Sensor Data Fusion for an Intelligent Fire Detection and Control System

This proposal deals with the application of advanced mathematical concepts for developing a general purpose, fire detection and suppression system that can be adapted for residential, commercial, and industrial environments. The system utilizes a suite of common (or future) sensors that are networked; it employs a distributed control strategy with local "intelligence," and uses advanced data and information fusion techniques; this is not a sensor "voting" rule. The Intelligent Fire Control (IFC) system offers improved reliability via redundancy, cross validation and triangulation. The IFC also offers greater robustness and is less vulnerable to failures in a single sensor or in a control console. While based on existing sensor hardware, continuous monitoring of the sensors' output allows the control algorithms to utilize the sonsors' proportional output rather than a fixed set-point (threshold) control. By continuously performing situational assessment before, during, and after a fire event, the IFC is capable, via event-anticipation, of providing high sensitivity with a low false alarm rate. Furthermore, the IFC is designed to continuously select/deselect necessary fire abatement means for most effective suppression while avoiding unnecessary damage due to water, etc. The capability for autonomous self inspection, fault detection, and fault tolerance via redundancy or compensation (for a faulty sensor) is built into the system. The Phase I objectives are to demonstrate (i) the feasibility of such a system via controlled lab experiments, and (ii) a higher detection probability at a lower false alarm rate than a system based on a set-point threshold with a discrete sensor. Extensive algorithm development and hardware testing will be conducted during Phase II.