MODELING AND UNCERTAINTY - COMPARISON BET WEEN 2 APPROACHES OF ESTIMATION OF THE RELIABILITY OF NUMERICAL MODELING RESULTS

The problem developed and analyzed in this paper is that of the estima tion of the uncertainty associated with the results obtained by numeri cal simulation codes of physical systems induced from input data. In t he first section, a general problem is developed and the authors descr ibe briefly the methods used to calculate the uncertainty domain. They present two classic methods, one is a probability method, the Monte-C arlo method; the other, a determinist method, a differential analysis with finite differences. In the second section, two examples of models of thermal behaviour of building are used to underline the advantages and the drawbacks of both methods. These models are extracted fractio ns of larger models allowing a simplified presentation of the methods proposed. The authors determine the uncertainty domains on outputs sim ultaneously with the two methods. They show that the obtained domain b y differential analysis is lightly pessimistic, but very near to those of the Monte-Carlo method, in the case of first model, linear, like i n the case of second, highly nonlinear. The differential analysis is c learly more economical in calculation time and makes it possible to id entify sensitive data with significant bearing on output uncertainty. Nevertheless, it is emphasized that for this method it is essential to enter into the calculation code formalism. Globally, the authors conc lude that a relative superiority of the differential analysis exists, particularly in the case of large codes where Monte-Carlo use would be prohibitive in calculation time.