Sewer Blockage Sensor Placement Optimization and Prioritization using AI, Advanced Data Analytics, and AGOL Dashboards

PURPOSE Urban Utilities, WCS Engineering and Optimatics pioneered an innovative sensor deployment optimization and prioritization decision-support platform to help prevent dry weather overflows resulting from pipe blockages. Harnessing Artificial Intelligence infrastructure, cloud computing and custom Python optimization scripts, a sensor deployment roadmap based on comprehensive data was generated from thousands of model simulations to optimize sensor placement, maximizing high-risk pipe coverage with minimum sensor units. ArcGIS Online dashboarding simplified complex data into an easy-to-use interface, which, paired with deployed sensors' early warning system, enables Urban Utilities' staff to make optimized sensor placement decisions and quickly respond to emergency situations. BENEFITS The presentation introduces a groundbreaking convergence of intelligent algorithms, custom Python scripting, hydraulic modeling, and dashboarding; offering a tangible and accessible solution to utilities as evidenced by its successful implementation in Urban Utilities' Breakfast Creek collection system catchment. This sensor deployment optimization and prioritization framework is impactful, customizable, adaptable, and scalable for any utility regardless of the size or complexity of their collection system. INTRODUCTION Urban Utilities engaged WCS Engineering (WCS) to apply Optimizer-ICMTM (Optimizer) and customized scripts to select sensor placement in the pilot catchment of Breakfast Creek (spanning 407 km of sewer main) that would maximize coverage of gravity sewers with higher risk of blockage and consequence of failure (CoF). The primary intent of the sensor placement is to provide an early warning system prior to dry weather overflows and to provide an emergency response framework that identifies potential overflow locations when a sensor high-level alarm is triggered. Secondary benefits of sensor placement include the potential for enhanced model calibration data and real-time system performance forecasting. DETAILS Leveraging Optimizer's capabilities of intelligent algorithm optimization technology and cloud computing, the hydraulic model of the Breakfast Creek catchment was simulated over 12,000 times. Each model run represented a scenario in which a single pipe blockage was simulated during peak dry weather flow (PDWF). Optimizer captured depth timeseries data for each blockage model run. The raw output from the Optimizer blockage analysis included extensive, large databases that required customized post-processing using Python scripts to consolidate and translate the data into more manageable database formats. Those databases then informed the custom sensor placement optimization and prioritization algorithm developed by WCS in Python. The algorithm first traces downstream of every potential sensor location, checking sensor alarm levels against the blockage surcharge response levels to determine the coverage reach of each sensor. The algorithm then optimizes and prioritizes the sensor deployment by computing a Sensor Effectiveness Score (SES), a custom quantity cumulatively calculated from parameters related to the pipes covered by each sensor. Through the tracing and prioritization of sensors, the algorithm engine collects and synthesizes key information and metrics pertaining to the coverage of each sensor. Among other quantities, the following critical parameters are cumulatively stored, for each sensor: pipes covered, lengths of pipes covered (total, unique, and by risk category), potential overflow locations covered, maximum preventable overflow rate, overflow location environment, and the time between sensor high-alarm detection to overflow. The results from the tracing, optimization, and prioritization algorithms were then condensed into an interactive, data-driven, decision-support platform using ArcGIS Online dashboards. Sensitivity analyses were undertaken to test the effect of key inputs and assumptions on the sensor placement prioritization results. These included running scenarios with varying inputs for sensor placement site access constraints, asset risk score assumptions, SES weightings, minimum response time and minimum freeboard from alarm activation to overflow. The preferred SES calculation adopted was risk multiplied by length. The risk score is a function of CoF, structural grade and historical pipe blockages. The CoF incorporates the location of overflow which was previously calculated using the Optimizer-ICMTM blockage model (Moore, et al., 2021). A dynamic ArcGIS Online dashboard was built to allow Urban Utilities to access the complex tracing and ranking results in an easy-to-use decision-support platform, allowing Urban Utilities' staff to focus their time on where it's most efficient — making informed decisions under critical, time-sensitive conditions. The dashboard gives a crystal-clear view of the sensor placement prioritization results (including all parameters related to each sensor location), potential network vulnerabilities, and how to best address them in a planning and emergency scenario. The dashboard interface clearly displays the prioritized order in which sensors may be deployed to maximize risk reduction. It also allows the user to select any alternative, non-optimal sensor location in the map to visualize the resulting coverage. If a proposed sensor location manhole is deemed inaccessible