The Application of SSA-BP Neural Networks in Multisensor Fire Detection Systems

This study addresses the issue of improving fire detection accuracy and integrating multi-sensor information in intelligent fire detection systems by proposing an enhanced BP neural network method based on Sparrow Search Algorithm (SSA) optimization. By introducing SSA optimization of weights, adaptive learning rates, and local search mechanisms, the BP neural network is optimized to enhance its training speed and generalization capability. Experimental results demonstrate that compared to the BP neural network's prediction accuracy of 86.43% for flame probability, the optimized SSA-BP neural network achieves a 9.1% increase in prediction accuracy, reaching 95.53%. Additionally, the accuracy of predicting smoldering fires increases from 93.99% to 96.0009%, indicating a 2% improvement in prediction capability. This method holds certain reference value for the further development and application of intelligent fire detection systems.