New mathematical approaches to the quantification of uncertainty affecting the measurement of U-value

De Simon, Lia (2017) New mathematical approaches to the quantification of uncertainty affecting the measurement of U-value. PhD thesis, University of Nottingham.

PDF (Thesis - as examined) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (23MB) | Preview


This thesis describes the development and validation of a new computational procedure for the calculation of thermal transmittance (U-value) of existing building elements from the measurement of surface heat flux, and surface and nearby air temperatures.

The U-value plays a key role in the determination of the final energy consumption of a dwelling, and, as in the current political scenario reducing carbon emissions is a growing concern, obtaining accurate and quick measurements of thermal transmittance is of particular relevance to the precise representation of the energy performance of the building sector.

The calculation method developed is an extension of the RC network, a model based on the discretisation of building elements in resistors and capacitors in analogy with electrical circuits.

The advances proposed in this work extend the discrete RC networks to a model based on the full heat equation, with continuous, spatially varying thermal prop- erties. The solution algorithm is inserted in a Bayesian framework that allows the reformulation of the problem in terms of probability distributions. Two solution schemes have been confronted: Markov Chain Monte Carlo and Ensemble Kalman Filters approximation.

The model proposed has been validated on synthetic data, laboratory data collected in an environmental chamber on a solid and cavity wall, and in-situ data collected in 3 different locations (2 solid walls and 1 insulated steel frame construction).

The results show that the model offers an improved characterisation of the heat transfer through the building elements, furthermore, the algorithm can be used to analyse different wall constructions without the necessity of changing the structure of the model, as opposed to the standard RC networks, and, finally, it offers the practical advantages of the uncertainty reduction on thermal transmittance (from 14-25% to 7-10%) and a diminution of the necessary monitoring period from a minimum of 3 days to 1 day or less.

These advantages, in turn, benefit the building performance evaluation on different levels: in first instance, the practicality of measuring thermal transmittance in-situ is improved, thus making it easier to monitor the actual envelope performance and, secondly, the uncertainty reduction on the U-value leads to important reductions on the uncertainty surrounding the energy consumption predictions associated with a dwelling.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Wood, Christopher J.
Iglesias, Marco A.
Tetlow, David
Keywords: Bayesian inference, RC-Networks, U-value, energy efficiency
Subjects: T Technology > TJ Mechanical engineering and machinery
Faculties/Schools: UK Campuses > Faculty of Engineering > Built Environment
Item ID: 46738
Depositing User: De Simon, Lia
Date Deposited: 29 Nov 2017 04:40
Last Modified: 29 Nov 2017 07:15

Actions (Archive Staff Only)

Edit View Edit View