A Predictive Maintenance System for Field Cabinets Using Artificial Intelligence

İngilizce (OpenAlex)

Outdoor telecommunication field cabinets are exposed to environmental conditions that can lead to equipment degradation, network instability, and increased maintenance costs. However, current monitoring approaches are predominantly reactive and lack predictive intelligence at the environmental level. This study presents the design and implementation of an Internet of Things (IoT)-based intelligent monitoring and predictive maintenance system for network cabinets used in telecommunication infrastructures. These cabinets, which house critical network equipment, are highly sensitive to environmental conditions such as temperature and humidity. The proposed system integrates DHT22 sensors with a Raspberry Pi Zero 2 W microcomputer to collect real-time environmental data and transmit it to a centralized database via a FastAPI-based service. To ensure data reliability, preprocessing techniques such as outlier removal, interpolation, and normalization were applied. Three different artificial intelligence algorithms were developed to detect abnormal patterns in time series data: Isolation Forest, Autoencoder, and Long Short-Term Memory (LSTM) Forecast. Results indicate that the Isolation Forest effectively identifies sudden spikes, while the LSTM model achieves superior temporal awareness by detecting deviations with a Mean Squared Error (MSE) below 0.0002 for temperature trends. Field tests conducted on an actual cabinet in Istanbul confirmed that the developed architecture can detect anomalies at an early stage, improve network reliability, and reduce maintenance costs. The study shows that the integration of IoT-based sensors with AI-powered analytics provides a scalable and energy-efficient solution for proactive monitoring of telecommunication infrastructures.