Optimizing Domino And Marble Track Designs

ISEF Category: Engineering Technology: Statics and Dynamics

Subcategory: Computational Mechanics  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A tiny change in shape can make a domino run fail or a marble path work perfectly. That means geometry, not just materials, can control the outcome. With simulation, you can ask the computer to search for the best layout, then test whether the real object behaves the same way.

What Is It?

This project uses a physics simulator to predict how rigid objects move, collide, and settle. Think of it like a chess engine for motion, except the pieces are dominoes, ramps, or marble-track parts. You give the simulator a goal, like a marble reaching a target zone or a domino line finishing cleanly, and it searches for shapes or angles that make that goal more likely.

The key idea is differentiation. In simple terms, the simulator can tell you how a small design change affects the final result. That lets you tune the geometry automatically instead of guessing by hand. JAX helps because it can run fast math and gradient-based optimization, which means the computer can adjust the design step by step toward a better answer.

Then you build the optimized design and compare it with a naive version. The real test matters because simulations always simplify the world. Your project becomes a check on how well the model matches reality, and where it breaks.

Why This Is a Good Topic

This is a strong science fair topic because you can measure a clear outcome, compare two designs, and test whether a computer model predicts real motion. It connects to robot path planning, product design, and engineering simulation, which makes the project feel real instead of theoretical. You can also learn a lot from one project, including modeling, optimization, error analysis, and validation.

Research Questions

• How does optimized geometry change the success rate of a domino or marble-track layout compared with a hand-designed layout?
• What is the effect of small changes in ramp angle on the final position of the marble?
• Does a simulator-trained design transfer better to real hardware than a design chosen by trial and error?
• To what extent does adding friction uncertainty change the optimizer's chosen layout?
• Which objective function gives the most reliable real-world performance, target position or path stability?
• How does the number of control points in the geometry affect simulation accuracy and final design quality?

Basic Materials

• Laptop or desktop computer with enough memory for local simulation.
• JAX installed in Python.
• 3D printer or access to a school maker space.
• PLA filament or other standard 3D printing material.
• Dominoes or small marbles, depending on your chosen layout.
• Measuring tape or ruler.
• Protractor or digital angle finder.
• Smartphone camera for motion recording.
• Flat test surface or tabletop.
• Tape, clamps, or putty for holding parts in place.

Advanced Materials

• High-resolution 3D printer with repeatable dimensional accuracy.
• Force sensor or load cell for impact measurements.
• High-speed camera or a phone with slow-motion mode.
• Motion capture software or calibration board.
• Precision scale for checking printed part mass.
• Interchangeable track materials with known friction differences.
• Computer with GPU access for faster JAX runs.
• Laser level or digital leveling tool for alignment checks.
• Environmental sensor for temperature and humidity logging.
• CAD software for parametric geometry design.

Software & Tools

• Python: Runs the simulator, optimizer, and data analysis scripts.
• JAX: Computes gradients that help tune the geometry automatically.
• NumPy: Stores numeric data and supports analysis of trial results.
• Matplotlib: Plots loss curves, trajectories, and comparison graphs.
• ImageJ: Measures positions frame by frame from video when you need manual validation.

Experiment Steps

1. Define one final-state target, such as a landing point, stopping zone, or completion outcome.
2. Choose the smallest set of geometry parameters you want the optimizer to change.
3. Build a simple simulator first, then check that it reproduces obvious real-world cases.
4. Add an objective function that scores how close each layout comes to your target.
5. Compare optimized designs with hand-made baselines using the same test conditions.
6. Plan a validation method that measures where simulation and reality agree, and where they do not.

Common Pitfalls

• Letting the optimizer change too many geometry parameters, which makes the search unstable and hard to interpret.
• Ignoring friction and bounce differences between printed parts and the simulator, which causes the real layout to miss the target.
• Testing only one optimized design, which hides whether the result was luck or a repeatable improvement.
• Changing the starting release point between trials, which confounds geometry effects with launch variation.
• Using video without a fixed camera angle and scale marker, which makes position measurements drift across runs.

What Makes This Competitive

A strong version of this project does more than prove that optimization works. It compares several objective functions, tests sensitivity to noise, and checks whether the simulated optimum still wins after printing. You can raise the level by measuring how prediction error changes across different shapes, materials, or contact conditions. Careful validation and clean statistics matter more than fancy visuals.

Project Variations

• Optimize a domino chain so the final domino lands in a specific zone instead of just falling down.
• Tune a marble track to hit a target exit point while keeping speed changes small.
• Compare one optimizer that uses gradients with one that uses random search for the same geometry problem.

Learn More

• MIT OpenCourseWare, Mechanics and materials-related courses: Search MIT OpenCourseWare for rigid body dynamics, optimization, and computational mechanics lecture notes.
• JAX documentation: Read the official guides to automatic differentiation and accelerated numerical computing.
• USGS publications and educational pages: Search for friction, slope stability, and measurement basics that help with physical validation and experimental design.
• NASA NTRS: Search for papers on trajectory optimization, simulation validation, and numerical methods used in engineering models.
• Journal of Computational Physics: Search for review articles on differentiable simulation and physics-based optimization through your school library or PubMed-style journal access if available.