Multi-criteria fire detection systems using a probabilistic neural network

The Navy program, Damage Control Automation for Reduced Manning (DC-ARM), i s focused on enhancing automation of ship functions and damage control syst ems. A key element to this objective is the improvement of current fire det ection systems. As in many applications, it is desired to increase detectio n sensitivity and, more importantly, increase the reliability of the detect ion system through improved nuisance alarm immunity. Improved reliability i s needed such that fire detection systems can automatically control fire su ppression systems. The use of multi-criteria-based detection technology con tinues to offer the most promising means to achieve both improved sensitivi ty to real fires and reduced susceptibility to nuisance alarm sources. A mu lti-signature early warning fire detection system is being developed to pro vide reliable warning of actual fire conditions in less time with fewer nui sance alarms than can be achieved with commercially available smoke detecti on systems. In this study, a large database consisting of the responses of 20 sensors to several different types of fires and nuisance sources was gen erated and analyzed using a variety of multivariate methods. Three data mat rices were developed at discrete times corresponding to the different alarm levels of a conventional photoelectric smoke detector. The alarm times rep resent 0.82%, 1.63% and 11% obscurations per meter. The datasets were organ ized into three classes representing the sensor responses for baseline (non fire), fires and nuisance sources. A robust data analysis strategy for use with a sensor array consisting of four to five sensors for early fin detect ion and nuisance source rejection was developed using a probabilistic neura l network (PNN) that was developed at the Naval Research Laboratory for che mical sensor arrays. The analysis algorithms described in this paper evalua te discrete samples and develop classification models that examine individu al chemical signatures at discrete points. Published by Elsevier Science S. A.