Browse by author
Lookup NU author(s): Dr Xiang XieORCiD
This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
The emerging concept of digital twins outlines the pathway toward intelligent buildings. Although abundant building data carries an overwhelming amount of information, if not well exploited, the redundant and irrelevant data dimensions result in the overfitting problem and heavy computational load. Taking the fault detection and diagnosis process for building HVAC systems as the case, this study adopts a symbolic artificial intelligence technique to identify informative sensory dimensions for building-specific faults by exploring the symbolic representation of labelled time-series. To preserve this ad-hoc temporal knowledge in the digital twin ecosystem, machine-readable fault tags are defined to label corresponding sensor entities. A digital twin data platform is developed to annotate the real-time data with fault tags and produce filtered low-latency data streams associated with a specified tag to automate this process. This study informs a digital twin-based approach to automatically identify and pick up informative data to support dynamic asset management.
Author(s): Xie X, Merino J, Moretti N, Pauwels P, Chang J, Parlikad AK
Publication type: Article
Publication status: Published
Journal: Automation in Construction
Print publication date: 01/02/2023
Online publication date: 09/12/2022
Acceptance date: 28/11/2022
Date deposited: 29/11/2022
ISSN (electronic): 0926-5805
Altmetrics provided by Altmetric