TÜBİTAK 1707 Contract R&D · 2025 – 2027 · ongoing
Beyçelik Gestamp Otomotiv Sanayi A.Ş.
AutoEncoder and IoT-based water use anomaly detection
A deep learning model that learns water use aims to catch leaks and faults without needing labelled fault data.

Scope
An AutoEncoder-based deep learning model that learns the plant’s water use behaviour is being developed. The model learns the normal consumption pattern and produces an anomaly signal when consumption deviates from it. An IoT-based flowmeter and sensor network is being installed on site, and the data is transferred to a cloud platform.
Requirements and challenges
- Mains leaks and faulty valves being noticed months later on the bill
- Heavy, distributed water use in pressing, washing, cooling and surface treatment lines
- Anomaly detection in plants without fault records
- Client
- Beyçelik Gestamp Otomotiv Sanayi A.Ş.
- Location and date
- TÜBİTAK 1707 Contract R&D · 2025 – 2027 · ongoing
- Sector
- Automotive and metal
- Sensors
- Flow (electromagnetic / ultrasonic), line pressure, temperature (PT100), conductivity
- Software / hardware
- AutoEncoder (deep learning), IoT data acquisition, cloud monitoring, instant alerts




