University of Edinburgh Featured PhD Programmes
Peter MacCallum Cancer Centre Featured PhD Programmes
University of Glasgow Featured PhD Programmes

Reliability analysis, fault diagnosis and optimization in dynamic small data environment using explainable AI

Faculty of Engineering and Informatics

This project is no longer listed on and may not be available.

Click here to search for PhD studentship opportunities
Dr Qichun Zhang Applications accepted all year round Self-Funded PhD Students Only

About the Project

The smart system consists of the man-made machine and the intelligent decision making. Mostly, the machine will be designed with automation purpose and the decision making can be achieved based on big data training and AI technology. However, it is very difficult to define if the data is big enough or sufficient. The man-made machine somehow is unique due to the specified design requirement which means that the decision has to be made within the limited data amount. In addition, the smart system would interact with external environment in real-time which leads to the dynamics of the system. Therefore, it is significant to develop a novel framework to investigate the smart system with small data including the dynamics.

Nowadays, the smart system becomes more and more complex both in structure and functional logics. It has been shown that the reliability analysis and fault diagnosis are challenging. The dependent states and the events make the fault isolation more difficult and the cost increases dramatically dealing with the weak component in the system. The optimisation can be achieved effectively once the reliability and diagnosis can be analysed in real-time. Note that most of the current methods gave the probability for the reliability and diagnosis which means that the causing reasons hide behind.

Motivated by explainable AI (XAI), the fault diagnosis and reliability for dynamic smart system can be designed where the AI mechanism would explain the fault at multi-operational point. In particular, random forest and multi-layer CNN can be adopted for reliability analysis and fault diagnosis, and the optimisation can be implemented with probability density function shaping since the reliability can be described as a distribution model. In addition, due the limited data, the measurement distribution must be non-Gaussian which results in difficulty to optimise the performance as most of the existing optimisation results are based on Gaussian assumption.

The following 3-year schedule has been made for completing the project as a PhD research programme. In year one, the literature reviews for the existing explainable AI methods and reliability analysis. Developing basic fault diagnosis methods for dynamics systems. In the second year, the novel XAI framework considering the uncertainties and randomness will be developed in the second year in order to achieve the fault diagnosis and the stochastic distribution optimisation will be developed for the system optimisation in terms of reliability. In the third year, the presented AI hybrid model and algorithms will be validated by simulation and experiment at automotive pilot system at University of Bradford.


Mifeng Ren, Qichun Zhang & Jianhua Zhang (2019) An introductory survey of probability density function control, Systems Science & Control Engineering, 7:1, 158-170

Došilović, Filip Karlo, Mario Brčić, and Nikica Hlupić. "Explainable artificial intelligence: A survey." 2018 41st International convention on information and communication technology, electronics and microelectronics (MIPRO). IEEE, 2018.

Xing, Liudong, Gregory Levitin, and Chaonan Wang. Dynamic System Reliability: Modeling and Analysis of Dynamic and Dependent Behaviors. John Wiley & Sons, 2019.

FindAPhD. Copyright 2005-2021
All rights reserved.