Design of an intelligent fire protection system

Fire protection systems utilize detectors that process signals from fire sensors using threshold-based methods and generate a fire signal based on a logical function. Artificial neural networks can enhance these detectors by processing information from a network of sensors after being trained. To train these neural networks, extensive data sets are necessary, which can be obtained through fire simulations on a supercomputer. Field tests are costly, subject to random factors, limited to one or two rooms, and do not provide a comprehensive picture of fire development. Thus, designing intelligent fire systems falls under model-based design. Through modeling, large data sets were generated for training fire system algorithms, expanding the range of tasks they can address. A group of neural networks is proposed for optimizing the placement of multi-parameter sensors, identifying the type of burning material, detecting fires at early stages, and localizing the fire zone to select appropriate extinguishing agents. Artificial neural networks enable the prediction of fire development, mapping hazardous factors' distribution to find optimal evacuation routes. An example of model-based design for a ship fire protection system is provided.