Skip to main content


Read more.


Thesis defences

PhD Oral Exam - Hongwen Dou, Building Engineering

Fault detection and diagnosis (FDD) of multiple-dependent faults of chillers

Date & time
Wednesday, March 29, 2023
10 a.m. – 12 p.m.

This event is free


School of Graduate Studies


Daniela Ferrer



When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.


As an indispensable system in modern buildings, the heating, ventilation, and air conditioning (HVAC) system usually takes a large proportion of building energy usage. However, faults of HVAC system in modern buildings are becoming a growing issue. These faults can aggravate equipment degradation or even damage equipment, and sequentially lead to the increase of maintenance cost and the significant energy waste. Therefore, it is paramount for the HVAC system to run as effectivity and fault free as possible; as a result, research and development into HVAC faults turns to an effective approach to improve building energy efficiency. This dissertation is a contribution to the research of fault detection and diagnosis with a new topic of chiller multiple dependent faults in a real building.

The forward fault detection model and the backward fault diagnosis model are developed to identify and isolate chiller system-level and component-level faults respectively, based on the measurement data from an institutional building. The second law of thermodynamics is applied to analyse the energy flow over an electric chiller, for the purpose of selecting target variables. Benchmarking grey-box models are developed to predict target variables, and sequentially to estimate the impacts of regressor variable faults on target variables. A fault symptom is detected when the residual of a target variable, the difference between the measured value and the predicted value, exceeds the corresponding threshold. Then, a backward rule-based approach is presented to identify if (i) the fault symptom is correct (i.e., a variable has abnormal values), or (ii) the fault symptom is incorrect (i.e., the symptom of target variable is caused by impacts generated by other faulty variables due to the dependency between variables), or (iii) both target and regressor variables are abnormal. The proposed model for chiller multiple-dependent faults is validated by a case study with a cooling plant serving an institutional building, where some faults are artificially inserted into the measurement data file. This dissertation also explored the effectiveness of transfer learning method in the application of improving deep learning model performance.

Back to top

© Concordia University