Generating Synthetic Sensor Data to Facilitate Machine Learning Paradigm for Prediction of Building Fire Hazard
Wai Cheong Tam, Thomas G. Cleary, Yujun Fu
This paper presents a learning-by-synthesis approach to facilitate the utilization of machine learning paradigm to enhance situational awareness for fire-fighting in buildings. An automated Fire Data Generator (FD-Gen) is developed. The overview of FD-Gen and its capabilities are highlighted. Using CFAST as the simulation engine, time series for building sensors including heat detectors, smoke detectors, and other targets at any arbitrary locations in multi-room compartments with different geometric configurations can be obtained. An example case is provided. Synthetic data generated for a wide range of fire scenarios from the example is utilized in a supervised machine learning technique. Preliminary results demonstrate that the proposed models can help to predict building fire hazards in real-time.
September 16-20, 2019
Suppression, Detection and Signaling Research and Applications Symposium (2019 SUPDET)