Paper Reports on Using Detailed Chemistry to Advance the Fire Dynamics Simulator
“Detailed Fire Chemistry with Fire Dynamics Simulator” has recently been published and presented at Interflam 2025. Key findings from this study support the “Fire Modeling Development and Validation” project, led by UL Research Institutes’ Fire Safety Research Institute. Jason Floyd, principal research engineer, co-authored this paper with Chandan Pal from the National Institute of Standards and Technology (NIST) and George Washington University, and Randal McDermott from NIST.
Fire Modeling within Fire Protection Engineering
Fire modeling is an important aspect of fire protection engineering design and fire research. It is used to design safer buildings, develop fire protection systems, and analyze previous fire events. Understanding various quantities, such as carbon monoxide and soot, produced by a fire is important. However, in the Fire Dynamics Simulator (FDS), a current, practical, and widely used computer model, predicting these quantities is challenging.
Currently, FDS uses either one- or two-step, fast chemistry for practical applications. The one-step method uses fixed post-flame yields for carbon monoxide and soot. This performs well for well-ventilated fires, provided the modeler selects an appropriate set of yields for the fuel and fire size being modeled. The two-step method first oxidizes fuel carbon to carbon monoxide and soot, and then, if oxygen remains, it oxidizes the carbon monoxide and soot to carbon dioxide. This model increases production of soot and carbon monoxide in under-ventilated fires; however, it relies on an empirical parameter to define the initial split between soot and carbon monoxide and still relies on post-flame yields for the second step.
A long-term goal for FDS and other fire models is to predict, rather than prescribe, both the production of soot and carbon monoxide yields and the heat release rate of a fire. These are interdependent. Soot production and incomplete combustion affect radiative and convective heat transfer to surfaces. The burning rate of surfaces impacts local oxygen concentration and the potential for local or global extinction, as well as increased soot and carbon monoxide production. Reignition events following extinction can lead to backdrafts or smoke explosions (a proximate cause for many firefighting line-of-duty injuries or fatalities).
At practical engineering scales, models still struggle with these predictions. In FDS, this struggle involves specifying a number of empirical parameters for which there is limited guidance available for real-world fuels.
Further Developing the Fire Dynamics Simulator
Until recently, it has been challenging to delve deep into the details of extinction and reignition, soot formation, or other phenomena related to detailed chemical processes for engineering-scale simulations. The computational cost was prohibitive. That is no longer the case. The advent of relatively inexpensive, high-performance computing opens the possibility of incorporating more detailed chemical information into simulations. While still computationally intensive, a skeletal chemical mechanism (a few tens of reactions rather than one or two) is now tractable. The latest development of FDS 6.10 has enabled FDS to model detailed chemical mechanisms. This opens a pathway to improving the prediction of carbon monoxide, soot, extinction, and reignition. With FDS 6.10, there is now a pathway of highly detailed chemistry at small scales to inform the development of skeletal mechanisms for engineering scales.
“The ability to conduct detailed chemistry with the FDS opens up a number of avenues for future research and improvement of predictions of under-ventilated and post-flashover fires.”
–Jason Floyd
Principal Research Engineer
UL Research Institutes | Fire Safety Research Institute
This paper consists of three major parts. The first section presents an implementation discussion on how the CVODE solver was integrated into the FDS. This new tool was added to assist in handling complex chemical reactions. The second section presents a series of zero-dimensional problems comparing the FDS to the Cantera tool for the full GriMech 3.0 mechanism. The verification tests demonstrate that the FDS can accurately predict fire ignition, the behavior of chemical reactions over time, and the end states for varying initial conditions of fuel, air, and temperature. Additionally, to accelerate chemistry calculations and avoid some processes sitting idle, an optimized load-balancing strategy was implemented, and the results of this strategy are presented. Finally, in the third section, FDS is applied to the modeling of bench-scale diffusion flames in 2D and 3D. Two geometries are considered for these experiments, demonstrating the FDS's ability to predict the extinguishing concentrations for methane and propane flames with nitrogen or carbon dioxide as the diluent.
While more work is needed to better understand the grid resolution requirements for various mechanisms, the results of this effort demonstrate the great promise of these new capabilities.
