Leak Detection in Water Networks With EPANET

ISEF Category: Environmental Engineering

Subcategory: Water Resources Management  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

A tiny leak can waste thousands of gallons before anyone notices. Water networks act a lot like a spiderweb, so one weak point can affect pressure far from the break. You can test where sensors should go so the system spots leaks faster. That makes this a great project if you like maps, data, and smart design.

What Is It?

This project studies how well a water network can detect leaks when you place pressure sensors in different spots. EPANET is a computer program that simulates how water moves through pipes, how pressure changes, and how a leak can disturb the system. Think of the network like a set of connected straws. If one straw gets a hole, the pressure pattern changes, but some sensor locations will notice that change better than others.

A genetic algorithm is a search method inspired by natural selection. You start with many possible sensor layouts, test them in the simulation, keep the best ones, and mix them to make new layouts. Then you repeat the process. In this project, you are not building the real water system. You are asking a design question, which sensor layout gives the strongest leak signal with the fewest sensors?

Why This Is a Good Topic

This is a strong science fair topic because you can change one clear variable, sensor placement, and measure one clear outcome, leak-detection sensitivity. It connects to a real problem that cities care about, since water loss costs money and strains infrastructure. You can learn network modeling, optimization, and data analysis without needing a wet lab. That makes it a serious engineering project that still fits a student workspace.

Research Questions

• How does the number of pressure sensors affect leak-detection sensitivity in a synthetic water network?
• What is the effect of sensor placement near junctions versus near endpoints on leak detection accuracy?
• Does a genetic algorithm find better sensor layouts than random placement?
• To what extent do pipe length and network size change the best sensor configuration?
• Which leak locations are hardest to detect with pressure sensors in the model?
• How does the assumed leak size change the ranking of sensor placements?

Basic Materials

• Laptop or desktop computer with enough memory to run EPANET models.
• EPANET software.
• Spreadsheet software such as Excel or Google Sheets.
• Python with NumPy and pandas.
• Python package for genetic algorithms, such as DEAP.
• Graphing software for plots and network visuals.
• Water network test case file or a synthetic network you build yourself.
• Notebook for tracking model settings, sensor layouts, and results.

Advanced Materials

• Access to a calibrated municipal network dataset or published benchmark network.
• EPANET toolkit for scripted runs.
• Python with SciPy, NumPy, pandas, and NetworkX.
• DEAP or another evolutionary computation library.
• GIS software for mapping sensor locations onto a network.
• Statistical analysis software for permutation tests or mixed models.
• Version control system such as Git for tracking simulation changes.
• High-performance computer access if you plan many simulation runs.

Software & Tools

• EPANET: Simulates water flow, pressure, and leak effects in a pipe network.
• Python: Automates sensor placement trials, leak scenarios, and result analysis.
• DEAP: Runs the genetic algorithm that searches for strong sensor layouts.
• Excel or Google Sheets: Organizes simulation outputs and compares candidate layouts.
• ImageJ: Not needed here, so skip it unless you also analyze network maps as images.

Experiment Steps

1. Define the network you will model, then choose whether you want a synthetic layout or a published benchmark system.
2. Decide the leak signal you will measure, such as pressure drop, detection rate, or false alarm rate.
3. Set the rules for candidate sensor layouts, including how many sensors each layout may use.
4. Build a baseline set of random or hand-picked layouts so you have something to beat.
5. Design the genetic algorithm search plan, including how you score each layout and how you compare winners.
6. Plan a validation test that checks whether the best layout still works when leak size, location, or demand pattern changes.

Common Pitfalls

• Changing the network model and the sensor count at the same time, which makes it hard to tell what caused the result.
• Scoring layouts only by average performance, which can hide sensor setups that fail badly on a few leak locations.
• Testing the genetic algorithm on one leak pattern, which can overfit the layout to a single case.
• Ignoring false positives, which makes a layout look better than it really is.
• Forgetting to compare against a simple baseline, which leaves you unable to prove the algorithm helped.

What Makes This Competitive

A stronger project does more than find one good layout. It compares several network shapes, leak sizes, or demand patterns, then tests whether the same sensor strategy still works. You can raise the level by using clear validation, strong statistics, and a fairness check against random placement. A top project often asks not just what works, but why it works in certain parts of the network.

Project Variations

• Test how sensor placement changes when the network has loops versus mostly straight branches.
• Compare pressure sensors with flow sensors to see which detects leaks earlier in the same synthetic network.
• Use a different optimization method, such as simulated annealing, and compare it with the genetic algorithm.
• Study how night demand versus peak demand changes leak detectability in the model.

Learn More

• EPA Water Infrastructure resources: Search the EPA site for drinking water distribution system guidance and leak loss topics.
• EPANET User Manual: Find the official documentation through EPA or search for the EPANET manual PDF.
• USGS Water Science School: Read background on water supply, distribution, and water loss topics on the USGS site.
• NOAA Climate.gov: Look for articles on drought, water stress, and infrastructure impacts.
• PubMed: Search for review articles on leak detection in water distribution systems and pressure-based monitoring.
• MIT OpenCourseWare: Search for free courses in optimization, networks, and algorithm design.