Intelligent decision support system for automated fire monitoring in technical systems

Modern automated fire monitoring systems play a key role in preventing catastrophic consequences in technical systems and critical infrastructure facilities. This paper examines contemporary fire detection methods, including the use of wireless sensors, geographic information systems, machine learning technologies, and neural networks. Special attention is given to the application of YOLOv8 algorithms for real-time fire and smoke detection based on video camera images, including those installed on unmanned aerial vehicles and mobile platforms. The study covers the development, training, and optimi-zation of an intelligent decision support system (DSS) inte-grated with the YOLOv8 model. During the experiments, the impact of the number of training epochs, data structure, and preprocessing methods on model accuracy was analyzed using mAP50, Precision, Recall, and F1-score metrics. The results demonstrated that increasing the training dataset, including negative examples, and adapting hyperparameters significantly improve detection accuracy. The developed system provides automatic operator notifi-cations, activation of fire prevention measures, and initiation of emergency response protocols. The paper also discusses prospects for further system development, including fire spread prediction, risk analysis, and integration with emergency management systems.