A Computational Approach to Predicting Thermal Conductivity
Proposal by Noreen Saeed,
Monday, 13 May 2019 | 10:30am-12:30pm
Ramesh Krishnamurti (Chair), Professor, Computational Design, School of Architecture, CMU
Jonathan A. Malen, Professor Mechanical Engineering, CMU
Chris Harrison, Assistant Professor, Human Computer Interaction Institute, CMU
The work documented here pertains to the study of material properties, how they interrelate, and how these relationships can be exploited to assess the building fabric. This work lays out the foundations for a thesis proposal, stemming from my interest in sustainable design and from a need for tools that enable green building design and evaluation. The specific material property addressed in this project is thermal conductivity. A building’s thermal properties are key to predicting its performance, and energy consumption. For new buildings, thermal properties are used in determining its certification such as LEED etc. For older buildings, evaluation of thermal properties offers a useful insight into how much insulation to add, and to predictions of energy consumption. However, physically establishing the thermal conductivity of a building’s external walls is difficult and is often a time-consuming process. This proposal explores ways to use machine learning techniques for predicting the thermal properties of materials. It proposes a method to make this estimation by learning from quicker and easier to measure material properties such as dielectric properties and sound properties and establish their correlation with thermal conductivity.
The complete proposal document is available here: