2017/10/18 by Filippo Monari, Monari, Filippo
Computer Science · Decision Sciences · Engineering · #Advanced Multi-Objective Optimization Algorithms #Building Energy and Comfort Optimization #FOS: Computer and information sciences #Methodology (stat.ME) #Probabilistic and Robust Engineering Design #Structural Health Monitoring Techniques
paper · pdf · doi:10.48550/arxiv.1710.06671
openalex publication_date 2017/10/18 · openalex created_date 2022/09/13 · openalex updated_date 2026/07/28
The construction of computer models (mathematical models implemented in\ncomputer codes), with respect to observed phenomena, is usually undertaken by\nbuilding different variants depending on modeller sensibility, and choosing the\none yielding the best fit of the field data, according to Root Mean Squared\nError (RMSE) based measures. Usually a particular model is chosen because of\nits marginally lower RMSE, and not because of its actual higher adequacy,\nrisking that its capability of extrapolating predictions is poor. This work\naims at improving the current practice in the creation of computer models by\nproposing an approach similar to those employed in statistical modelling,\nwherein starting from the simplest hypothesis, effective model upgrades are\nidentified by analysing discrepancies between observations and predictions, and\ndifferent model variants are compared according to robust likelihood based\ncriteria. The method, focused on models with high dimensional outputs and\nboundary conditions and centred on Bayesian calibration, is demonstrated on\nnumerical experiments considering a series of building energy models. The\nobject of the modelling is a test facility used for round robin tests in the\ncontext of the International Energy Agency (IEA), Energy Building and\nCommunities (EBC), Annex 58.\n