A topology-derived flow inference approach to optimize sensor placement for effective inflow and infiltration detection in sewer networks

Abstract

Inflow and infiltration (I&I) in urban sewer networks increase both the delivery loads and overflow risks, thereby compromising environmental safety. While sensor-based detection of I&I is promising, a key limitation in most current applications is their reliance on fixed, isolated detection thresholds derived from individual sensors. This approach prevents the synthesis of information from multiple sensors and inherently limits the detectability of I&I events. To bridge this gap, this study introduces a novel topology-derived flow inference (TDFI) method that quantifies the pipe-specific minimum detectable I&I flow by synthesizing data from upstream sensors in sewer networks. Based on this method, a new detection threshold metric is formulated for sensor placement strategy (SPS) optimization, accompanied by an efficient Sequential Backward Selection (SBS) approach that deterministically constructs hierarchical sensor subsets through an iterative removal of the least-contributing sensors. Evaluation across three case studies (one synthetic and two real-life) demonstrates that the TDFI-based optimization framework identifies robust SPSs, leading to significantly improved overall detection performance for minor I&I events. A key finding is that prioritizing sensor density in high building-density areas significantly enhances overall detection performance. Comparative analysis shows that SBS achieved performance close to that of the DE-based optimizer while providing deterministic layouts under various budgets, offering a computationally efficient complementary approach. The core contribution of this study is to propose a topology-informed framework for SPS optimization in support of I&I detection, integrating upstream flow inference to maximize the detection capability of sensor layouts.