Catch machine faults before they stop you.
PredictaMaint combines short-time Fourier transforms with a Kolmogorov–Arnold Network to read machine vibration, diagnose faults early and explain its findings, light enough to run on site in real time.
The problem
Rotating machinery fails without much warning.
Turbines, pumps and compressors drive energy and industrial systems. Their vibration signals are irregular and change with operating conditions, so early faults are hard to see, and unplanned downtime is costly.
fault diagnosis accuracy on a machinery test rig.
precision and recall across fault classes.
standard models outperformed: SVM, Random Forest and XGBoost.
Source: the founding research paper (ASEAN Journal of Science and Engineering, 2025).
How it works
How it works
A clear path from data to decision.
Sense
Vibration sensors stream data from the machine.
Transform
STFT turns raw signals into time–frequency features.
Diagnose
A hybrid KAN model classifies the machine's condition.
Act
Maintenance teams get early, explained alerts.
Product
What makes it different.
Every feature is grounded in the founding research.
Hybrid KAN + STFT
Captures irregular, changing vibration patterns.
97.5% accuracy
Precision and recall above 95%.
Explainable
Shows which signal features drive each diagnosis.
Edge-ready
Lightweight enough for real-time on-site use.
Who it's for
Who it's for.
Power plants & utilities
Protect turbines and generators.
Water & oil facilities
Monitor pumps and compressors.
Manufacturing plants
Reduce unplanned downtime on rotating equipment.
The research
Built on peer-reviewed research.
- Hybrid Kolmogorov–Arnold Network (KAN) with STFT for rotating machinery diagnosis.
- Achieved 97.5% accuracy with precision and recall above 95%.
- Outperformed SVM, Random Forest and XGBoost on a machinery test rig.
- Interpretable and lightweight, suitable for real-time predictive maintenance; cited 12 times so far.
Khalaf, M. I., Al-Haddad, L. A., et al. (2025). Enhancing Predictive Maintenance in Energy Systems Using a Hybrid KAN with STFT Framework for Rotating Machinery. ASEAN Journal of Science and Engineering, 5(2).
Team
The team behind it.
Founded in 2025 by a student team from the faculty below.
Faculty of Engineering Technologies (Power Mechanics Engineering)
Rotating machinery, vibration analysis and AI diagnostics.
FAQ
Questions we hear often.
What sensors do we need?
Standard vibration sensors (accelerometers) on the machines you want to monitor.
Can it run without the cloud?
Yes. The model is light enough to run on site in real time.
Is this a finished commercial product?
It is a research-based student startup moving from validated prototype toward pilots with real users.
Can we partner or run a pilot?
Yes. We are looking for organisations willing to test the solution in real conditions and share feedback.
Bring PredictaMaint to your organisation.
We are onboarding pilot partners now. Read the research and get in touch.