Model Citizens: Open-Source Fire Research
A team of FM Global researchers, with help from scientists and academic institutions from around the world, has completed a five-year effort to develop cutting-edge fire modeling software.
The FireFOAM modeling software is an important new tool for researchers. It will allow FM Global to make the most of its large-scale tests and give its clients the very best advice on how to mitigate risk.
The software integrates key physical models relating to fluid mechanics, heat transfer, combustion and multiphase flows. The model is the first to incorporate the complex relationship between fire and water when a fire-suppression system is activated. It can predict the impact a fire-suppression system will have on a fire’s fuel source and on the fire’s growth, allowing FM Global to make more accurate fire protection recommendations.
The software takes advantage of advances in mathematics, modeling and computing, including modern object-oriented code structure, multi-physics modeling approaches, unstructured meshing and parallel computing. The software can create accurate and robust simulations in the areas of fire growth, material flammability and fire suppression.
“Fire modeling won’t replace fire testing. It will allow us to develop a better test, learn more from the test results and apply those results to a wider range of fire scenarios.”
The project was launched in 2008 by FM Global’s Dr. Sergey Dorofeev, research area director, fire hazards and protection. He saw modeling as an important complement to the large-scale fire testing research FM Global has always done.
Modeling allows FM Global to design better large-scale sprinkler protection tests. Simulating multiple scenarios using different protection schemes, water supply pressures, and sprinkler configurations, sizes and locations would point FM Global researchers toward the optimum protection configurations for various stored commodities. Large-scale tests could then be performed to validate the protection concept.
Modeling also addresses some of the limitations of large-scale tests, such as applying the test results to variations in the storage scenarios. It can be used, for example, to simulate a test with a taller storage rack or a scenario where the fire starts in a different location. These simulations greatly expand the applicability of the knowledge gained through large-scale fire tests.
The result is fewer large-scale tests, transferrable data and more reliable fire protection recommendations, helping FM Global clients reduce risk.
“Fire modeling won’t replace fire testing,” explains Dorofeev. “It will allow us to develop a better test, learn more from the test results and apply those results to a wider range of fire scenarios.”
But developing such a model was a daunting task. Fire itself and the interaction of water and fire are so complex, Dorofeev expected it would take decades to create an effective model. He set a five-year strategic goal of developing FM Global’s fire model with a level of predictive capabilities that would allow FM Global to reduce its reliance on large-scale fire testing in fire protection engineering. With this practical five-year goal, he began assembling a team of researchers to do the work.
“In research, five years is an eternity,” explains Dr. Karl Meredith, the technical team leader in fire modeling, who joined FM Global as a senior research scientist to work on this project. “Researchers usually work on grants that last six months or a year. It was impressive that FM Global had the foresight into the effort that would be involved and dedicated the resources needed for this kind of effort. Professionally, it was exciting to come here and work on a project like this.”
Dorofeev also wanted to tap the knowledge, expertise and enthusiasm of the academic world and embraced an open-source modeling framework. This allowed researchers from around the world to contribute, as well as benefit, from the research being done.
He coordinated his team’s efforts with the work already being performed at the University of Maryland (USA), Kingston University in London (U.K.), the University of Edinburgh in Scotland, Ghent University in Belgium and Worcester Polytechnic Institute in Massachusetts (USA). Industry and governmental labs, including Sandia National Laboratories in New Mexico (USA) and Oak Ridge National Laboratory in Tennessee (USA), also contributed to the project.
“We decided to create the model open source so we could leverage our effort and develop the science,” Dorofeev said. “We needed to understand the physics of the phenomena involved, use prior knowledge and study the key ingredients that needed to be in the models. By making it open source, we could turn to universities and industry for their insights. It gives everyone a chance to contribute and certainly gets everyone excited about working on it.”
The open-source effort has led to the annual Open-Source Fire Modeling Workshop in Norwood, Mass., USA. The event brings leading fire modeling researchers and scientists together for two days of workshops and presentations. Last year, attendees came from five countries and included representatives from Shell Research, the (U.S.) Federal Aviation Administration and eight universities. Attendees participate in workshops detailing the latest research being done at FM Global and advances made in fire modeling around the world.
Rack Storage Protection
The fire modeling Dorofeev envisioned would require an unprecedented amount of research. The team decided to start that research focusing on the rack storage of commodities in cardboard boxes. This represented the most common bulk-storage scenario of FM Global’s clients and would create the most applicable data. Much of FM Global’s large-scale fire testing looks at this storage configuration, which leads to sprinkler recommendations for a variety of commodities.
The research was organized into three main areas: fire modeling, material flammability and water suppression. Teams tackled each of these areas independently and cooperatively, conducting tests and developing models for the key physical phenomena that occur in a fire. Eventually, all the research was brought together into the computer model named FireFOAM.
Initial work originated in the fire modeling team, where Dr. Yi Wang, research group manager, wrote the FireFOAM solver. It was capable of describing the dynamics of gas-phase fire and fire spread over solid fuels.
“We had to capture all the physics, but simplify it enough so that we could run a model in a reasonable amount of time,” Wang said. “We really started from scratch. When we started I didn’t think we could get where we are today in five years.”
Meredith and his team tackled water transport modeling, the physics of water suppression and its application in rack storage protection. His work incorporated a multitude of physical properties, including splashing, spray, evaporation, absorption, heat transfer and the way water pools and runs off cardboard boxes.
“All of these things are governed by physics, and we had to identify which ones were the dominant factors that control the fire,” Meredith said. “Just in water transport there were 14 model areas and each of those areas had a number of sub-models. And all the other teams had similar numbers of models and subsets that they were looking at.”
Technical team leader Dr. Marcos Chaos led the material flammability research exploring the properties of cardboard and other fuels and modeling the chemical breakdown of the material as it burns. His team also conducted small- and medium-scale experiments on flame spread to help validate the model.
Dr. Yibing Xin, the technical team leader in water suppression, led the sprinkler technology program. His team explored the physics of water suppression in a variety of intelligent fire tests. They studied sprinkler properties, measuring water flux, droplet sizes and spray density.
Each area of research required bench-, medium- and large-scale tests for model development and validation. This work could not be completed without the expertise of the personnel from the FM Global Research Campus in West Glocester, R.I. (USA).
Once all the model pieces were complete, it was up to Yi Wang and his team to bring it all together. He was responsible for assembling the FireFOAM code and evaluating the integral model. The model was put to the ultimate test in 2012. FireFOAM was used to model a rack storage sprinkler suppression scenario at full scale. The large-scale test was run and the results compared with the model output. The model produced remarkably similar results.
“Bringing this modeling tool to fire protection was very challenging,” Dorofeev said. “There are so many different aspects to it. It was extremely complex.”
FireFOAM is continuing to be improved and may never truly be finished. In many ways, the research has only just begun. The fire characteristics of the commodities stored have not been fully addressed. There are thousands of materials being stored in a multitude of ways all over the world, making it important to classify them into a finite number of standard commodities, with the corrugated cardboard box being one of the most dominant forms of storage.
“We’ve only started with the standard cardboard box; we haven’t put anything in the box yet,” explains Chaos, the flammability team leader. “We know that most fires of cartoned materials are initially driven by the cartons themselves. So it’s very important that we understand fire growth with cardboard so that activation of the first sprinklers can be predicted.”
“But the contents of the boxes will eventually become fuel for the fire, so that is a very important piece as well. We are currently working to add content to the boxes that represents our standard plastic commodity and try to understand that.”
Another effort moving forward will be to enhance the current model, refining the calculations and getting the model results even closer to those of the large-scale tests. The team will also be looking to develop new models and apply the current modeling software to fire research now being done. The applications of the current model include in-rack sprinkler protection, roll paper, data center protection and the rack storage of complex commodities.
The in-rack sprinkler project is already under way. Small-scale tests are being conducted to understand in-rack sprinkler characteristics; and medium-scale tests are ongoing to determine the water density needed to suppress the fire. Modeling is used to develop the optimum projection concept, including sprinkler type, their locations and the water pressure needed. Large-scale testing will then be conducted to validate the protection concept. The results will ultimately help create the data sheets that will drive the recommendations FM Global makes to its clients.
“Having an efficient model at hand changes the paradigm of fire protection design, which is currently based on multiple large-scale fire tests. The new ‘test-model-validation’ paradigm is based on smaller and smarter tests, modeling to optimize protection and limited large-scale validation. Modeling can help us get closer to the answers,” Dorofeev concludes.