Orthogonal Gas Sensor Arrays with Intelligent Algorithms for Early Warning of Electrical Fires

Abstract

The feasibility of utilizing chemical sensor arrays and multivariable analyses as the basis for an early-warning combustion alarm for electrical fires was evaluated. During the pre-combustion phase of electrical fires, electronic components will heat up, resulting in an out-gassing of chemical vapors, which generally will precede the formation of smoke, scorching and fire. A variety of materials (PVC, Teflon^®, Kapton^®, and silicone rubber) that are frequently used as wire insulation were subjected to electrically induced thermal excursions, thereby simulating an electrical failure and possible pre-combustion condition. The off-gassing vapors from the various coatings can serve as chemical signatures for a pending fire and were detected by an array of chemical sensors (e.g., an electronic nose). Principal component analyses and KNN identification algorithms applied to the sensor response patterns successfully identified the various vapor sources. A 20-sensor array including electrochemical sensors, quartz microbalance (QMB) sensors with different polymer coatings, and heated metal oxide sensors (MOXs) was evaluated and the optimal performance was obtained using the electrochemical and MOXs. The use of heterogeneous orthogonal sensors increased the information content of sensor array signals and a diminutive array can still identify fire materials and extent of damage. The small, lightweight, inexpensive and low power sensors used to detect vapors during pre-fire conditions were ideal for space or commercial aircraft applications.