TÜBİTAK
Predicting maintenance before failure
In a TÜBİTAK-backed R&D project we built and compared machine-learning models that predict industrial equipment maintenance needs.
Client
TÜBİTAK
Sector
R&D · Industrial AI
Technology
Python · XGBoost · scikit-learn · LSTM · Pandas
The challenge
Industrial maintenance runs either too early, wasting money, or too late, turning into failure. Which model family actually works on this data was an open question.
The solution
We built and benchmarked Random Forest, XGBoost, Support Vector Machines and LSTM on the same dataset, ran time-series analysis and documented the results in a scientific analysis report and an academic paper.
Results
- 4 model
- RF · XGBoost · SVM · LSTM
- TÜBİTAK
- backed R&D
- Makale
- academic output
Related work
300+
facilities live
Atalian
Running 300+ facilities on one platform
We digitised and standardised Atalian's field and facility operations with Ontime.
Göcek
bays, in production
Türkiye Çevre Ajansı
Protecting the Göcek bays with digital buoy reservations
We built DERİA, a smart buoy-reservation system combining GIS, IoT and AI, for the Turkish Environment Agency.
On-prem
data sovereignty
TCDD
Technical-document AI that never leaves the building
We deployed BelgeChat fully on-prem for complex technical manuals.
Have a similar project?
Bring your idea; we'll scope the right system and deliver it end to end.