Modelling of building fires coupled with turbulent, combustion, soot chemistry and radiation effects
綜合考慮湍流, 燃燒, 煙氣生成及幅射影響之建築火災模型
Student thesis: Doctoral Thesis
Currently, with the rapid advancement of the computer power, computer models for simulating smoke and fire movement in building fires are widely adopted among fire researchers or engineers. As increasing efforts have been devoted to incorporate sophisticated sub-models into the field model, researchers discovered that radiation and soot formation modelling are of very significant importance in building fire simulations. Modelling radiation effect and the soot augmentation has now become the centre of fundamental fire modelling research. In this thesis, a self-developed numerical model which is tailored for studying the behaviour of building fires has been developed and validated with full-scaled experimental data. The numerical model attempted to completely simulate the simultaneously occurring flow, turbulent, combustion, soot chemistry and radiation effect in the building fires. Numerical study on three widely adopted empirical/semi-empirical models has been carried out. Computational results with or without soot consideration are validated against a full-scale experimental data. The performances of the three soot models are evaluated in terms of prediction errors. In general, semi-empirical soot model by Moss et al. (1995) and Syed et al. (1990) produced the best agreement with experimental data. The model is considered as a suitable candidate for considering soot production in compartment fires. The results also demonstrated the feasibility of applying the semi-empirical soot models for simulating realistic compartment fires. Parametric studies on three different SN order schemes of the Discrete Ordinates Method (DOM) were also performed. The optimum choice of approximation order was investigated through sensitivity numerical studies of the S4, S6 and S8 approximations, which correspond to 24, 48 and 80 fluxes respectively. The predicted temperatures of the three approximations were compared with experimental data reported by Steckler et al. (1982) and Nielsen and Fleischmann (2000). The study proved that the optimum choice of approximation order depends highly on the soot generation and augmentation. The S4 scheme appeared as the optimum choice for the methane fire in Steckler’s experiment while S6 scheme was the best option for the LPG fire in the two compartments structure. Difficulties in resolving the different combustion and flow length scales and the need of Local Grid Refinement (LGR) in building fire studies had been discussed. Numerical techniques for implementing the LGR had been extensively reviewed. The Nested Overlapping Grid (NOG) algorithm was proposed. Interface treatment by Liu and Shyy (1996) was extended for compressible flow and employed for passing information between overlapping grids as boundary condition. The NOG technique was then critically validated by several well-known benchmarked problems. Predicted results agreed well with the benchmark solutions. Significant reduction of the required computational time and storage was found by using the NOG technique. Capability of the algorithm in resolving the different length scales and improving prediction accuracy had also been tested and clearly demonstrated. Finally, the NOG technique was applied to simulate the smoke movement and fire growth in buildings. The computed results compared well with the measured data. In addition, the results also exemplified the algorithm in capturing the important microscopic information with the critical region (e.g. the fire plume) whilst maintaining adequate macroscopic simulation of the other non-critical regions. The required computing time to solve the fire problems were found to be significantly reduced comparing the traditional approach of globally resolving the computational domain covering both the critical and non-critical region. The applicability and robustness of applying the NOG technique in building fire studies have been confirmed.
- Mathematical models, Fires