference is invisible without a measure like NDCI.
Validated across four subsystems.
Rather than a single case study, the thesis evaluates MOSOF + NDCI across four aircraft subsystems — Engine, Fuel, Electrical Power System, and Environmental Control System — using Cranfield's SESAC simulation platform with platform "symptom vectors" to model fault propagation. Repeated nested cross-validation prevents the optimistic bias that crops up in lighter evaluations.
The headline output is a defensible Pareto front whose knee solution is just 12 sensors distributed as Engine 5, Fuel 2, EPS 2, ECS 3, achieving approximately 0.69 normalised diagnostic performance. Under consistent learners, NDCI-guided suites are reliably more compact than mRMR-selected baselines while delivering better-balanced accuracy across three of the four subsystems.
From thesis to products.
The methods don't stay on the page. AcoustR — my lead venture — takes the same diagnostic-value-per-sensor thinking to its limit, building a product around a single, low-cost, high-information sensor (a microphone). Sensorry packages the MOSOF / NDCI workflow itself for engineering teams.
The research agenda continues: extending NDCI-style scoring to multi-modal sensor fusion, stakeholder-aware design where OEM, operator, and MRO weightings are explicit, and decision frameworks that surface uncertainty honestly.
A few terms.
For readers coming from outside the field, the four terms that matter most.
IVHM
The discipline of designing aerospace and industrial systems so they can monitor their own health, diagnose problems, and predict failures — using sensor data, models, and decision frameworks together. Cranfield's IVHM Centre is one of the field's leading academic homes.
MOSOF
The framework introduced in this thesis for choosing sensor networks under multiple competing objectives. Treats sensor selection as a Pareto-front search instead of a single-axis cost-versus-coverage trade-off.
NDCI
A per-sensor score for the share of system-level diagnostic capability that sensor contributes. Lets candidate sensors be compared on a common, normalised scale rather than by domain intuition.
Pareto front
In a multi-objective problem, the set of solutions where you can't improve any objective without making another worse. The "best" answer is rarely a single point — it's a curve or surface, and choosing on it is a stakeholder decision, not a calculation.
Read the full thesis.
The complete doctoral document — Cranfield University 2025, ~6 MB PDF. Includes full literature review, methodology, sub-system case studies, and supporting code references.
MOSOF with NDCI: A Cross-Subsystem Evaluation of an Aircraft for an Airline Case Scenario
NDCI Integration to Multi-Objective Sensor Optimisation Framework — An ECS Case
Understanding the Role of Sensor Optimisation in Complex Systems
Full publications archive
Examining, citing, or extending it?
If you're a researcher building on this thread, an academic considering hosting it, or an industry partner curious whether MOSOF / NDCI fits your problem, I'd be glad to talk.
hello@buraksuslu.com →