Efficient Probabilistic Assessment of Hygrothermal Performance: Sequential Monte Carlo and Decomposition Methods

Name: Tianfeng Hou

When: 18/09/2020 17:00
Language: en

Where: Justus Lipsiuszaal, 08.16, Blijde-Inkomststraat 21, 3000 Leuven

• Prof. dr. ir. Hans Janssen (promotor)
• Prof. dr. ir. Gustaaf Roels (co-promotor)

In building physics, deterministic hygrothermal models are widely used to simulate the building performance. Since significant uncertainties commonly exist for many of the involved variables, a probabilistic approach providing statistics of the probability distribution outputs is of crucial importance. However, when a highly accurate result is required, such probabilistic methodology (e.g. Monte Carlo analysis) 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 hygrothermal performance probabilistic assessment as much as possible, the targets of this research 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.

The thesis first focusses on the use of quasi-Monte Carlo based probabilistic assessment for hygrothermal performance, since it has the potential to outperform the standard Monte Carlo method. This part starts with a literature survey of different sampling strategies for probabilistic assessment of complex systems. In addition the quasi-Monte Carlo sampling strategies and related error estimation techniques are introduced in detail. Two applications — a numerical and a building physical application are respectively performed, with focus on both sampling efficiency and error estimation for the different sampling methods. In particular, the purely numerical examples are meant to illustrate the potential problems of applying quasi-Monte Carlo to the building physical application. And for the building physical application, both a thermal case study and a hygrothermal case study are performed. Finally, for getting a better understanding of the potential factors that may influence the performance of quasi-Monte Carlo methods, the effect of the different parameters and the smoothness of the target function are respectively investigated.

The second part of the thesis targets the investigation of using model order reduction methods for efficient hygrothermal simulation, given that it generally allows a large reduction of the simulation time without losing the dynamic behavior of the conventional models. First, a literature survey and the fundamental concepts of two model order reduction methods – proper orthogonal decomposition (POD) and proper generalized decomposition (PGD) are provided, focusing on their potential use for simulating building performance. Subsequently, the capability and robustness of POD and PGD are compared with the conventional finite element method based on both deterministic and probabilistic assessment of building thermal performance. Next, the POD method and the discrete empirical interpolation method (DEIM) are combined to construct a reduced order model to further lessen the computational expense of hygrothermal simulation. Two applications: a simple case study on non-linear heat conduction and a more complex case study on non-linear moisture redistribution are performed to assess the performance of POD-DEIM method. Further, the potential use of POD for highly non-linear hygrothermal simulations are investigated via two case study: HAMSTAD benchmarks 3 and 4. In particularly, HAMSTAD benchmark 3 centers on air transfer through a light weight wall and HAMSTAD benchmark 4 represents a very strongly non-linear and complicated case study, with focus on the analysis of a wall with a hygroscopic finishing material at the inside. Finally, our global conclusion on the main findings of the proposed methodology and future work about the follow-up research are put forward.

https://lirias.kuleuven.be/handle/123456789/659552

• Prof. dr. ir. Hans Janssen (promotor)
• Prof. dr. ir. Gustaaf Roels (co-promotor)
• Prof. dr. ir. Dirk Vandermeulen (chairman)
• Prof. dr. ir. Dirk Nuyens (secretary)
• Prof. dr. ir. Karl Meerbergen
• Prof. dr. Joris Degroote , Universiteit Gent
• Dr.ing. Steffen Freitag , Ruhr Universität Bochum