Mostafa commenced his tertiary study in Civil engineering at the University of Golestan, 2011. He continued his academic journey as a master student in water and wastewater engineering at the University of Tehran. As the top student, he graduated in 2018 with three published Q1 papers out of his thesis. Less than a month after his graduation, he was awarded two scholarships from Monash university for carrying out his Ph.D. research. His great passion for data science and environmental studies drove him to direct his study toward the current research under the supervision of Dr. Yihai Fang and Dr. Brandon Winfrey since January 2020.
Wastewater Treatment Plants (WWTPs) are a type of critical civil infrastructure that play an integral role in maintaining the standard of living and protecting the environment. Sustainable operation of WWTPs requires maintaining the optimal performance of their critical assets (e.g., pumps) at minimum cost. While various maintenance strategies, including Corrective and Preventive Maintenance, have been employed in the operation of WWTPs, Predictive Maintenance (PdM) has started to gain significant attention recently due to the advances in the Internet of Things (IoT) and Machine Learning. Based on asset condition monitoring data, PdM techniques are expected to predict the asset’s future performance and failures, which lead to more sustainable maintenance programs for individual assets. However, the knowledge is not well established on how PdM can be implemented to optimize maintenance decisions at the system level.
This research aims to develop a Digital Twin (DT)-driven framework to enhance current PdM practices in WWTPs. As a digital replica of physical systems, a DT promises to model complex system configurations and behaviors to facilitate simulation and prediction of system performance. The proposed DT-driven framework comprises three layers: the data integration layer, the analytics, and prediction layer, and the application layer. The data integration layer defines a standard data schema to harmonize the description and structure of WWTP asset data in digital platforms, including Building Information Modelling (BIM) and Computerized Maintenance Management System. Multiple Deep Learning methods will be assessed and employed at the analytics and prediction layer to detect physical condition anomalies, diagnose asset faults and, based on the system context of individual assets, predict the future performance of the system. Finally, the application layer presents a BIM-based interface for plant managers and operators to visualize scenario-based predictions and optimize maintenance decisions. This framework will be implemented and validated with monitoring and operational data from a real WWTP. The efficacy of the framework will be evaluated through qualitative and quantitative analyses. Findings from this research are expected to advance the knowledge in data-driven analytics and prediction in PdM and to extend the application of DT in critical civil infrastructures.