AIoT-powered building digital twin for smart firefighting and super real-time fire forecast
Weikang Xie, Yanfu Zeng, Xiaoning Zhang, Ho Yin Wong, Tianhang Zhang, Zilong Wang, Xiqiang Wu, Jihao Shi, Xinyan Huang, Fu Xiao, Asif Usmani
Abstract
• Propose a framework of AIoT-integrated Digital Twin for the full-scale three-story building. • Apply ADLSTM-Fire model to transform point-sensor arrays into spatiotemporal temperature field. • Super real-time fire forecast of hazardous floor regions to support smart firefighting and rescue operation. • Full-scale building fire experiments demonstrate smart system validity and generalization capability. Complex dynamics inherent of building fire poses big challenges to firefighting and rescue, especially with limited access to critical fire-hazard information. This work proposes the novel AIoT-integrated Digital Twin for the full-scale multi-floor building to manage the dynamics fire information. This system allows for super real-time mapping of actual building fires into accurate and concise digital fire scene at the cloud platform. By developing the ADLSTM-Fire model, we effectively transform discrete sensor-array data into high-dimensional spatiotemporal temperature fields in real-time, and furthermore, forecast future fire development and hazardous regions 60 s in advance. By comparing with benchmark numerical simulations, the Digital Twin system demonstrates the high reliability of super real-time fire-scene reconstruction and the capacity of fire-risk forecasting in supporting firefighting. The full-scale building fire experiment is employed to validate the generalisation capability of the proposed smart firefighting method. This work demonstrates the great potential and robustness of AIoT and digital twin in support smart firefighting and reducing fire casualties by information fusion.