Analysis of Multi-Criteria Fire Detection Data and Early Warning Fire Detection Prototype Selection

摘要

This report describes the analysis of Fire/Nuisance Source data and the selection of sensors for an early warning, multi-criteria, fire detection system for the Office of Naval Research (ONR) program on Damage Control: Automation for Reduced Manning (DC-ARM). In this work, the analysis of transient fire signatures is studied using a probabilistic neural network (PNN). Experiments are described to study the effects of various PNN training parameters and to determine the optimal sensor suite combination, which enables both early fire detection and high nuisance source rejection. Comparisons are made between the candidate sensor arrays, commercial fire detection systems, and sensor arrays proposed in previous reports Recommendations and directions for future research are also given.