Detect Cavitation Noise with a Piezo Hydrophone

ISEF Category: Engineering Technology: Statics and Dynamics

Subcategory: Naval Systems  ·  Difficulty: Intermediate  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

A propeller can look fine and still be damaged by tiny bubbles. Those bubbles collapse fast, make noise, and can wear down blades before your eyes catch the problem. That means sound can warn you earlier than video. Your project can test how early a simple hydrophone spots cavitation.

What Is It?

Cavitation happens when pressure drops enough that water forms vapor bubbles near a spinning propeller. Those bubbles do not stay quiet for long. They collapse with sharp pressure pulses, and those pulses create a noise pattern that changes as cavitation starts and grows.

Think of it like cracking bubble wrap under water. A smooth propeller sound turns rough, then noisy, then more chaotic. A hydrophone is just a water microphone. In this project, you build one from a piezo disc sealed in epoxy, place it in a bucket or tank, and record the sound while the propeller speed changes.

Machine learning can help you sort those sound patterns. You do not need a huge model. Even a simple classifier can learn the difference between normal flow noise and the first signs of cavitation if your recordings are clean and your labels are good.

Why This Is a Good Topic

This makes a strong science fair project because you can measure a clear signal, compare it against a visible reference, and ask a question with real engineering value. Cavitation matters for boats, pumps, and turbines, so your work connects to marine systems and energy equipment. You can also learn signal processing, classification, and experimental control without needing a university lab.

Research Questions

• How does propeller speed affect the sound features that appear at cavitation onset?
• What is the effect of hydrophone placement on the classifier’s ability to detect cavitation early?
• Does a simple ML model detect cavitation onset earlier than visual inspection?
• To what extent do blade shape or pitch changes alter the acoustic signature of cavitation?
• Which time-domain or frequency-domain features best separate normal flow noise from cavitation noise?
• What is the effect of water depth or tank geometry on cavitation detection accuracy?

Basic Materials

• Small DC motor with propeller or hobby test rig
• Bucket, tub, or clear tank for water testing
• Piezo disc sensor
• Epoxy or waterproof potting compound
• Audio interface or preamp for the sensor signal
• Smartphone or laptop for recording
• Tachometer or motor speed controller
• Ruler or marked stand for sensor placement
• Tripod or clamp for steady video recording
• Quiet workspace with stable lighting
• Notebook for labeling runs
• Safety glasses.

Advanced Materials

• Calibrated hydrophone
• Signal amplifier with known gain
• Data acquisition board or oscilloscope
• Variable-speed motor controller with tach feedback
• Transparent test tank with flow control
• High-speed camera or synchronized video system
• Force or torque sensor for load tracking
• MATLAB, Python, or LabVIEW for analysis
• Acoustic shielding materials
• 3D-printed sensor mount
• Reference propellers with different blade geometries.

Software & Tools

• Python: Runs audio cleaning, feature extraction, and model training with free libraries like NumPy, SciPy, and scikit-learn.
• Audacity: Lets you inspect recordings, trim clips, and check whether your sensor captured usable sound.
• ImageJ: Measures visual onset cues from video frames so you can compare sound against what you see.
• Google Colab: Gives you a free notebook environment for training and testing simple classifiers.
• TensorBoard: Helps you track model performance if you try a neural network later.

Experiment Steps

1. Define the cavitation event you will call your label, and decide how you will confirm it with video or another visual check.
2. Design the sensor mount so the hydrophone stays in the same place for every run.
3. Choose the motor settings and propeller conditions that will create a useful range from no cavitation to clear cavitation.
4. Plan the audio features you will extract, such as spectrum shape, amplitude changes, or burst counts.
5. Build a baseline classifier first, then compare it with a second model or a rule-based threshold.
6. Set up a fair comparison between sound-based detection and visual detection so you can measure which one triggers first.

Common Pitfalls

• Recording motor hum instead of cavitation, which makes the model learn the machine instead of the bubbles.
• Letting the piezo disc touch the bucket wall, which adds vibration artifacts that look like real acoustic events.
• Changing sensor depth between trials, which shifts the signal and breaks comparisons across runs.
• Labeling cavitation too loosely from video, which makes your training data noisy and weak.
• Training and testing on the same run conditions, which makes accuracy look better than it really is.

What Makes This Competitive

A strong version of this project does more than say sound changes when cavitation starts. It measures how early the signal changes, compares multiple feature sets or models, and tests whether the result holds across propellers or tank setups. If you build clean labels, honest controls, and a fair validation split, your work starts to look like real engineering diagnostics rather than a demo.

Project Variations

• Test whether the same hydrophone setup detects cavitation on different propeller blade counts or pitches.
• Compare a simple threshold method with ML models such as random forest, support vector machine, or a small neural network.
• Use a second sensor position or a second tank shape to see how the acoustic pattern changes with geometry.

Learn More

• NOAA Ocean Explorer: Search for free background material on cavitation, propellers, and underwater acoustics in marine systems.
• NASA Technical Reports Server: Search for open technical papers on cavitation, propeller noise, and fluid machinery diagnostics.
• PubMed: Search for review articles on acoustic cavitation detection and signal classification methods.
• MIT OpenCourseWare: Find free fluid mechanics and signal processing course materials that help with the physics and the data analysis.
• Journal of the Acoustical Society of America: Search the journal for peer-reviewed studies on underwater sound, hydrophones, and cavitation noise.