Ubiquitous Monitoring and Alert Fire Detection System in Data Center using Logistic Regression

Abstract

Data centers are the core of any organization. They house its data, host its services, ensure business continuity, and serve as the bedrock of decision-making. Therefore, round-the-clock operation is essential to maintaining the continuity of services provided by an organization. This requires robust, standardized security systems to restrict exposure to threats to infrastructure and data. Yet, existing data center alarm systems do not have robust monitoring and reporting capabilities that generate premature action on high-priority alerts. In this research paper, we propose a data-driven automated method to optimize fire detection in data centers through the use of machine learning algorithms and sensor networks to inspect large amounts of data and detect patterns of fires. The installation of the proposed system includes the employment of an ESP32 development board to handle real-time data and wireless communication with different sensors, such as smoke detectors, temperature, gas, and motion sensors, in order to facilitate end-to-end monitoring. We have achieved an intelligent monitoring system through the employment of machine learning for handling sensor data, adaptively setting thresholds, and initiating corresponding actions ranging from sending alerts to alarm sounds. The system is highly accurate (≈97–98%) with strong confidence-based filtering and minimal computational cost, which renders it ideally suited for real-time indoor use. The method provides consistent detection through diversified materials (carton, cloth, electrical), filling the gap in situations where visual information is not necessarily accurate or reliable. Experimental results demonstrate the efficiency of the system in the timely notification of fire cases, bearing witness to its practical application over vision-based or simulation-based approaches.