Zhidong Li
Menu
Water research

Zhidong Li

PhD

Associate Professor

Data Science Institute
University of Technology Sydney

Sydney, Australia

Advancing applied artificial intelligence for intelligent infrastructure, water systems, and real-world decisions.

Google ScholarORCIDResearch & collaboration

Research with
real-world purpose.

I work at the intersection of machine learning, infrastructure and the decisions that keep essential services running.

My research develops and applies data-driven methods for infrastructure reliability, water loss, wastewater assets and drinking-water quality. It brings statistical and machine-learning methods into conversation with engineering knowledge, operational data and practical constraints.

I am an Associate Professor at the UTS Data Science Institute. Before joining UTS, I was a Senior Research Engineer at CSIRO Data61. I received my PhD from the University of New South Wales in 2013.

The questions that interest me are translational: how to make a prediction useful, how to communicate its uncertainty, and how to connect research with a responsible decision in practice.

Research interests

Methods informed by the systems they serve.

Intelligent water infrastructure

Pipe failure prediction, predictive maintenance, asset condition assessment and renewal prioritisation.

Water quality & environmental intelligence

Water-quality forecasting, distribution analytics, catchment modelling and operational decision support.

Applied machine learning

Predictive and temporal modelling, graph learning, risk analysis and uncertainty estimation.

Responsible & translational AI

Interpretable analysis, model reliability and human-centred decision support in real-world applications.

Selected water research

Collaborative work across the water cycle.
Research questions, technical approaches and practical contributions.

6 projects

AI-driven water main failure prediction

Water networks

Understanding which assets need attention, before failure becomes disruption.

Research challenge
Age alone is a limited guide to pipe reliability. Utilities need to combine incomplete asset records with the environmental and operational conditions that shape failure risk.
Technical approach
My collaborative research applies Bayesian models, survival analysis, temporal modelling and graph learning to pipe-level risk assessment, drawing together asset characteristics and historical failure evidence.
Practical contribution
Risk estimates and network-level analysis can inform condition assessment, inspection and renewal priorities. The work connects predictive modelling with engineering decisions.

Collaborative research with CSIRO Data61 and water utilities, including Sydney Water.

Research paperPublic project context

Intelligent water leak detection

Sensing & water loss

Turning acoustic signals into evidence for a more focused field investigation.

Research challenge
Non-surfacing leaks are difficult to distinguish from the background noise of an operating water network. Monitoring also needs to be targeted to the places where it can be useful.
Technical approach
The research combines acoustic-signal analysis and machine-learning classification with pipe-risk information, bringing sensor evidence into the same decision process as asset data and field findings.
Practical contribution
The resulting analysis supports screening of likely leak signals and prioritisation of field checks, connecting sensing with maintenance decisions.

Collaborative water-sector research spanning utility operations, sensing and data analytics.

Research paper

Raw-water quality forecasting

Water quality

Giving treatment teams an earlier view of changing catchment conditions.

Research challenge
Source-water quality changes across time and place. Treatment planning benefits from understanding upstream observations alongside weather and catchment conditions.
Technical approach
Spatial-temporal forecasting brings historical water-quality observations together with environmental and weather information to model changing conditions across source-water catchments.
Practical contribution
Forecasts provide an additional evidence base for operational planning and treatment decisions. Interpretation remains grounded in the monitoring data and the conditions under which a model is used.

Collaborative work with water-sector and catchment partners.

Discuss this research
Sewer rising-main failure predictionWastewater assets

Sewer rising-main failure prediction

Research challenge
Infrequent failures and complex local conditions make rising-main inspection and renewal difficult to prioritise.
Technical approach
Machine-learning models bring pipe characteristics, soil, elevation, waterways, operational information and historical failures into a common risk assessment.
Practical contribution
The analysis supports inspection and renewal planning, with model interpretation and updating as part of the workflow.
Discuss this research
Gravity-sewer corrosion and digital twinsCorrosion & digital twins

Gravity-sewer corrosion and digital twins

Research challenge
Corrosion develops unevenly across sewer networks, while direct condition observations are often sparse.
Technical approach
The research connects sensing and condition evidence with deterioration models and estimates of remaining service life, within a digital-twin context.
Practical contribution
The aim is to support targeted assessment and maintenance by relating predictions to the physical condition of assets.
Discuss this research
Distribution water quality predictionDistribution water quality

Distribution water quality prediction

Research challenge
Upstream operations affect water quality throughout a distribution system, but direct observations are limited.
Technical approach
Data-driven models of travel time and chlorine residual link reservoir information, network topology and downstream observations, with explicit attention to uncertainty.
Practical contribution
Network-wide estimates and scenario analysis support disinfection, dosing and operational planning.
Discuss this researchResearch paper

From methods
to practice.

The value of applied AI lies in the decisions it helps people make.

My water research connects asset records, sensing and environmental observations with questions of inspection, maintenance, treatment and long-term planning. The emphasis is on useful models, explicit uncertainty and results that can be interpreted in their engineering context.

These projects are collaborative. Their development brings together researchers, utility specialists and technology partners; programme outcomes belong to those teams and the settings in which the work was evaluated.

Read CSIRO’s public infrastructure research case study

Selected publications

Research in water infrastructure and applied machine learning.

Google ScholarORCID
  1. 2022

    Machine Learning for Efficient Water Infrastructure Management

    Zhidong Li, Bin Liang, Yang Wang

    Humanity Driven AI · Springer · pp. 37–61

    doi:10.1007/978-3-030-72188-6_3
  2. 2021

    Failure Prediction for Large-scale Water Pipe Networks Using GNN and Temporal Failure Series

    Shuming Liang, Zhidong Li, Bin Liang, Yu Ding, Yang Wang, Fang Chen

    ACM CIKM 2021

    doi:10.1145/3459637.3481918
  3. 2020

    A Data Driven Approach for Leak Detection with Smart Sensors

    Bin Liang, Sunny Verma, Jie Xu, Shuming Liang, Zhidong Li, Yang Wang, Fang Chen

    IEEE ICARCV 2020

    doi:10.1109/ICARCV50220.2020.9305403
  4. 2020

    Data-driven Water Quality Prediction in Chloraminated Systems

    A. Peters, B. Liang, H. Tian, Z. Li, C. Doolan, D. Vitanage, H. Norris, K. Simpson, Y. Wang, F. Chen

    Water e-Journal · 5(4) · pp. 1–19

    doi:10.21139/wej.2020.022
  5. 2018

    Long-Term RNN: Predicting Hazard Function for Proactive Maintenance of Water Mains

    Bin Liang, Zhidong Li, Yang Wang, Fang Chen

    ACM CIKM 2018 · pp. 1687–1690

    doi:10.1145/3269206.3269321
  6. 2014

    Water Pipe Condition Assessment: A Hierarchical Beta Process Approach for Sparse Incident Data

    Zhidong Li, Bang Zhang, Yang Wang, Fang Chen, Ronnie Taib, Vicky Whiffin, Yi Wang

    Machine Learning · 95 · pp. 11–26

    doi:10.1007/s10994-013-5386-z

Academic &
research experience

  1. 2018–present

    University of Technology Sydney

    Associate Professor · Data Science Institute

    Joined UTS in December 2018 as Senior Lecturer; subsequently appointed Associate Professor.

  2. Previously

    CSIRO Data61

    Senior Research Engineer

    Applied machine learning, smart infrastructure and industry research collaboration.

  3. 2013

    University of New South Wales

    PhD · Sydney, Australia

Recognition

Selected awards recognising collaborative research.

  1. 2021

    UTS Medal for Research Impact

    Team recognition for research with impact beyond the academic community.

    UTS award record
  2. 2018

    Eureka Prize for Excellence in Data Science

    Awarded to the Smart Infrastructure Team at CSIRO’s Data61.

    Australian Museum winner archive
  3. 2018

    Australian Water Association recognition

    NSW Research Innovation Award and National Research Innovation Merit Award, recognising the Smart Pipes team.

    Public impact report

Let’s explore
a research question.

I welcome conversations about industry collaboration, joint research and the translation of AI into practical water and infrastructure decisions.

Dr Zhidong Li
Associate Professor
Data Science Institute, UTS

Sydney, Australia

Google ScholarORCID