Back to search

University of British Columbia

Automatic digital twin development : sparse nonlinear identification approaches

Abstract

dc:description

The digital twin, serving as a virtual replica of its physical counterpart, plays a critical role in advancing industrial simulation, monitoring, and control applications. To develop an effective industrial digital twin, three essential aspects need to be addressed, including modeling, synchronization, and interaction, with modeling being the foundation. This thesis focuses on the sparse nonlinear identification of digital twins, facilitating online model updating and enhancing advanced process control to address the three fundamental aspects of the digital twin technique. Currently, the development of a comprehensive and accurate digital twin model requires substantial time and effort. To address this challenge, the first section of this thesis focuses on digital twin modeling. In this section, the sparse identification of nonlinear dynamics (SINDy) approach is adopted to streamline the digital twin identification procedure. Within the digital twin modeling section, two special scenarios are considered, including input measurement noise and complex system nonlinearities. Correspondingly, the noise-robust, modified generalized SINDy (MGSINDy) algorithm and the integrated algorithm that combines the single-layer, feed-forward neural network with SINDy are introduced. Once the digital twin model is implemented, its accuracy may degrade over time. To resolve this issue, the second section focuses on enabling online digital twin model updating. In this section, a Kalman filter and GSINDy integrated algorithm is proposed to recursively select digital twin model features and identify the corresponding parameters, enhancing precise synchronization between the physical and virtual domains. Restricted interactive capacities pose another significant challenge in the digital twin development. To address this, the third section of the thesis focuses on facilitating effective interactions between user instructions and physical operations utilizing the model predictive control (MPC). Consequently, an extended Kalman filter-based recursive sparse nonlinear identification with MPC approach is proposed, promoting interactions between user instructions and physical operations. Finally, this thesis presents the practical steps involved in developing and implementing the advanced digital twin development algorithms within a real industrial digital platform. Throughout the thesis, proposed algorithms are applied to both industrial and simulated case studies to demonstrate their effectiveness in improving the accuracy of digital twin models, enhancing synchronization precision, and strengthening interactive capabilities.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Chemical and Biological Engineering
Grantor dc:publisher
University of British Columbia
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Wang, Jingyi
Contributors dc:contributor
  • Gopaluni, Bhushan
  • Cao, Yankai

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/90987

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Wang, Jingyi. Automatic digital twin development : sparse nonlinear identification approaches. doctoral thesis, University of British Columbia, 2025. http://hdl.handle.net/2429/90987