Use of factorial sensitivity analysis in multizone airflow model tuning

Multizone models provide a relatively simple, rapid means for simulating airflow and contaminant dispersion in complex buildings. However, an untuned model may not agree well with the actual performance of the building it represents. An efficient method for tuning a model, which combines a heuristic approach with factorial sensitivity analysis, is evaluated for the case of a hypothetical single-story building served by three independent air distribution systems. A detailed "virtual" building model is constructed using multizone modeling software and a set of design data and then detuned by changing airflows, flow resistances, and other parameters. A multizone model of the virtual building is constructed using the same design parameters and is tuned by successively correcting its parameters to match values from the virtual building as directed by the sensitivity analysis, which uses the percentage of incorrect interzonal flow directions as the response variable. The method results in improvementfrom 60% incorrect to 10% incorrect after six correction steps. Improvement of the model is confirmed by application of ASTM Standard D 5157-97 for four different simulated tracer release scenarios. Improvement is significantly more rapid using the structured approach than when corrections are made using randomly selected measurements.