Comparison and combination of factorial and Monte-Carlo design in sensitivity analysis

The sensitivity of building physics simulation programs to the uncertainty of input parameters is analysed. A comparison is drawn between the classical Monte-Carlo method and a new method based on fractional factorial design. Their numerical behaviour is analysed and their respective advantages are discussed. The new method provides more information for an equivalent number of simulation runs. It allows a real feedback process since the relation between the uncertainty of the input and output data is fitted on a linear model. It also allows a better understanding of the behaviour of the studied building. Both methods are illustrated with a three-storey building, in which ventilation is analysed for different ratios between the wind and buoyancy forces. However, the two methods can also be combined and a map of the uncertainty of the output for the whole domain, determined by the variation of a few input parameters, can be drawn.