As climate change intensifies rainfall and flooding, the need for smarter, more resilient sewer system management is increasingly urgent. A research team led by Prof. Tarek ZAYED, Member of Research Institute for Artificial Intelligence of Things (RIAIoT), Research Institute for Sustainable Urban Development (RISUD) and Research Centre for Resources Engineering towards Carbon Neutrality (RCRE), Professor of Department of Building and Real Estate, has developed a multi-tiered model integrating artificial intelligence (AI) and the Internet of Things (IoT) to enable more cost-effective and intelligent sewer system management. The model supports a wide range of functions, from predicting exfiltration severity and identifying leakage-prone areas to monitoring and forecasting overflow incidents in high-risk locations.
This smart management model uses deep learning algorithms to analyse condition of sewer pipelines with a high degree of accuracy. By applying these algorithms, ageing and defective sections can be identified more effectively, helping to prioritise locations for CCTV inspections.
A core component of the system is the team’s pioneering Exfiltration Severity Index (ESI), which quantifies and models pipe-level exfiltration severity, enabling sewer line managers to identify segments that prone to leakage in advance. Prof. Zayed explained that sewer exfiltration occurs when sewage escapes from a defective sewer system into the surrounding environment, which can contaminate soil or groundwater with pollutants such as pathogens and toxic compounds, posing risks to both ecosystems and public health.
The AI model incorporates a range of parameters, including pipe characteristics, climatic conditions and environmental impacts, to predict the likelihood and severity of exfiltration. This provides valuable insights for prioritising the most urgent maintenance work. The team’s research showed that the system achieved an 85% accuracy rate in severity assessment, significantly reducing the risk of groundwater contamination. In addition, through optimised maintenance scheduling, the predictive model improved operational efficiency by 50% to 60% and reduced emergency repairs by 30% to 40%. The team’s findings have been published in the journal Tunnelling and Underground Space Technology, under the title “Proactive Exfiltration Severity Management in Sewer Networks: A Hyperparameter Optimization for Two-tiered Machine Learning Prediction”.
In addition to leakage, blockage is another major cause of sewer system failures and, in more severe cases, flooding. To address this issue, Prof. Zayed’s team has applied IoT-based technologies to simulate the performance of sewer networks and overflow events under different blockage scenarios.
In collaboration with the Drainage Services Department of the Hong Kong Special Administrative Region (HKSAR) Government, the team installed water-level sensors across drainage networks in Kowloon, collecting real-world data and applying a range of data mining techniques for case study simulations, as well as for model calibration and validation. The IoT-based monitoring system delivered impressive results: sewer segments identified as having blockage issues showed an 85% improvement in overall performance following targeted cleaning.
The related study, “Performance Assessment of Sewer Networks under Different Blockage Situations Using Internet-of-things-based Technologies”, has been published in the journal Sustainability. The research was supported by the Research Grants Council’s General Research Fund and the Environment and Conservation Fund.
Press release: https://polyu.me/4tHS0VL
Online coverage:
Mirage - https://polyu.me/41nOG64
Ta Kung Pao - https://polyu.me/3OrIxmm
Wen Wei Po - https://polyu.me/3OjGn8a; https://polyu.me/41WeB4V
Bauhinia - https://polyu.me/4vivqER
隨着氣候變化加劇，極端降雨及水浸風險愈趨頻繁，發展智能且具韌性的排污系統管理尤為迫切。由香港理工大學建築及房地產學系教授、人工智能物聯網研究院、可持續城市發展研究院及碳中和資源工程研究中心成員 Tarek ZAYED 教授領導的研究團隊，成功研發一套結合人工智能（AI）與物聯網（IoT）的多層級模型，為排污系統管理帶來更具成本效益、更智能化的解決方案。該模型具備多項功能，能預測滲漏嚴重程度、識別高滲漏風險區域，以及監測和預測高風險地點的溢流事故。
這套智能管理模型採用深度學習算法，能精準分析污水管道狀況，從而助更有效識別老化及損壞的管道，完善進行閉路電視檢測的優先排序。
系統的核心創新之一，是研究團隊首創的「泄漏嚴重程度指數」。該指數可量化管道層級的泄漏嚴重程度，並建立相關模型，讓排水系統管理人員及早識別容易出現泄漏的管段。Zayed 教授指出，排污系統泄漏是指污水從受損或有缺陷的排污系統溢出並滲漏至周邊環境，可能導致土壤或地下水受到病原體、有毒化合物等污染物影響，對生態系統及公眾健康構成威脅。
該人工智能模型綜合考慮多項參數，包括管道特性、氣候條件及環境影響，以預測泄漏發生的可能性及其嚴重程度，從而為維修工作的優次安排提供重要依據。研究結果顯示，該系統在泄漏嚴重程度評估方面的準確率達 85%，可顯著降低地下水污染風險。此外，該預測模型可用於優化保養日程，將營運效率提升 50% 至 60%，並減少 30% 至 40% 的緊急維修工作。相關研究成果已刊載於期刊《隧道與地下空間技術》(Tunnelling and Underground Space Technology)，論文題為「渠道網絡泄漏嚴重程度主動管理：基於超參數優化的兩級機器學習預測」。
除泄漏問題外，堵塞亦是導致排污系統失效的主要原因之一，嚴重時更可能引發水浸。為應對此問題，Zayed 教授團隊進一步應用物聯網技術，模擬排污網絡在不同堵塞情景下的運作表現及溢流情況。
團隊與香港特別行政區政府渠務署合作，在九龍區排水網絡安裝水位感測器，收集真實環境數據，並運用多種數據探勘技術進行案例模擬、模型校準與驗證。這套物聯網監測系統取得顯著成效：曾出現堵塞問題的排污管段經針對性清理後，其整體運作表現提升達 85%。
相關研究論文「利用物聯網技術評估不同堵塞情況下的渠道網絡效能」已發表於《可持續性》（Sustainability）期刊。此研究獲研究資助局的「優配研究金」及環境及自然保育基金的資助。
網上報導：
Mirage - https://polyu.me/41nOG64
大公報 - https://polyu.me/3OrIxmm
文匯報 - https://polyu.me/3OjGn8a；https://polyu.me/41WeB4V
紫荊雜誌社 - https://polyu.me/4vivqER
