Efficient probabilistic assessment of building performance: sequential Monte Carlo and decomposition methods
- November 13 (Fri) at 16:00 - 18:10, 2020 (JST)
- Dr. Tianfeng Hou (Postdoctoral Researcher, iTHEMS / Postdoctoral Researcher, Prediction Science Laboratory, RIKEN Cluster for Pioneering Research (CPR) / Postdoctoral Researcher, Data Assimilation Research Team, RIKEN Center for Computational Science (R-CCS))
- via Zoom
The use of numerical simulations for complex systems is common. However, significant uncertainties may exist for many of the involved variables, and in order to ensure the reliability of our simulation results and the safety of such complex systems, a stochastic approach providing statistics of the probability distribution of the results is of crucial importance. However, when a highly accurate result is required, the conventional Monte Carlo based probabilistic methodology inherently requires many repetitions of the deterministic analysis and in cases where that deterministic simulation is (relatively) time consuming, such probabilistic assessment can easily become computationally intractable. Hence, to reduce the computational expense of such probabilistic assessments as much as possible, the targets of this seminar are twofold: (1), to exploit an efficient sampling strategy to minimize the number of needed simulations of Monte Carlo based probabilistic analysis; (2), to investigate a surrogate model to reduce the computational expense of single deterministic simulation.
This seminar contains two parts and will be accompanied by a set of illustrative building physical case studies (analysis of the heat and moisture transfer through building components). The first part of this seminar focusses on the use of quasi-Monte Carlo based probabilistic assessment for building performance, since it has the potential to outperform the standard Monte Carlo method. More specifically, the quasi-Monte Carlo sampling strategies and related error estimation techniques will be introduced in detail. In addition, questions on under which conditions the quasi-Monte Carlo can outperform the standard Monte Carlo method will be answered by a set of analyses. The second part of this seminar targets the investigation of using model order reduction methods for optimizing the deterministic simulation, given that it generally allows a (large) reduction of the simulation time without losing the dynamic behavior of the conventional models (such as the transient finite element analysis). Particularly, the fundamental concepts of one common model order reduction method – proper orthogonal decomposition (POD) will be provided, and its potential use for simulating (building physical) problems with different levels of non-linearity and complexity will be illustrated.
*Please contact Keita Mikami's mail address to get access to the Zoom meeting room.