Statistical Learning for Fluid Network Leak Characterization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and characterizing leaks in fluid networks, such as water distribution systems, are inefficient as they require numerous sensors and expert operators, are prone to environmental disturbances, and cannot accurately determine leak severity or location, leading to ineffective prioritization of maintenance.
Innovation Solution
A method involving a statistical learning model trained on a database of hydraulic behavior data, including leak and no-leak scenarios, using flow rate and pressure sensors to detect and characterize leaks by determining their area and flow rate, which allows for more precise and reliable leak detection and characterization with fewer sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vibro-acoustic listening methods are used to detect leaks, then leak location can be identified, but the method requires a large number of sensors and expert operators, and cannot characterize leak severity
Solution Approach 1:
The patent replaces the mechanical/acoustic listening system with a hydraulic measurement system using flow rate sensors and pressure sensors. Instead of using microphones to listen to leak sounds, the system measures hydraulic parameters (flow rates and pressures) at various points in the network and uses statistical learning to infer leak characteristics, thereby eliminating the need for numerous acoustic sensors and expert operators.
Solution Approach 2:
The patent introduces statistical learning models as an intermediary between the hydraulic sensor measurements and the leak characterization results. The model processes the hydraulic behavior data to infer leak location, area, and flow rate, acting as a mediator that transforms raw sensor data into meaningful leak characteristics without requiring direct physical measurement of the leak itself.
2Reliability
If sectorization methods are used to detect leaks, then leak presence can be detected by comparing inlet and outlet flow rates, but precise leak location and severity characterization are not achieved
Solution Approach 1:
The patent applies segmentation by dividing the fluid network into multiple areas and placing flow rate sensors and pressure sensors at strategic points within each area. This allows the system to not only detect leak presence through flow rate comparison but also to pinpoint the specific area containing the leak and characterize its severity, overcoming the limitations of traditional sectorization methods.
Solution Approach 2:
The patent implements feedback by using pressure sensors to measure hydraulic conditions and feeding this information back into the statistical learning model. The model continuously processes flow rate and pressure data to refine leak characterization, enabling both precise location identification and severity assessment that simple flow rate comparison cannot provide.
3Reliability
If traditional detection methods are used, then leaks can be detected, but accurate characterization of leak area and flow rate is not possible, preventing prioritized maintenance
Solution Approach 1:
The patent changes the measurement parameters from simple presence/absence detection to multi-parameter measurement including flow rate, pressure, and their variations over time. By measuring multiple hydraulic parameters simultaneously and processing them through statistical learning models, the system extracts detailed leak characteristics such as leak area and leak flow rate, transforming qualitative leak detection into quantitative leak characterization.
Solution Approach 2:
The statistical learning model serves as an intermediary that processes raw hydraulic measurements and outputs detailed leak characterization information. The model infers leak area and leak flow rate from the hydraulic behavior data, acting as a bridge between sensor measurements and actionable maintenance information, thereby preventing loss of critical leak severity data.
Data Source
AI summary
Method for characterizing a leak in a fluid network, the fluid network including several interconnected areas, in which the fluid network is equipped with at least one flow rate sensor at the inlet of the fluid network and with at least one other hydraulic sensor, of the flow rate or pressure sensor type, configured to provide hydraulic behavior data (Qi, Pi), in which the fluid network is provided with a digital mapping comprising at least the geometry of the fluid network and the location of said hydraulic sensors, and in which a statistical learning model receives as input a set of hydraulic behavior data (Qi, Pi) and provides as output at least one leak characterization data among the leak area (Zf) and the leak flow rate (Qf).