You keep adding sensors, but diagnostic confidence does not improve.
Or nobody can explain any more why the current configuration was chosen. More channels have not meant better answers.
Industry · University–industry R&D
Hard trade-offs, worked on together.
Some engineering decisions have no single right answer: which sensors to fit, which signals to trust, how much monitoring a plant or a fleet can justify. I would like to work on problems like these with companies. Any collaboration runs through Kırıkkale University (KKÜ): joint TÜBİTAK projects, contracted R&D through the university, and student projects on real company problems.
The methods are peer-reviewed and published, so you can read them before we talk. The first step is an email, free and without commitment.
Türkçe: Sanayi işbirliği sayfası →
§ 01 / Problems worth bringing
If one of these sounds familiar.
The first four are the kind of problem the published doctoral methods address. The last three come from the research threads I am building at KKÜ. Any of them is worth an email.
Sensors and signals
Teams disagree on which signals actually matter.
False alarms, missed faults or redundant measurements are adding to maintenance cost, and the argument cannot be settled with evidence.
You need a defensible sensor or telemetry strategy before a design gate or tender.
A redesign gate, a tender or an R&D programme review is coming, and the sensor stack has to be specified and argued for.
You must justify a configuration to finance, safety or leadership.
“It seemed right” is no longer enough. You need the trade-offs written down, with the cost of each option visible.
Newer threads
Rotating equipment too numerous for permanent sensors.
Pumps, motors and compressors in the middle tier of a plant. Low-cost acoustic monitoring with a phone or a commodity microphone, as a screening layer rather than certified condition monitoring.
Related project: SensorryMaintenance decisions that hide their uncertainty.
A model returns one label, but you need to know how sure it is before you stop a line or ground an aircraft. Calibrated uncertainty, for example conformal prediction, for fault diagnosis and prognostics.
Engineering documents that cannot leave the building.
Local, private language models for engineering documents, running on ordinary CPU hardware, with speed measured and quality checked rather than assumed, in Turkish as well as English.
§ 02 / What a collaboration produces
Options, not a single answer.
The exact outputs depend on the route and the problem. For a sensor or monitoring question, a collaboration can produce the following, each written so that your team can use it.
- NDCI contribution rankingEvery candidate sensor scored for what it adds to telling faults apart: separation power, sensitivity to fault severity, and uniqueness.
- Sensor redundancy mapWhich signals carry overlapping information, and where a channel could go without losing diagnostic coverage.
- Coverage-gap analysisFaults or operating conditions that the current suite cannot separate, and what it would take to see them.
- Pareto-optimal configuration optionsA set of configurations across your objectives (diagnostic performance, cost, reliability or others), with the knee and the trade-offs made explicit.
- Prototype codeWorking code, usually Python, that runs on your data and sits next to the models your team already uses where that is possible.
- A written methodThe formulation, assumptions and data choices written down, so your team can re-run the analysis and defend the result after the project ends.
Fig. 01 · Pareto frontSchematic · not project data
§ 03 / Ways to work together
Routes that run through the university.
As a full-time academic at a public university, I take on industry work only through university channels. The routes below are established ones. Some come with public funding, and every formal route sets out its terms in writing. Programme rules and calls change, so the official pages are always the reference.
Start here · No cost
An initial technical conversation
An email exchange and, if useful, a call. We look at the problem, the data that exists and the decision it feeds. If the problem is not the right shape for these methods, I will say so. There is no commitment, and no confidential data is needed at this stage.
Funded · TÜBİTAK
Joint TÜBİTAK projects
1505 University–Industry Collaboration Support Programme. Your company applies together with KKÜ. The university team runs the project, led by a full-time academic with a PhD. Applications are open all year.
1501 Industrial R&D Projects and 1507 SME R&D Start-up. Company-led grants in which R&D services or consultancy bought from a university are an eligible cost. I take part through a university contract.
1505 excludes companies in which a member of the project team, or a first-degree relative, is a partner or manager.
Contract · Through KKÜ
Contracted R&D and expert services
Scientific opinions, projects and expert services that a company requests are contracted through the university’s revolving fund (döner sermaye; Law 2547, art. 37 and 58(e)). University–industry R&D, design and innovation projects run through a dedicated revolving-fund account under art. 58(k), with the permission of the university’s executive board.
KKÜ’s Technology Transfer Office (KKUTTO) is the university’s office for industry collaboration and intellectual property. KKÜ’s internal research-project scheme (BAP) also includes industry-cooperation projects.
Students · Supervised
Student projects on company problems
A real problem from your company can become a supervised student project: an undergraduate graduation design project, a TÜBİTAK 2209-B industry-oriented undergraduate project (with an industry advisor at your company alongside the academic advisor), or an MSc thesis. For doctoral-scale work, the 2244 Industrial PhD programme is a joint application by the university and a company.
The company brings the problem, the data and a contact person (for 2209-B, the industry advisor). The students do the work themselves, under supervision.
Place · Teknopark
Companies at Kırıkkale Teknopark
Kırıkkale Teknopark is a technology park in Kırıkkale, and Kırıkkale University is its largest shareholder. Companies based there can work with the university through the same routes.
Technopark companies are also among those that can partner with a university in the 2244 Industrial PhD programme.
Terms
Intellectual property, publication and confidentiality are agreed through the university, in writing, before the work starts: in the university agreement or in the funding application. KKÜ’s procedures and the programme rules set those terms; I cannot promise them personally.
The methods I bring are already public. How your data, your results and any new intellectual property are handled is part of that agreement.
§ 04 / How it works
Four steps, in writing.
-
Step 01
Email a short description
A few sentences on the system, the decision in front of you, the timeline and what a wrong answer would cost. No NDA is needed; please hold back confidential data for now.
-
Step 02
A scoping conversation
We check whether the problem fits these methods, what data exists, and which route suits it. If it is not a fit, I will say so.
-
Step 03
Agree the format with KKUTTO
The route is set up through the university: a TÜBİTAK application, a revolving-fund contract or a student project, with scope, IP, publication and confidentiality agreed in writing.
-
Step 04
Short cycles, written results
Each cycle ends with something written: the result, the code that produced it, and what changed since the last one. Depending on the route, students may be part of the team.
§ 05 / Why talk to me
Read the methods before we talk.
Published, peer-reviewed methods.
MOSOF, the Multi-Objective Sensor Optimisation Framework, and NDCI, the Normalised Diagnostic Contribution Index, are published in Sensors (MDPI), with reference code on GitHub.
There is no black box. You can judge the approach, and whether it suits your problem, before you spend any time on a call.
2026
MOSOF with NDCI: A Cross-Subsystem Evaluation of an Aircraft for an Airline Case Scenario
2025
Normalised Diagnostic Contribution Index (NDCI) Integration to Multi Objective Sensor Optimisation Framework (MOSOF)—An Environmental Control System Case
2023
Understanding the Role of Sensor Optimisation in Complex Systems
3 vs 6
Sensors needed to reach ≥95% of maximum diagnostic coverage: NDCI against mRMR, environmental control system (2025 paper).
88.6%
Engine diagnostic accuracy with NDCI, against 69.0% for mRMR, under nested cross-validation (2026 paper).
12 / 96
Sensors in the knee suite, from 96 candidates across engine, fuel, electrical power and environmental control (thesis; the 2026 paper reports a knee of 12–14).
Simulated These results come from simulation, not from a flying aircraft: Cranfield’s SESAC platform for the environmental control system study, and a high-fidelity virtual aircraft model for the cross-subsystem study. Your system will differ; finding out how is what the scoping conversation is for.
Five years of industrial delivery.
Before the PhD, I founded and led SSL Elektrik-Elektronik in Batman, Türkiye, from July 2017 to April 2022: industrial automation and high-voltage systems for state-run oil and gas, algorithm-driven control in hazardous environments, and IoT sensor integration into legacy machinery.
The flagship project was a phased retrofit of cracking-tower control systems at a state-run petroleum facility, with an energy-efficiency gain of about 20%. That work shapes how I collaborate: a result only counts if the people running the plant can act on it.
§ 06 / Sectors
Where the methods travel.
The methods are not tied to one domain. They apply wherever the question is what to measure and how to weigh cost against diagnostic value. These are the sectors I know best.
Aerospace and defence
Health management (IVHM), sensor-suite design and diagnostic coverage. The doctoral work was on aircraft engine, fuel, electrical power and environmental control systems.
Energy and power systems
High-voltage and distribution assets: where to measure, which alarms to trust, and how to monitor more assets without fitting more hardware.
Oil and gas
Process plant, control-system retrofits and monitoring in hazardous environments, where I spent five years before the PhD.
Manufacturing and heavy industry
Rotating equipment, legacy machinery and sensor retrofits, including low-cost acoustic screening for assets too numerous to instrument permanently.
Probably not a fit.
Saying so early saves everyone time.
- Cost cutting along a single axis that pretends there is no trade-off.
- Work whose deliverable is a slide deck rather than a result, a method or code.
- Anything I could not defend in writing, under my own name.
- Private consulting outside the university. All my industry work runs through KKÜ.
§ 07 / Questions
Common questions.
Do you still offer private consulting?
No. Since joining Kırıkkale University, I take on industry work only through university channels, so there are no private engagements, fees or retainers. Older links to the consulting page now lead here.
Do we need to share confidential data to start?
No. A description of the system and the decision is enough to judge fit, and anonymised or synthetic data is fine for a first look. Anything confidential is shared under the university agreement, once there is one.
Are the methods proprietary?
No. MOSOF and NDCI are peer-reviewed and published, with reference code on GitHub. What a collaboration adds is the adaptation to your system, your objectives and your constraints.
Is the work limited to aerospace?
No. The published results are on aircraft subsystems, and my industrial background is in oil and gas and high-voltage systems, but the multi-objective structure applies wherever sensors, diagnostics or costly trade-offs are the bottleneck.
Can the work fit our existing tools?
Usually. Prototype code is normally Python and is written to sit next to the models or simulations your team already runs. Where a clean integration is not possible, I will say so at the scoping stage.
Can my company apply for TÜBİTAK 1505 with KKÜ?
Yes, if it is a capital company based in Türkiye, whether an SME or a large firm; sole proprietorships cannot apply. Your company applies as the customer organisation together with the university, and the two sign a cooperation agreement before applying. Companies in which a member of the project team, or a first-degree relative, is a partner or manager are excluded. The current rules are on TÜBİTAK’s 1505 page.
§ 08 / Get in touch
Describe the decision.
A few sentences on the system, the decision in front of you, the timeline and what a wrong answer would cost are enough for a first reply. No deck needed. I read every message.
KKÜ Institutional[email protected]
General[email protected]
Subject lineIndustry collaboration
LanguagesEnglish or Turkish
Time zoneTürkiye (UTC+3)
Türkçe: Sanayi işbirliği sayfası