APPENDIX 3 : Conference Paper "The Development and Application <strong>of</strong> Group Solversin the SMARTFIRE Fire Field Model", Ewer J., Galea E., Patel M. and Knight B.,Proceedings <strong>of</strong> Interflam '99, Edinburgh, UK, 1999, Vol. 2, pp 939-950.efficiency may be desired, this result was anticipated because there are still many calculationsperformed in the “de-coupled” region for material properties and simple calculated variables.It is anticipated that this figure can be improved somewhat by increasing the use <strong>of</strong> “group”activated calculations within the rest <strong>of</strong> the CFD code.In Case 3 both static and dynamic groups are used with the majority <strong>of</strong> the extended regionbeing continuously evaluated for applied processing strategy. In this case an overallprocessing time saving <strong>of</strong> 5h 54m 55s (or 37.3%) was achieved when compared to thestandard whole field SOR solvers processing all cells equally. It should be noted that much<strong>of</strong> this saving is due to the “de-coupled” void group which, as shown in Case 2, saves 26.1%<strong>of</strong> the processing. The remaining 11.2% saving is due to the optimisation <strong>of</strong> processingwithin the extended region which targets less solver processing in cells with relatively lowvelocity flow. The fact that this saving is comparatively less than for the “de-coupled”region is also anticipated. This can be explained by considering the work performed in the“de-coupled” and dynamic groups. In the “de-coupled” group, it was not necessary to buildthe system matrix coefficients for the member cells whereas cells in a group that performsone (or more) iterations must build the system matrix coefficients. Building the systemmatrix coefficients is relatively costly compared to solving the matrix .CONCLUSIONSThe results indicate that there are large potential savings to be gained in thesimulation <strong>of</strong> fire modelling scenarios by the targeting and optimisation <strong>of</strong> processing effortin fully de-coupled, suitably stratified or geometrically related flows. Furthermore, thesesavings need not result in compromised accuracy <strong>of</strong> the final solution. The techniquesdeveloped and presented here resulted in considerable run-time savings <strong>of</strong> up to 37% <strong>of</strong>processing time. It is anticipated that this figure can be improved significantly when a betterunderstanding <strong>of</strong> the balancing required between groups and variables is achieved.As group solvers are a new concept, there was little or no expertise to guide in the optimalselection <strong>of</strong> number <strong>of</strong> groups to use, the choice <strong>of</strong> group membership conditions and therelative amounts <strong>of</strong> processing used in each group. Furthermore there are a number <strong>of</strong>remaining group solver control options which were not varied during the test simulations.It is anticipated that in large scale simulations, which may involve whole buildings, there arelikely to be much greater savings possible with intelligent use <strong>of</strong> group solvers that can targetthe processing only on the active flow and fire regions until the solution characteristics inother regions become significant.Current research effort is directed at gaining a better understanding <strong>of</strong> when it is appropriateto use groups and how best to balance the processing between groups in order to obtainoptimal convergence and simulation times. Dynamic groups have been shown to givemodest performance improvements but more work is needed to determine if there are anyfurther benefits possible due to combined solution monitoring and dynamic knowledge basedcontrol <strong>of</strong> the processing within both the static and dynamic groups. Whilst the use <strong>of</strong> groupsolvers increases the complexity <strong>of</strong> the knowledge based control, it is also most likely toprovide the most significant savings and most reliable solutions.Appendix 11.3 Page 143-11 11
ACKNOWLEDGEMENTSAPPENDIX 3 : Conference Paper "The Development and Application <strong>of</strong> Group Solversin the SMARTFIRE Fire Field Model", Ewer J., Galea E., Patel M. and Knight B.,Proceedings <strong>of</strong> Interflam '99, Edinburgh, UK, 1999, Vol. 2, pp 939-950.The authors gratefully acknowledge the financial support <strong>of</strong> the UK EPSRC throughresearch grant GR/L56749 and the Loss Prevention Council.REFERENCES1. Galea E.R., “On the field modelling approach to the simulation <strong>of</strong> enclosure fires ”, Journal<strong>of</strong> Fire Protection Engineering, vol 1 (1), 1989, pp 11-22.2. Taylor S., Petridis M., Knight B., Ewer J., Galea E.R. and Patel M., “SMARTFIRE: AnIntegrated CFD code and expert system for fire field modelling”, Fire Safety Science,Proceedings <strong>of</strong> the 5 th Int. Symp., Ed: Hasemi Y., 1997, pp 1285-1296.3. Ewer J., Galea E.R., Knight B., Patel M., Janes D., Petridis M., “Fire Field Modellingusing the SMARTFIRE Automated Dynamic Solution Control Environment”, CMS Press,Paper Number 98/IM/41, ISBN 1899991387, London, 1998.4. Taylor S., Galea E.R., Patel M., Petridis M., Knight B. and Ewer J., “SMARTFIRE: AnIntelligent Fire Field Model ”, Proc. Interflam 96, Cambridge, UK, March 1996, pp 671-680.5. Ewer J., Galea E.R., Taylor S., Patel M.K. and Knight B., “SMARTFIRE: An IntelligentCFD Based Fire Field Model”, To appear in Journal <strong>of</strong> Fire Protection Engineering, 1999.6. Ewer J., Knight B. and Cowell D., “Case Study: An Incremental Approach to Reengineeringa Legacy FORTRAN Computational Fluid Dynamics Code in C++ ”, Advancesin Engineering S<strong>of</strong>tware, vol. 22, 1995, pp 153-168.7. Spalding D.B., “A General Purpose computer Program For Multi-Dimensional One- andTwo- Phase Flow”, Mathematics and Computers in Simulations, North Holland (IMACS),Vol. XXIII, 1981, 267.8. FLOW3D Release 2.3.3 Reference Guide, CFD Dept AEA Harwell UK, Feb 1991.9. Kumar S., Gupta A.K. and Cox G., “Effects <strong>of</strong> Thermal Radiation on the Fluid Dynamics<strong>of</strong> Compartment Fires”, Fire Safety Science - Proc. <strong>of</strong> the Third Intl. Symp., 1991, pp 345-354.10. Lewis M.J., Moss M.B. and Rubini P.A., “CFD Modelling <strong>of</strong> Combustion and HeatTransfer in Compartment Fires”, Fire Safety Science, Proc. <strong>of</strong> the 5 th Int. Symp., Ed: HasemiY., 1997, pp 463-474.11. Galea E.R., Knight B., Patel M., Ewer J., Petridis M., and Taylor S., “SMARTFIREV2.01 build 365, User Guide and Technical Manual”, Smartfire CD and bound manual,1999.12. Steckler K.D., Quintiere J.G. and Rinkinen W.J., “Flow Induced By Fire in aCompartment”, NBSIR 82-2520, National Bureau <strong>of</strong> Standards, Washington, 1982.Appendix 11.3 Page 143-12 12