Deep learning for the next generation (highly sensitive and reliable) ECLSS fire monitoring and detection system

Abstract

Fire detection is a critical component of the Environmental Control and Life Support System (ECLSS) on board space habitats and remains an important research area with considerable practical relevance. Significant advances have occurred over the years in ECLSS design and automation, and its operation has become more complex and requires ever more effective environmental monitoring systems in the space habitat. In this work, we develop a novel fire detection method using deep Long-Short Term Memory (LSTM) neural networks and variational autoencoder (VAE) to meet the increasingly stringent requirements of fire detection for future space habitats in terms of sensitivity and reliability. To examine the performance of our method, termed Deep Learning for Environmental Monitoring (DeLEM), and specialized for Fire Detection in this work (DeLEM-FD), we develop a series of computational experiments using a high-fidelity fire simulator. We evaluate and compare the performance results of our proposed fire detection with alternative methods, including the standard LSTM, cumulative sum control chart (CUSUM), exponentially weighted moving average (EWMA), and two currently used fixed-temperature heat detectors. The simulation-based results indicate that the DeLEM-FD robustly outperforms the other detection methods with statistically significant shorter alarm time lags, no missed detection, and no false alarms. In future work, we propose to examine the performance of our detection method on real-fire experiments conducted by the National Institute of Standards and Technology.