Augmented machine learning for sewage quality assessment with limited data

作     者：Jia-Qiang Lv Wan-Xin Yin Jia-Min Xu Hao-Yi Cheng Zhi-Ling Li Ji-Xian Yang Ai-Jie Wang Hong-Cheng Wang 

作者机构：State Key Laboratory of Urban Water Resource and EnvironmentSchool of EnvironmentHarbin Institute of TechnologyHarbin150090China School of Civil and Environmental EngineeringHarbin Institute of Technology ShenzhenShenzhen518055China College of the EnvironmentLiaoning UniversityShenyang110036China 

出 版 物：《Environmental Science and Ecotechnology》 (环境科学与生态技术(英文))

年 卷 期：2025年第23卷第1期

页      面：193-203页

核心收录：

学科分类：083002[工学-环境工程] 0830[工学-环境科学与工程（可授工学、理学、农学学位）] 08[工学] 

基　　金：support by the National Natural Science Foundation of China(No.52321005,No.52293445) Shenzhen Science and Technology Program(KCXFZ20211020163404007,Grant No.KQTD20190929172630447) 

主　　题：Machine learning Hybrid model Mechanistic augmentation Sewer system Sulfide and methane 

摘      要：Physical,chemical,and biological processes within sewers significantly alter sewage composition during *** leads to the formation of sulfide and methanedcompounds that contribute to sewer corrosion and greenhouse gas *** modeling of these compounds is essential for effective sewer management,but the development of machine learning(ML)models is hindered by differences in data accessibility and sampling frequencies of water quality *** we present a mechanistically enhanced hybrid(ME-Hybrid)model that combines mechanistic modeling with data-driven *** model harmonizes datasets with varying sampling frequencies and generates synthetic samples for ML training,thereby enhancing the monitoring of methane and sulfide in *** optimal MEHybrid model integrates the backpropagation neural network with mechanistic frequency *** demonstrate that the ME-Hybrid model outperforms pure ML and linear interpolation in capturing fluctuating trends and extremes of sulfide concentrations,achieving a coefficient of determination(R2)of *** samples generated through mechanistic augmentation closely approximate real samples in modeling performance,statistical distribution,and data *** enables the model to maintain high predictive accuracy(R20.76)for sulfide even when trained on only 50%of the ***,the ME-Hybrid model successfully assesses sewer methane concentrations with an R2 of 0.94,validating its applicability and generalization *** results provide a reliable methodological framework for modeling and prediction under data *** facilitating better monitoring and management of sewer systems,the ME-Hybrid model aids in the development of strategies that minimize environmental impacts,enhance urban resilience,and ultimately lead to sustainable urban water systems.