Hardware and software modules development for multi-sensor fire detection and remedial control using fuzzy logic and convolutional neural networks

Abstract

Despite the fire safety campaigns, the rate at which fire outbreaks continue to rise requires an engineered solution system that can detect a fire early before it escalates and send a notification system to the fire service crew. This paper addresses the situation by designing and developing a fuzzy logic-based multisensor fire detection.. The designed solution also has a web-based and SMS notification system with trained convolutional neural networks for proximity and wide-area fire detection. The Raspberry pi 4 model processes data using three fire signatures, smoke flame and heat, using the MQ2 sensor, IR flame sensor, and DHT22 senor, respectively using the fuzzy logic algorithm to determine the fire status. With broader coverage, the convolutional neural network ensure wide-area fire detection using the Raspberry pi came 5MP. By incorporating an email, a web-based, and SMS notification system, the fire service crew gets real-time notification of fire scenes, the location of the fire scenes, and map integration in the web app to help them navigate to the residence. The residents of the fire scene also get alerts via the mobile app and SMS. The efficiency and response time of this paper's final engineered solution system showed significant improvements with early fire detection and notification. The final experiment and performance evaluation results the CNN has an accuracy of 96% while the fuzzy logic is 98%.