Dynamic fault modelling and prediction in cyber-physical systems

Download Statistics

Downloads

Downloads per month over past year

Adeyemo, Hayatullahi Bolaji ORCID: https://orcid.org/0000-0001-5229-9591 (2024). Dynamic fault modelling and prediction in cyber-physical systems. University of Birmingham. Ph.D.

Adeyemo2024PhD.pdf
Text - Accepted Version
Restricted to Repository staff only
Available under License All rights reserved.
Download (4MB) | Request a copy

Abstract

Cyber-Physical Systems (CPS) are distinguished by their intricate interplay between physical components and computational processes. Each constituent entity plays an integral role in advancing the system towards its overarching objectives, resulting in a tightly coupled and complex relationship within the CPS framework. As CPS becomes increasingly pervasive in critical domains such as healthcare, transportation, robotics, smart grids, and manufacturing, ensuring their reliability and resilience against faults and failures is paramount. This thesis explores the challenges and methodologies associated with fault detection in CPS environments. There is a rise in the number of studies on CPS fault detection, with a notable emphasis on security aspects. There remains a relative dearth of research dedicated to non-intrusive faults, which may arise from natural system variability or unintended malfunction. The limited attention to nonintrusive faults underscores the need for a more comprehensive understanding and exploration of these types of faults to ensure a well-rounded and robust fault detection framework for CPS. One of the biggest challenges of fault detection is establishing what is considered the normal behaviour and the abnormal behaviour of the system with limited intervention of humans. The proposed method’s usefulness is proved through extensive simulation and experiments. The existing fault detection methods are studied, categorised, and analysed. We present the concept of surrogate modelling in CPS; wherein surrogate models are utilised to emulate the behaviour of complex physical systems. The integration of surrogate modelling techniques enhances fault detection capabilities by providing insights into system dynamics and facilitating rapid prototyping of detection algorithms. We leverage the capability of fault injection to generate different system dynamics to understand the behaviours of the system under different fault scenarios. Expert models are trained to help in fault detection with each model trained to detect and/or predict specific faults. The approach is further enhanced by adaptively training additional expert models in case the existing models are unable to detect the unknown faults. Through empirical evaluations and case studies, this thesis validates the effectiveness of the proposed fault detection methodologies against CPS application. The results demonstrate the feasibility and benefits of proactive fault management strategies in enhancing the resilience of CPS against various faults. This thesis contributes to the advancement of fault detection in CPS by providing proactive fault detection strategies, which can help CPS designers and engineers to fortify system resilience and mitigate the impact of unforeseen faults on critical infrastructure and operations.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):