SiF 2018– The 10 th International Conference on Structures in Fire FireSERT, Ulster University, Belfast, UK, June 6-8, 2018 DISCUSSION ON A SYSTEMATIC APPROACH TO VALIDATION OF SOFTWARE FOR STRUCTURES IN FIRE João Daniel Romeiro Ferreira 1 , Thomas Gernay 2 , Jean-Marc Franssen 3 ABSTRACT A validation exercise has been performed for the thermo-mechanical software SAFIR ® based on a dual approach. First, the examples proposed in the German National Annex of EN 1992-1-2 have been calculated with SAFIR and, second, the references to SAFIR found in scientific publications have been analysed with respect to the level of accuracy estimated by the authors for the results of SAFIR. The aim of this paper is not to present the detailed results of this validation exercise (which are published in open access reports). The focus is more on a critical analysis of the examples presented in the German National Annex which indeed paved the way to a systematic approach to validation of SiF software but could be improved, namely by including sensitivity analyses on the discretisation, a better description of some input data particularly in the material models, a presentation of the means used to obtain some reference results and a clear definition of the failure criteria to be used for determining fire resistance times. Furthermore, extensive analysis of the literature showed that the wide field of application of a typical SiF software requires an extension of the domain covered by the standard on concrete structures. Keywords: Structures, finite element modelling, benchmarking, validation, fire tests 1 INTRODUCTION Numerical models are used more and more in structural fire engineering. This trend is supported by extensive research efforts toward the development of computational models. However, it has to be recognised that, compared to CFD software, the validation of advanced tools for the thermo- mechanical analysis of structures in fire has received little attention, which may be partly due to a lack of clear guidelines and standards. Some efforts have been undertaken ([2], [12]), but systematic validations of software used in the field remain the exception rather than the norm. Validation of structural fire analysis codes can rely on benchmarking examples such as those provided in the German National Annex of EN 1992-1-2 [1], which has attracted the attention of the authors and will be commented in this paper. On the other hand, confidence in numerical models may rely on their validation against experimental tests, but to infer validity from such comparisons, a statistically relevant database of tests should be considered, which is rarely the case. This paper describes a systematic approach for software validation based on a combination of standard benchmarking examples and literature tests review. The work, applied to the particular software SAFIR ® [3] but valid in general, aims to: 1. Critically discuss the numerical modelling of the DIN benchmarking examples, while proposing relevant sensitivity analyses; 1 Research Engineer. Department of Architecture, Geology, Environment & Constructions (ArGEnCo), Liege University, Belgium. e-mail: jdferreira@uliege.be 2 Assistant Professor. Department of Civil Engineering, Johns Hopkins University, Baltimore, USA. e-mail: tgernay@jhu.edu 3 Professor. Department of Architecture, Geology, Environment & Constructions (ArGEnCo), Liege University, Belgium. e-mail: jm.franssen@uliege.be