Smartfire future developments

The SMARTFIRE group undertake considerable and diverse research into issues associated with fire, combustion and CFD modelling technology. This research includes:

The group is involved in applying all of these capabilities in aviation, building, rail and marine environments. This work is often undertaken in collaboration with other research organisations. For an indication of some of the projects we are working on visit our FSEG Research Projects pages.

Listed below are a range of new features that are currently being developed. Most of these features exist in prototype format within the SMARTFIRE software and are scheduled for release in subsequent versions.

Group Solvers / Adapted Group Solvers: Back

Group solvers are a means of reducing computational overheads in fire simulations. In this way group solvers make fire modelling more accessible to the fire engineer and to the education environment. The group solver is used to obtain numerical solutions to the algebraic equations associated with fire field modelling. The purpose of the technique was to reduce the computational overheads associated with traditional numerical solvers, typically used in fire field modelling applications, by fine tuning the amount of numerical processing that is performed on particular geometric or logically related groups of cells.

As part of the work undertaken, it has been demonstrated, using the group solver technique, that a reduction of around 37% of the processing time can be achieved for complex fire simulation scenarios. Group solvers are a conceptual extension of the simple linear, iterative, algebraic equation solvers usually referred to as JOR or SOR. These solvers involve the repetitive updating of the solution of a conserved quantity within each cell based on the contributions from nearest neighbouring cells, a portion of the previous solution value and the source quantity for each cell. The contributions from neighbouring cells represent the convection and/or diffusion of a physical property throughout the solution domain whilst the source is the creation or destruction of the physical property in a cell. In the typical whole-domain JOR or SOR solver, the solution in each and every cell of the domain is updated repetitively until the difference between successive updates is sufficiently small.

Clearly, if the solution domain contains many cells that are far removed or de-coupled from any active flow region for a portion of the simulation, then not all of these JOR or SOR calculations are performing any useful advancement of the solution. This is especially true of many of the large complex geometries often used in fire field modelling (e.g. whole building simulations). The group solver concept allows the domain to be partitioned into "geometric" or "logical" (i.e. solution dependant) groups of cells that then use the iterative point-by-point update procedures described above. The difference for group solvers is that each group can have a unique set of control parameters to configure the maximum number of iterations, convergence tolerance and/or linear relaxation to be used. Since, in an unstructured code, a group does not need to be limited to some pre-configured geometric region it is possible to further extend the group solver techniques by allowing groups to determine their own cell-membership as the solution develops.

In many large scale fire scenarios there are likely to be remote areas of the domain which may not experience significant fire effects (due to ventilation paths) until late in the simulation or there may be areas which are fully decoupled from the main flow of the system (for example closed rooms). The exploitation of such scenario features has the potential to massively reduce run times and memory usage. The development of Adapted Group Solvers has enabled larger geometry CFD cases to be simulated on serial systems, with savings in memory usage and runtime. These technologies use the physical characteristics within the solution domain to group the cells and control the solution process using predefined parameters. Each group is then solved using the most appropriate solver criteria to attain a converged solution.

Methods that have been investigated include Restricted Domain Method (RDM), Zone Applied Propagation (ZAP), Free Group Solvers (FGS), Group Re-indexing and Flow Finder. Through identification of computationally expensive processes - both within the solution algorithms and areas of the physical domain - Adapted Group Solver methods have been applied to limit the overheads and produce savings in solution run-time of between 18% and 53%, with less than a 1% reduction in accuracy.

A multi-storey geometry used to test the various Group Solver techniques.	Velocity results used to compare Group Solver consistency.
Performance of the various Group Solver technologies.

Benefits of Group Solver Research - Adapted Group Solver optimisations allows for:

The Group Solver techniques and User Interface are currently undergoing testing and final developmental fine tuning.

CAD Interface Enhancements (the SMARTFIRE Scenario Designer): Back

While the CAD Interface was introduced into SMARTFIRE V4.0, research in this area continues. It’s main aim is to further develop feature recognition techniques to assist in importing of buildings from external CAD files (commonly used DXF file) for the solution of fire scenarios being simulated using SMARTFIRE. These developments are intended to improve project turn around times.

The semi-automatic approach of importing CAD drawings would reduce the initial overheads involved in geometry specification phase. The aim of the technique is to automatically recognize more of the drawn objects such as doors, windows, walls, etc, for both layered and non-layered CAD files. These objects will then be combined to form the room objects, which are further combined to from building objects. Results from the research have shown that the quality and nature of the initial CAD drawing is critical to the level of automation possible. An interface, and the underlying feature recognition techniques to address the above task, have been implemented in the Scenario Designer. This is currently being used to assess various potential methodologies that would provide a usable and practical tool for engineers with both “old” and “new” CAD drawings. The final phase of the task, allowing the alteration of the imported building objects in the SMARTFIRE Case Specification component, has also been implemented and tested on a range of cases. Work is currently progressing to enhance the methodology so that more extreme forms of DXF floor plans can also be handled.

2D-CAD drawing of multi-room structure being passed through the SMARTFIRE CAD interface with third dimension being added.

Section of multi-room structure to be modelled in SMARTFIRE has been extracted and meshed.

Automated Solution Control : The Intelligent Control System (ICS) and the Experiment Engine: Back

This research is directed at the use of Artificial Intelligence (AI) techniques to assist in dynamic solution control of fire scenarios being simulated using SMARTFIRE. Its main aim is to improve the automatic convergence capabilities of the software while further decreasing the computational overheads. The technique automatically controls solver relaxation using an integrated production rule engine with a blackboard (current solution status) to monitor and process the required control changes as the solution progresses. Results from the research have showed great potential for considerable savings in simulation run-times when compared with control sets from various sources, e.g. CFD experts, etc., and furthermore higher accuracy solutions are obtained in the process. The approach demonstrates enhanced solution reliability due to obtaining acceptable convergence within each time step as compared to the conventional approaches. The control architecture is described in the figure above. There are two alternating stages in Intelligent Control System (ICS)-controlled simulations:

Monitoring Phase, where no control actions are performed but the time steps are monitored for signs of developing problems
Search Phase, where the modifications are proposed and tested. The search results are then evaluated and eventually the best control parameters applied.

In complex three-dimensional fire modelling examples, this approach has reduced the overall iteration count by 50%, which in return is an overall saving of some 50-60% of the CPU time. Furthermore, solutions are generated that have meet the convergence criteria 100% compared to traditional approaches where solution may reach the maximum number of iterations within a time step and simply move onto the next time step.

The second development is the new Experiment Engine. This is a computational module that can act as an expert assistant to provide high level solution control from the point at which the geometry and the nature of the simulation have been specified. The concept here is that the Experiment Engine will be able to run a coarse mesh, simplified model to determine such issues as the likely stability, ranges of data and solution complexity. This information can then be used to enhance the performance of the mesh generation (via cell budget control and local refinement) and the subsequent numerical processing and solution control (such as configuration of time stepping and relaxations).

Unstructured Meshing: Back

At the heart of SMARTFIRE is an unstructured mesh CFD code. This means that SMARTFIRE can solve the fire equations on very complex computational meshes. This enables the software to model fires in complex irregular structures such as large warehouses with pitched roofs or the extremely complex environment of an aircraft cabin. Furthermore SMARTFIRE will be able to support any type of computational mesh composed of arbitrary polyhedral volumes including but not limited to tetrahedral and hexahedral bricks. However, construction of some unstructured meshes may require the use of third party meshing software. Work is currently underway to develop user-interfaces that allow mesh specification to be performed in as simple a manner as the current regular mesh generation is within SMARTFIRE.

Unstructured tetrahedral mesh elements (Pyramids), has been used in the unstructured mesh version of SMARTFIRE to simulate the effects of fire within aircraft. The mesh was generated using a third party mesh generator.

Simulation of fire within an aircraft cockpit using an unstructured mesh. Depicted is a VR representation of the aircraft with a temperature isosurface. Some notable features in the geometry include the pilots chair and ducting in the above ceiling region. Visualised using MayaVi.

SMARTFIRE V4.0 simulation of the Swiss Air MD11fatal aircraft fire. SMARTFIRE used to simulation the spread of fire behind the ceiling panel. Eventually smoke enters the flight deck.
Real Media (low quality) [34 sec]
Real Media (good quality) [1.87 MB]

Fire Product Deposition Modelling: Back

Fires typically generate a number of narcotic and toxic gasses. Sometimes these products of combustion have complex behaviour which means that the dispersion and transport of the products (within the flow domain) is non-trivial. For HCL, the gas reacts with certain wall surfaces to effectively remove it from the gaseous phase. This needs to be accurately modelled in order to predict the gaseous quantities of HCL which can be convected into other regions of the geometry and so cause harm to occupants.

Galloway and Hirschler’s deposition model, typically used in zone models, is modified and applied to a field fire model in order to predict the decay of HCL within fire enclosures. The modified model still uses empirical formulas, but the HCL deposition mechanisms have been simplified from three processes to two processes (see Figure below).

The effect of HCL flux to the wall boundary, on the time to reach equilibrium (i.e. for the wall surface HCL density to rise and reach the equilibrium) is addressed in this modified model. The new model correctly predicts the absorption of HCL by wall surfaces with the SMARTFIRE Fire Field Model.

SMARTFIRE modelling HCL absorption showing surface map when using PMMA walls

SMARTFIRE modelling HCL absorption showing surface map when using Concrete walls

Coupled CFD Fire Field Modelling and Zone Modelling: Back

In very large geometries (e.g. airport terminals, underground stations) there may be regions of the geometry which will only experience minor fire impact or which may be of little direct relevance to the solution outcome. In such cases it may be desirable to use a coupled field-zone modelling approach to simplify the calculations and reduce the computational overhead associated with simulations based on a pure field modelling approach. The aim of this project is to develop such a coupled field-zone modelling approach using SMARTFIRE as the core field modelling engine.

Enhanced SMARTFIRE to EXODUS data linkage: Back

The current SMARTFIRE to EXODUS data linkage relies on the creation of appropriate sub-zones within both models. These zones might be very large by comparison to the control volumes used within the CFD model. In such instances there might be considerable inaccuracy (due to data averaging) within the zones, particularly when the zones are close to the fire sources or where complex flows are likely within a zone.

An enhanced SMARTFIRE to EXODUS data linkage is being developed which attempts to preserve more of the computational accuracy of the CFD modelling. The evacuation software would have access to a database of results data for each time step which it can interrogate to determine arbitrary zone averages. These arbitrary zones could be as small as a single evacuation node space.

Sprinkler and Water Mist Modelling: Back

Sprinkler and Water Mist models developed by FSEG for the EU funded FIREDASS project has been ported to the SMARTFIRE software. Ongoing research is aimed at investigating droplet interaction with the gas environment, droplet interaction with burning surfaces and suppression effects. Although Sprinkler and Water Mist system modelling is now available, Fire Suppression and enhanced coupling between sub-models will be made available to users in future releases of the software after thorough validation has been performed.

Investigating the effect of sprinkler head design for Water Mist modelling within SMARTFIRE Research Version.

Solid Fuel Combustion: Back

Research undertaken by the SMARTFIRE development team into the development of solid fuel combustion models based on the pyrolysis process and including charring effects is being incorporated into the SMARTFIRE environment.

Continuing Research into Toxic Gas Generation: Back

Research by the SMARTFIRE developers continues into the application and development of toxic gas generation models based on the Local Equivalence Ratio (LER) concept. The LER based toxicity model has been fully implemented in SMARTFIRE and is currently being extended to model other gaseous species.

The model was first used to predict the CO, CO₂ and Optical Density of smoke for a combined CFD analysis and evacuation model performed in SMARTFIRE and EXODUS.

The Equivalence Ratio φ, is defined as the fuel mass flow rate divided by the air mass flow rate normalized by the stoichiometric ratio of fuel to air. φ is used to describe the vitiated condition within a fire room Correlation between species yields and φ. Experimental data support that the yields of combustion products are correlated with φ. A Correlation between species yields and φ have been derived from experimental data for various building materials. This yield correlation is used to predict the generation of combustion products in room fires. Preliminary numerical tests showed that this approach was very promising.

The model performed very well in the Cable fire scenario with accurate prediction of the product species.

SMARTFIRE modelling CO mass fractions from a cable fire

Graph showing predicted and experimental CO2 volume fractions for a small room fire test scenario.

New Features planned for SMARTFIRE v4.2 Back

A range of new features for SMARTFIRE V4.2 are currently in development. More information concerning these features will be provided here in the near future.