Exeter, UK & Online (12:30-13:30 CET)
Machine learning is transforming predictive asset maintenance by turning raw data into actionable insights with remarkable precision. This talk highlights the pivotal role of machine learning in modern maintenance strategies for water systems. By analysing past failures in water distribution networks (WDNs), machine learning models predict pipe failure probabilities based on static attributes—such as length, diameter, age, and material—and dynamic factors like pressure, flow, and strain. These models not only identify the most influential factors but also enable more precise and efficient predictive maintenance practices.
This hybrid-seminar will demonstrate how domain knowledge enhances machine learning algorithms, leading to the development of a semi-supervised approach to address data scarcity. It will also tackle the challenge of imbalanced datasets, a frequent issue in WDNs. Finally, innovative ensemble learning techniques for pipe failure prediction will be showcased, offering insights into cutting-edge advancements in this field.
For more information (and to register) visit events.teams.microsoft.com/event/from-data-to-insights.
