Overview

Wastewater sewer networks show strong daily, rainfall driven, and operational variability, which makes anomaly detection challenging when labelled failure data is unavailable. This webinar presents a physically informed anomaly detection framework that conditions wastewater flow on known physical drivers such as daily usage patterns, pump activity, and short lag rainfall before applying residual based detection.

By explaining predictable variation, the method improves robustness to regime shifts and increases the interpretability of detected anomalies. A lightweight diagnostic layer assigns anomalies to wet weather or dry weather contexts to support practical operational triage. Using one year of publicly available telemetry from United Kingdom sewage pumping stations, the physically informed model explains a substantial share of flow variability, with both daily timing effects and rainfall effects jointly significant. Residual based anomalies form a concise and operationally meaningful subset of events, with dry weather anomalies particularly indicative of potential asset or inflow issues. Quantitatively, the proposed framework reduces rainfall‑period false‑alarm alerts and increases the proportion of actionable dry‑weather anomalies compared with an unconditioned baseline detector, demonstrating measurable technical improvement. These results provide initial empirical evidence that physically informed conditioning on known drivers can improve anomaly relevance and operational usefulness in wastewater network monitoring.

Speaker

Rostyslav Pietukhov, Global Head of Insights & Analytics, Kubrick

Ross is a Senior analytics and AI leader, currently Global Head of Insights & Analytics with Kubrick Group, with 12+ years’ experience delivering enterprise-scale transformation across multiple industries. He leads Insights & Analytics at Kubrick Group, shaping AI and data strategy, helping organisations embed data-driven decision-making into core operating models through agentic AI adoption and workforce upskilling. PhD researcher and lecturer at the University of Warwick, focused on supply chain analytics and management. Trustee at Free UK Genealogy, contributing to technology, data, and AI strategy. he has previously held roles at Rolls-Royce, Unilever, and Capgemini, building deep expertise in analytics, process optimisation, and large-scale digital and operational transformation.

McKenzie Howells, Solutions Lead, Kubrick

McKenzie is a Solutions Lead and Data Engineer at Kubrick Group with a background in Python Programming Language and numerical modelling, working on projects spanning data-driven financial and supply chain MATLAB, human nutrition and health, as well as energy policy. His career goals are to combine his knowledge of mathematical and computational methods with his ability to solve complex issues and explain them through an impactful way. He thrives in dynamic and collaborative environments where people have the same vision and enthusiasm for their work and thrives in situations that require high-level communication skills. He has a BS in Mathematics from Cardiff University.

Amrinder Rai, Supply Chain Analytics Lead, Kubrick

Amrinder is a Supply Chain Analytics Lead at Kubrick Group. Amrinder has led and delivered technical solutions to complex problems across telecoms, utilities, retail and manufacturing. Before joining Kubrick, he led supplier risk research within an academic-industry collaboration. Building on a background in data science and engineering, he has progressed to map these technologies directly to end business outcomes through a specialisation in supply chain and risk management, along with a solutions and product mindset. ​

The material presented has not been peer-reviewed. Any opinions are the presenter’s own and do not necessarily represent those of IChemE or the Process Management and Control Special Interest Group. The information is given in good faith but without any liability on the part of IChemE.

Time

09:00–10:00 GMT.

Software

The presentation will be delivered via Microsoft Teams. We recommend downloading the app from the Microsoft website, rather than using the web portal.