The role of falsification in the validation of numerical models

Abstract

ABSTRACT Numerical modelling consists of simulating with informatic technologies the conditions of experiments that would require big resources in terms of time and costs. For example, structural engineering considers numerical models of structures because otherwise it would be too expensive to build and destroy a building to study its behaviour under collapse conditions. In this regard, numerical modelling has become a valid alternative to physical experiments since they fall within the scientific method as they need validation that consists in discarding models when the results are not considered reliable. The criterion of falsification is therefore the basis of the selection process not of the ‘most truthful’ model, but of the ‘model least’, that is, less affected by errors. The paper discusses the sources of errors at the base of numerical models by proposing the importance of validation as the application of the falsification method proposed by Karl Popper.