Objective - To develop and maintain robust, validated fire models—and the associated visualization and analysis tools—for
(1) performance-based design,
(2) forensics (fire reconstruction), and
(3) fire research applications in the built and natural environments.
What is the New Technical Idea? The development first of zone fire models, like the Consolidated Fire and Smoke Transport model (CFAST), and then high-fidelity, physics-based fire models, like the Fire Dynamics Simulator (FDS), has been driven by a need to better understand compartment fire dynamics for the purpose of protecting lives and property. CFAST and FDS continue to play a key role in performance-based design of buildings, saving billions of dollars annually in fire protection costs . Consequently, these modeling capabilities must be maintained, with the simulation tools evolving to keep pace with changes in computing technology. At the same time, advanced fire models like FDS are facing new challenges.
Starting with the World Trade Center investigation , and later the investigations of the Charleston Sofa Super Store  and the Rhode Island Station Night Club  fires, the NIST Fire Dynamics Simulator has been used to reconstruct flame spread behavior and tenability conditions in burning structures. To date, investigators have relied heavily on full-scale testing to generate realistic heat release rate curves to be used in the model. But full-scale tests are both extremely expensive and sometimes simply impossible to perform for the desired conditions (consider zero gravity conditions in space, for example). Improving accuracy and reducing uncertainty in forensic analyses, therefore, requires an ability to predict large-scale heat release rates.
Roughly speaking, fire modeling can be broken down into two distinct (but coupled) problems: (1) pyrolysis, or the thermal degradation of the solid fuel into a volatile gas, and (2) the combustion of the volatile fuel, which generates the heat that feeds back to the solid. This feedback loop is highly nonlinear. Thus, while the fire modeling community has a handle on each problem separately, when the problems are coupled—as they are when predicting full-scale heat release rates—the output from the models is unreliable.
Improving reliability will require advancements in the prediction of local heat feedback to the surface of the solid, which is generally controlled by radiation. And since radiation is largely controlled by local soot concentration and temperature, we will require improvements to local soot emissions prediction. In the solid phase we will require improvements to our ability to measure the appropriate material thermal and kinetics properties. We will also require improvements to our ability to account for the complexity of material geometry. For example, fire resistant coatings often intumesce, or swell, as a way of slowing the heat transfer to the burning solid. Additional complexities in the solid phase include melting, which can form secondary pool fires, and volatile transport in porous media. Neither of these phenomena are present in the current FDS solid phase model.
FDS was originally designed to model simple plumes and compartment fires where external boundary conditions are well defined. Over the past decade, however, FDS has been thrust into the role of modeling wildland-urban interface (WUI) fires [cite] and pollutant dispersion at the community scale [cite]. In these outdoor flows, the external boundary conditions are continuously varying in both space and time. To further complicate matters, for a given problem we rarely know precisely what the boundary conditions are—we must rely on estimates of mean winds and turbulence intensity. These estimates may be obtained from data stations inside the domain or from larger-scale simulations. In either case, the external boundary conditions are not supplied—the internal data must be assimilated into the FDS calculation.
While FDS has made modest progress in modeling outdoor flows at the community scale, several research questions remain. The principal issue is in how the wind field data is to be assimilated into the FDS calculation. FDS currently uses a form of data assimilation called nudging, which forces the mean wind field to track toward a specified target value—the specific value, as mentioned, comes either from measurements or a larger-scale weather model. There are three outstanding questions for the nudging model: (1) whether to use a global force or a local force field, (2) how the time-scale parameter is to be specified to correctly capture second-order flow statistics (wind gusts, etc.), and (3) how to eliminate spurious fluctuations at outflow boundaries.
In most outdoor flows the terrain is not a perfectly flat surface. In fact, most of the interesting—and potentially dangerous—wildfire flow situations involve slopes and canyons. So, in addition to challenges with external boundary conditions, FDS inherits the challenges that come with modeling flows over complex geometry. These issues are being studied in the context of complexities in compartment fire dynamics (see previous discussion on forensics), but atmospheric boundary layer flows have an added challenge in that the dynamics is dramatically affected by the global state of stability of the boundary layer (stable, neutral, unstable). And since the problems of interest for atmospheric flows involve time scales that span hours or days, FDS must be capable of handling the effects of the diurnal cycle, including variation in surface heating (night vs. daytime) and atmospheric moisture (humidity and clouds).
Modeling flame spread in wildland fires and WUI fires must also involve the modeling of ember transport. It has been well-documented [cite] that spot fires from embers are a dominate flame spread mechanism in wildfires. Further, WUI fires may spread by embers that come from burning homes as well as vegetation. FDS has the capability to transport embers as Lagrangian particles. These particles obey specified drag and heat transfer laws and may be linked to material properties that allow the embers to burn as they are transported. The primary research focus from a modeling point of view is verification and validation of the current modeling capabilities.
Flame Spread on Complex Materials:
The fire community has made considerable progress on measurement of macroscopic material properties and kinetics constants for “simple” materials. The most commonly studied material is a clear plastic known as PMMA (poly methyl-methacrylate). One of the reasons PMMA is popular with researchers is that it does not form a char, which simplifies the kinetics to a single one-step reaction. Additionally, PMMA does not intumesce (swell or expand) upon heating, so the geometry stays simple under controlled laboratory conditions. But even PMMA can melt and create a pool fire, thus changing the global heat release rate and altering the thermal feedback cycle in unpredictable ways. In real fires, even this simple melting problem renders the current FDS mathematical framework useless. Further, upon heating in a fire most common materials form char, foams, or add other levels of geometric complexity that make it impossible to translate bench scale physical property measurement to real-world application. Figure 2 shows the pyrolysis of polyethylene forming liquid and bubbles, which significantly impact the “effective” thermal properties of the material.
The immediate research need is to establish a baseline set of validation cases for the spread of fire over PMMA and similar polymers, pulling from experiments in the existing literature [Kashiwagi, etc.]. It has been demonstrated that FDS is capable of predicting the burning rate of these materials in devices such as the cone calorimeter. However, preliminary attempts to model PMMA flame spread [Leventon] have suggested that length and time scale resolution requirements are intractable for the current code—simple cases take months to run on high-performance computers. The current numerical framework in FDS utilizes structured Cartesian grids with fully explicit time integration. These two factors largely account for the computational workload. It is evident that improved numerical approaches will be necessary to achieve accurate and reliable flame spread calculations for complex materials in real-world engineering applications.
In parallel with work to characterize complex solids is work to improve the prediction of radiation heat flux from fires impinging on vertical walls. There are several examples of these fire scenarios in the current validation guide, but the accuracy of the model depends on a variety of factors, including the geometry, fuel type, and radiative properties of the fuel and exhaust products. The current gray gas or 6-flux narrow band model might need to be re-examined to determine if they can accurately predict near-wall radiative heat flux.
What is the Research Plan? In addition to general support for the public release of both FDS and CFAST, the project is separated into 9 research tasks, all of which are critical to the project’s objective. The first 6 tasks are FDS development. Task 7 is FDS support for a WUI grant. Task 8 is experimental work for FDS validation. And Task 9 is CFAST maintenance.
Task 1 Complex Geometry in FDS:
In previous years, we have developed a cutcell-immersed boundary (CC-IBM) method to handle complex geometry in FDS. The present task is to finalize the implementation in the FDS code trunk and to hook up all the boundary conditions, output options, and special features available to users in the baseline Cartesian code. This will include modification of the radiation solver to correctly treat surfaces with arbitrary orientation. The radiation solver work will be done in collaboration with Aalto University in Finland. The majority of the work for FY18 will focus on verification and validation tests for the complex geometry routines. The key deliverable from this task will be the release of FDS 7 beta.
Task 2 High-end Visualization in Smokeview:
The introduction of non-Cartesian geometry into FDS has necessitated an overhaul of the way Smokeview draws geometry. In FY18, further work is needed to support boundary files and slice files around complex geometry. It is critical that the visualization is a seamless transition for the user. In addition to complex geometry, FDS is continuously pushing the limits of computing power and generating ever larger datasets to be visualized. Smokeview is under continuous development to keep pace with these large data visualization needs. And as more outdoor flows are modeled, there is more and more a need for Smokeview to provide better ways to visualize wind data and to show terrain texture maps to better identify the community landmarks. This task is critical to FDS-SMV 7 beta bundle release.
Task 3 Flame Spread on Simple Materials:
In this task, we will capture recent work on modeling PMMA flame spread into the FDS Validation Guide. This will represent the only example of small scale flame spread. It is critical that we archive this work so that future modifications of the code do not erase the progress we have made to date on this problem. As discussed in the introduction of this proposal, the couple pyrolysis and gas phase combustion problem is currently extremely delicate, and also too costly for practical engineering applications. We expect this exercise to drive us to develop a more tractable and practical approach to flame spread modeling.
Task 4 Carbon Monoxide (CO) Prediction with Flame Suppression:
We continue our work on verification and validation of carbon monoxide (CO) prediction in FDS. Most fire deaths are attributed to CO poisoning. This task is aimed at significantly improving the accuracy and reliable of CO prediction, which is critical to effective use of these models for predicting and recreating fire scenarios. The chemistry submodel in FDS has been generalized to handle detailed chemical mechanisms. The present task is to develop an extinction model capable of handling multiple fast reactions. The goal is to implement both the multi-step CO mechanism and the multi-step extinction model as the default reaction scheme for all CO cases in the FDS validation suite, from small scale (Smyth burner) to a full-scale compartment (NIST Full-Scale Experiments)
Task 5 Development of an Improved Local Radiant Fraction Model:
It has recently been discovered that FDS over-predicts the global radiant fraction for high resolution simulations using the default method for radiant emission. This method applies a correction factor to the emission term in the radiation transport equation in an attempt to force the global radiant fraction to integrate to a predefined value. Since the convective heat release dominates the plume dynamics, it is critical that FDS gets the global radiant fraction correct. It is particularly challenging to design a method that properly handles both simple plumes and under-ventilated compartment fires.
Task 6 Development of a Parallel Fast Fourier Transform (PFFT) Pressure Solver:
One of the limitations of the FDS pressure solver is that it only operates on a single mesh or domain block assigned to one processor of a distributed memory parallel computer (super-computer). While the discrete FFT algorithm is nearly optimal for use in solving the pressure Poisson equation on a single mesh, extending this algorithm to parallel computing is not trivial. However, PFFT solvers have been successfully implemented in other codes and would eliminate mesh-to-mesh velocity errors for simple block domains that usually used in outdoor flows.
Task 7 Wildland Fire Spread Validation:
There is presently an effort to merge back the functionality of the Wildland Fire Dynamics Simulator (WFDS) into the latest FDS trunk. Over the past year, several simple verification tests for drag and mass loss rates from subgrid Lagrangian particles—which represent vegetation in the model—have been implemented into the FDS V&V suite. In FY18, we will continue to add to the V&V suite and develop the input hooks and modifications to the FDS pyrolysis model needed to accommodate the multi-step kinetics used for evaporation, burning, and char oxidation in the WFDS vegetation model.
Task 8 Experimental Support for Measurement and Computation of Fire Phenomena (MaCFP):
The FDS development team is involved in a community effort to improve physics-based computational models and modeling practice. The effort is called the working group on Measurement and Computation of Fire Phenomena, or MaCFP. In FY18, we will start to contribute data for burning rates of liquid pool fires (methanol, ethanol, etc.) with detailed heat flux measurements.
Task 9 Improved Radiation and Tenability in CFAST:
This task is to add improved accuracy and functionality to the CFAST model. CFAST is a two-zone fire model used to calculate the evolving distribution of smoke, fire gases and temperature throughout compartments of a building during a fire. CFAST is a vital design tool for the engineering community because its calculations can be run in minutes as opposed to hours or days for FDS. Over the last couple years, CFAST has been updated to be in compliance with modern Fortran standards and combustion chemistry that is more consistent with FDS. This task is to improve the computation of radiation view factors to targets in CFAST, beginning with a suite of verification and validation cases that can be used to test any proposed improvements to the model. In addition, the task will add calculation of tenability consistent with ISO standards 13571 using existing outputs from the CFAST model.