PhD Thesis - staffweb - University of Greenwich

PhD Thesis by John Ewer.difficulties choosing a sufficiently restrictive, but also optimal, set of solution control parametersfor a previously unseen simulation scenario where those control parameters are used for thewhole of the simulation. Furthermore the results and timings from selected fire simulationscenarios demonstrate that run-time adjustment of control parameters can lead to savings of upto 50% for overall processing time when compared to some "safe" initial set of relaxationparameters used throughout the simulation. Clearly such savings are highly problem specific butthe principle of choosing a known "safe" set of control parameters and then adjusting thecontrols as required by the most up-to-date solution status is highly recommended and likely tobe of great benefit to both expert and intermediate users. Ultimately it is anticipated that theseexperiences of run-time solution control will be automated so that reliable CFD simulation andmonitoring is made available to all classes of user from novices to experts.6.2 Indicative test casesOne of the important factors, when considering CFD simulations, is the time required to arriveat the results. This is an easy quantity to measure and gives a reasonable indication of theeffectiveness of the interactive control. The problem with attempting to assess the benefits ofinteractive control, of the solution parameters, is that the effectiveness will be highly dependenton the quality of the initial solution parameters. The factors which determine a good choice ofthe initial control parameters are prior knowledge of the simulation of similar cases, a reasonableunderstanding of the behaviour of the particular CFD code in question and an adjustment (basedon engineering judgement) to account for the particular simulation being conducted. There are,however, no hard and fast rules to prescribe a suitable set of initial control parameters.This situation is further complicated by the highly complex nature of CFD simulations whichmeans that the simulation controls required to start a simulation are likely to be too extreme forthe later stages of the simulation. There are also potential transient characteristics of the flowsolution which typically require even tighter control regimes to prevent the simulation frombecoming unstable.6-91