Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Predicting incident solar radiation on building’s envelope using machine learning

Alammar, Ammar, Jabi, Wassim ORCID: and Lannon, Simon ORCID: 2021. Predicting incident solar radiation on building’s envelope using machine learning. Presented at: 12th Symposium on Simulation for Architecture and Urban Design (SimAUD 2021), Virtual, 15-17 April 2021.

[thumbnail of SimAUD-2021-Final Paper-.pdf]
PDF - Published Version
Download (21MB) | Preview


The assessment of the impact of solar radiation on building envelopes has typically been achieved by using simulation software, which is time consuming and requires advanced computational knowledge. Given the increased complexity of large scale-projects and the demand for performative buildings, new innovative methods are required to assess the design efficiently. In this paper, we present an alternative and innovative approach to assessing solar radiation intensity on an office building envelope using two machine-learning (ML) models: Artificial Neural Network (ANN) and Decision Tree (DT). The experimental workflow of this paper consists of two stages. In the first stage, a generative parametric office tower and its urban context were designed and simulated using Grasshopper software to create a large synthetic dataset of the solar radiation that strikes the office room envelope with several types of analyses. In the second stage, the generated datasets were imported into two ML algorithms (ANN and DT) to create a model for training and testing. The comparison of these two ML models proved that input data types have a significant impact on the accuracy of the prediction and model selection. DT was found to be more accurate than ANN because the data is mostly categorical, which is the most suitable learning background for DT algorithms.

Item Type: Conference or Workshop Item (Paper)
Status: In Press
Schools: Architecture
Date of First Compliant Deposit: 8 April 2021
Date of Acceptance: 25 March 2021
Last Modified: 28 Nov 2022 12:22

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics