Below are the brief introductions and profiles of lectures to be presented at the IAHR Training Course on Artificial Intelligence Tools for Transient Flow Analyses organised by the IAHR Working Group on Transient Flows.
IAHR Training Course on Artificial Intelligence Tools for Transient Flow Analyses
Ali Haghighi: Artificial Neural Networks for Leak Detection
About Ali Haghighi: Ali Haghighi is a Professor of Civil Engineering and Hydroinformatics at the University of Kaiserslautern Landau (RPTU), Germany, having previously served as a Professor in the Department of Civil and Environmental Engineering at Shahid Chamran University of Ahvaz, Iran. His research centers on the convergence of classical hydraulic modeling and data-driven methods, with a focus on leak detection, system identification, and physical-parameter estimation in pressurized pipe networks. Over the past two decades, he has published extensively on Inverse Transient Analysis (ITA), surrogate modeling, and machine learning architectures aimed at solving ill-posed inverse problems in water systems. He has contributed to developing hybrid computational frameworks that combine transient wave dynamics with artificial neural networks (ANNs) and optimization algorithms to improve computational efficiency under real-world noise and sparse data conditions. He regularly serves as a peer reviewer for leading international journals in hydraulic engineering and water resources management.
About the presentation: Anomaly (Leak) Detection in Water Distribution Networks via Data-Driven Inverse Transient Analysis: A Neural Network Perspective. Inverse Transient Analysis (ITA) provides a physically grounded mechanism for detecting pipe anomalies such as leaks, wall deterioration, and unexpected valve configurations by analyzing high-frequency pressure wave propagation. However, classical optimization-based ITA often suffers from severe computational burdens and convergence issues due to high dimensionality, non-linear friction, and viscoelastic damping effects. This two-part session addresses these challenges through the lens of machine learning and neural networks. The 40-minute lecture covers the theoretical foundation of transforming traditional ITA into data-driven models. It presents neural network paradigms capable of handling both classification (leak localization) and regression (severity/size estimation) tasks, highlighting strategies to maintain physical consistency under noisy and sparse sensor configurations. The subsequent 40-minute hands-on tutorial demonstrates a practical execution workflow. Participants will process simulated pressure wave signals, construct a surrogate neural network model, and map transient responses back to system anomalies. The session concludes with a performance comparison highlighting the inference speed gains of neural surrogates over classical optimization techniques for network-scale applications.
Amy Wood: AI in Safety-Critical Hydraulic Simulation: An Industry Perspective
About Amy Wood: Amy Wood is a Senior Product Manager for the Flow Simulation team in the Datacor Engineering Software Group. She’s focused on building the roadmap, modernizing UX processes, and expanding the team’s ability to deliver value to users. She’s passionate about scaling systems that empower technical users and believes that great product strategy starts with listening closely to the people solving real-world problems. Amy holds a bachelor's degree in chemical engineering and a PhD in mechanical engineering, where her research emphasized product development for complex systems.
About the presentation: Transient analyses inform decisions with physical consequences: pipe ratings, surge protection, and safety system design. A tool that is usually right is not good enough when an engineer has to stand behind the result. That standard shapes how we decide where AI belongs, and where it does not.
The Datacor Flow Simulation team behind Impulse will share our perspective on three questions:
The trust problem - what AI assistance has to prove before it can touch a safety-critical engineering deliverable
What engineers actually ask for - findings from our customer research, including the consistent signal that users want time back in their workflow more than they want AI-solved physics
Where Datacor is heading - current and planned investments, and the architectural work behind them
André Artelt: Explainable Machine Learning for Water Distribution
About André Artelt: André Artelt is a post-doctoral researcher at Bielefeld University and a visiting researcher at the University of Cyprus. His research interests are centered around Trustworthy AI in Critical Domains, with a special focus on eXplainable AI (XAI) (in particular on counterfactual explanations) and water distribution systems. Besides basic research, he also works on applications of (X)AI in domains such as the aforementioned water distribution systems, transportation, and decision support systems for business owners.
About the presentation: Machine Learning (ML) and AI in general have shown great potential for several applications in water distribution systems (WDS). In my talk, I will give a brief overview of different applications and promising ML/AI techniques in this context. I will then elaborate on the need for trustworthy AI and discuss technical requirements and potential approaches for creating trustworthy AI in the context of WDSs. In particular, I will focus on fairness requirements and the use of explainable AI for making decisions transparent to the human operator. I will illustrate those algorithmic approaches in event diagnosis and control settings, highlighting their potential and discussing limitations of current methods. Finally, during a small hands-on session at the very
end of my talk, I will demonstrate some of those presented concepts in Python, together with some Python tools and packages for supporting research on AI methods for WDSs.
Dragan Savić: Digitalizacion and Agentic Artificial Intelligence Applications
About Dragan Savić: Professor Dragan Savić is a Global Advisor on Digital Sciences and a former CEO of KWR Water Research Institute in the Netherlands, as well as a Professor of Hydroinformatics at the University of Exeter, UK. With more than 40 years of experience in engineering, academia, and consultancy, he is internationally recognized for his work on hydroinformatics, artificial intelligence, digital transformation, and smart water systems. He founded and formerly directed the University of Exeter’s Centre for Water Systems. Professor Savić is a Fellow of the Royal Academy of Engineering and a Distinguished Fellow of the International Water Association.
About the presentation: Artificial Intelligence (AI) has moved from theory to transformative practice in the water sector. This talk traces the history of AI in water, from early expert systems and decision-support tools to today’s machine learning and Large Language Models driving predictive maintenance, demand forecasting, and real-time water quality monitoring. We will explore practical applications across utilities, infrastructure, and research, highlighting case studies where AI has improved efficiency, resilience, and sustainability. By connecting past developments with current innovations, the session will provide scientists with a critical perspective on how AI can shape the future of water management.
Jize Zhang: Uncertainty of Machine Learning in Civil Infrastructure
About Jize Zhang: Jize Zhang is an Assistant Professor in the Department of Civil and Environmental Engineering (primary) and the Division of Emerging Interdisciplinary Areas (joint) at The Hong Kong University of Science and Technology. He received his Ph.D. from the University of Notre Dame, an M.S. from Carnegie Mellon University, and a B.S. from Xi'an Jiaotong University. Before joining HKUST, he was a Machine Learning Research Scientist in the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory. His research broadly lies at the intersection of uncertainty quantification and artificial intelligence/machine learning for infrastructure systems. Applications include risk assessment and sensing for civil structures, engineering design optimization, and natural hazard risk assessment. He has co-authored over 60 peer-reviewed journal papers and premier AI conference papers at venues including NeurIPS, ICML, and CVPR. He has secured external research funding from multiple agencies including the Research Grants Council of Hong Kong, the National Natural Science Foundation of China, and the Innovation and Technology Commission of Hong Kong. He is an active member of ASCE, SIAM, and IEEE, serving on multiple technical committees.
About the presentation: Machine learning models are increasingly deployed in settings where a prediction alone is insufficient: scientific discovery, engineering design, and safety-critical decision-making all demand an answer to the question: how much should I trust this output? This tutorial provides a practical and conceptual introduction to uncertainty quantification (UQ) for modern machine learning models. We begin by distinguishing aleatoric from epistemic uncertainty and clarifying what each implies for data collection, model refinement, and downstream decisions. We then survey the dominant methodological families, including Bayesian neural networks and approximate inference, deep ensembles, Monte Carlo dropout, Gaussian processes, and evidential and quantile-based approaches, emphasizing the assumptions, computational costs, and failure modes of each. The second half turns to evaluation and reliability: calibration metrics, proper scoring rules, and distribution-free guarantees via conformal prediction, along with the persistent challenge of uncertainty estimation under distribution shift and out-of-distribution inputs. Attendees will leave able to select, implement, and assess UQ methods appropriate to their own problems.
Moez Louati: State of the Art on Artificial Intelligence and Transient Flows
About Moez Louati: Dr. Moez Louati is an Adjunct Associate Professor in the Department of Civil and Environmental Engineering at HKUST and a lecturer-Researcher at IHE-Delft. His B.Eng and MPhil degrees in Mechanical Engineering were obtained from the National School of Engineering of Sfax, Tunisia (ENIS). He received his PhD from HKUST in 2016 and received the SENG Overseas Research Award. Dr. Louati has more than ten years’ experience in hydraulic transients and smart urban water systems. Dr. Louati has ample experience in analytical analysis, modeling, computational fluid dynamics (CFD), experimental work, and machine learning. He also contributes to consultancy and advisory services for water utilities. He has conducted numerous field tests in Hong Kong in collaboration with water supplies and drainage services departments (WSD & DSD) on several DMAs and rising mains systems, and he has designed three water lab facilities. In 2023, he received HKUST’s tech-ship startup fund and co-founded Hyele Limited, a startup company developing novel technology for pipeline condition assessment. The startup is successfully incubated at the Hong Kong Science and Technology Park.
Muhammad Waqar: Graph Neural Networks
About Muhammad Waqar: Dr. Muhammad Waqar is a Lecturer in the Department of Construction and Quality Management at the Hong Kong Metropolitan University, Hong Kong. He completed his PhD in Civil Engineering from the Hong Kong University of Science and Technology (HKUST) in 2022 and received the RedBird Academic Excellence Award. His primary research focuses on asset management and condition assessment of pipelines using acoustics and ultrasonic non-destructive testing (NDT). Bridging the gap between theoretical modeling and field deployment, he pairs extensive expertise in fluid systems modeling with 7+ years of practical experience in real-field testing and implementations.
About the presentation: Transient pressure waves are ubiquitous in water distribution networks, and their repeated occurrence can contribute significantly to cumulative pipe fatigue and eventual failure. Localizing the sources of these transients is therefore important for identifying problematic assets or operations and enabling targeted investigation and mitigation. This lecture and accompanying tutorial will introduce the application of Graph Neural Networks (GNNs) to passive transient source localization in water distribution networks. The lecture will explain how the network topology and transient measurements can be combined to identify the pipe containing a transient source and estimate its location. Participants will learn how to represent a water distribution network as a graph, prepare the required input data, generate physics-informed training samples, apply a trained model, and interpret the localization results. The session will also discuss practical implementation considerations, limitations, and opportunities for applying the framework to real monitored water networks.
Sam van der Zwan: Artificial Intelligence Tools in Wanda Software
About Sam van der Zwan: Sam van der Zwan is a Senior Advisor/Researcher at Deltares, specialising in hydraulic transient flow modelling. He is the Product Owner of WANDA, Deltares' hydraulic transient simulation software, where he leads product development, API design, and innovation initiatives. Sam combines expertise in engineering, software development, and applied research to develop advanced modelling solutions for pipeline systems. He is also an active contributor to international knowledge exchange, serving as a member of the Technical Advisory Committee of the International Conference on Pressure Surges. He is also, together with colleagues, exploring the integration of AI technologies into engineering workflows.
About the presentation: Artificial Intelligence has evolved at a remarkable pace over the past year, especially in the field of agentic AI. While AI is often presented as the solution to every problem, the reality is more nuanced. What can AI genuinely contribute to hydraulic transient and water hammer studies, and where do its limitations lie? In this workshop, we will explore the practical application of AI in surge analysis and hydraulic transient modelling. We will discuss not only the opportunities offered by AI, but also important considerations such as model reliability, engineering responsibility, data security, and the risks of reliance on automated systems instead of knowledge transfer to new generations. We will demonstrate how AI can be used as a practical assistant for engineers working with water hammer software. Through a live demonstration of the first versions of the WANDA AI Agent, we will showcase how AI can help identify modelling mistakes, guide users during model development, answer domain-specific questions, and support engineering workflows. We will also present early examples of AI-assisted design optimisation, including simple surge vessel optimisation studies. The goal of this session is to provide a realistic view of how AI can enhance hydraulic transient analysis today, while highlighting the challenges that must be addressed before AI can become a trusted engineering partner.
Scott Lang: AI in Safety-Critical Hydraulic Simulation: An Industry Perspective
About Scott Lang: Scott Lang is the Development Manager for the Flow Simulation team in the Datacor Engineering Software Group, where he directs the implementation of algorithms and architecture for the simulation of thermal fluid systems. Previously lead developer and solution engine architect for xStream, Datacor's tool for analyzing transient flow in real-gas piping networks, he supports the fluids industry by publishing technical papers at a variety of conferences and serving on committees for the Hydraulic Institute and ASME PVP – including as Vice-Chair of HI's Viscosity Corrections committee and as Technical Program Representative for the 2027 PVP conference Fluid-Structure Interaction committee. Scott is a licensed Professional Engineer and holds a Bachelor of Science in Engineering with Mechanical and Electrical specialties from the Colorado School of Mines..
About the presentation: Transient analyses inform decisions with physical consequences: pipe ratings, surge protection, and safety system design. A tool that is usually right is not good enough when an engineer has to stand behind the result. That standard shapes how we decide where AI belongs, and where it does not.
The Datacor Flow Simulation team behind Impulse will share our perspective on three questions:
The trust problem - what AI assistance has to prove before it can touch a safety-critical engineering deliverable
What engineers actually ask for - findings from our customer research, including the consistent signal that users want time back in their workflow more than they want AI-solved physics
Where Datacor is heading - current and planned investments, and the architectural work behind them
Slavo Velickov: Artificial Intelligence Tools in Bentley Software
About Slavo Velickov : Dr. Slavco Velickov is the Global Advancement Director for Water Infrastructure at Bentley Systems, where he leads the advancement of digital twin solutions for the global water sector. A chartered engineer with more than 25 years of experience, he has contributed to water, wastewater, and stormwater infrastructure projects worldwide. His expertise includes hydraulic modelling, hydroinformatics, digital twins, artificial intelligence, water-loss management, and infrastructure resilience. Dr. Velickov holds a PhD from Delft University of Technology, the Netherlands, and postgraduate specializations in hydroinformatics and business development.
Trey Walters: AI in Safety-Critical Hydraulic Simulation: An Industry Perspective
About Trey Walters: Founder of Applied Flow Technology (a Datacor company), a developer of simulation software for fluid transfer systems. Developed commercial software in the areas of incompressible and compressible pipe flow, waterhammer, slurry systems, and pump system optimization. Teaches customer training seminars around the world. Has 40 years of experience in thermal/fluid system engineering. BSME (1985) and MSME (1986), both from the University of California, Santa Barbara. Fellow of the ASME.
About the presentation: Transient analyses inform decisions with physical consequences: pipe ratings, surge protection, and safety system design. A tool that is usually right is not good enough when an engineer has to stand behind the result. That standard shapes how we decide where AI belongs, and where it does not.
The Datacor Flow Simulation team behind Impulse will share our perspective on three questions:
The trust problem - what AI assistance has to prove before it can touch a safety-critical engineering deliverable
What engineers actually ask for - findings from our customer research, including the consistent signal that users want time back in their workflow more than they want AI-solved physics
Where Datacor is heading - current and planned investments, and the architectural work behind them