Hierarchical Temporal Memory Continuous Learning Algorithms for Fire State Determination

Abstract

An ultimate goal of placing fire detection systems in buildings and structures is to allow for the rapid detection of fire and accurate faster than real time prediction of ensuing fire behavior so that relevant information can be delivered to the appropriate stakeholders. In the near-term, development of detection systems with decreased detection time, better discrimination against nuisance and false alarms, and real-time monitoring of the fire state is a critical interim step. Building comfort and efficiency systems are increasingly incorporating a greater quantity of sensors and these sensors are installed at a greater density than any fire sensor with the exception of the sprinkler. While currently used primarily for building management purposes, the application of these, or similar types of building sensors, for rapid fire detection, fire state determination, and fire forecasting offers great potential. This paper discusses the potential benefits of the application of Hierarchical Temporal Memory algorithms for fire state determination in a continuous learning environment based on its application to a series of live fire experiments.