📘 Module 9 — Troubleshooting & Field Repairs

9.1 Engineering Design Process

The Iterative Loop: Design, Build, Test, Analyze, Redesign

The Engineering Design Process (EDP) is not a checklist — it is a feedback-driven cycle that professional engineers and top Science Olympiad teams treat as their core methodology. Understanding why the loop exists, and how to execute it rigorously, is the difference between a team that improves systematically and one that guesses and hopes.

The classical loop has five phases:

1. Design — Propose a configuration based on physics principles and prior data.
2. Build — Construct the vehicle to spec, noting any fabrication deviations.
3. Test — Run the vehicle under controlled, documented conditions.
4. Analyze — Interpret results against predictions; identify what matched and what didn't.
5. Redesign — Formulate a specific hypothesis and modify the design accordingly.

Then the loop repeats. The key word is loop — not a straight line from concept to competition. Teams that treat the EDP as a checklist ("we built it, we tested it once, we're done") are forfeiting systematic improvement.

Why Documentation Prevents Accidental Regression

The most common mistake in iterative design is losing a working configuration. Imagine your team discovers a skirt geometry that produces excellent lift at low fan speed. In the next iteration, you change both the skirt geometry and the fan mounting angle. Performance drops. You're now unsure whether to revert the skirt, the mount, or both — and you no longer have a record of the exact original dimensions.

This is accidental regression: inadvertently moving backward while trying to move forward.

Prevention requires a design log: a written (or digital) record of every configuration tested. Each entry should capture:

• Configuration ID (e.g., "v3.2")
• What changed from the prior version (e.g., "Replaced fabric skirt with polythene, sealed with hot glue, 3 mm overlap at seams")
• Test conditions (surface, battery charge level at start, ambient temperature)
• Results (run time, distance, observed anomalies)
• Hypothesis for next iteration

This documentation discipline transforms testing from "poking the vehicle and seeing what happens" into structured experimentation.

Hovercraft-Specific Design Iterations

In hovercraft design, iterations typically cluster around three subsystems: lift, thrust, and structural integrity. Each subsystem has its own design variables:

Lift subsystem variables:

• Skirt material (polythene film, foam, fabric)
• Skirt attachment method and seal quality
• Plenum hole geometry (number, size, placement)
• Hull surface area (determines required pressure: P = F/A)

Thrust subsystem variables:

• Fan model and blade pitch
• Motor voltage/current draw
• Propeller angle relative to hull
• Shroud presence and geometry

Structural variables:

• Hull material and thickness (foam board, balsa, corrugated plastic)
• Weight distribution (affects required lift pressure)
• Fastener methods (tape, glue, screws)

A mature iterative process focuses one design cycle on one subsystem. Do not change the skirt and the motor in the same test — you will not know which change drove the result.

Setting Entry and Exit Criteria

Professional engineering projects use entry and exit criteria for design phases. Science Olympiad teams benefit from the same discipline:

• Entry criteria for test: Battery charged to >90%, surface swept clean, cones placed at target distance, data sheet prepared.
• Exit criteria to advance to next iteration: Run three trials; if standard deviation of run times exceeds 15% of mean, investigate root cause before changing the design.

Without exit criteria, teams tend to redesign prematurely — changing the vehicle after one bad run rather than determining whether the bad run was a fluke.

The Role of Predictive Modeling

Each cycle should begin with a prediction: "I expect this change to reduce run time by approximately X seconds because..." Comparing predictions to results calibrates the team's mental model of the system. When predictions are consistently accurate, the team has deep understanding of their hovercraft. When predictions are consistently wrong, there is a hidden variable — perhaps skirt flex under load, or battery voltage sag — that the mental model is missing.

For example: if increasing fan speed by one setting (say, from 60% to 70% throttle) is expected to reduce 3-meter run time by ~0.4 s (based on F = ma estimates and prior data), but the actual improvement is only 0.1 s, the likely culprit is air leakage eating the excess thrust before it translates to forward acceleration.

Competition Tip: Freeze the Design Early

A critical but counterintuitive EDP principle for competition is design freeze: stopping all modifications 1–2 weeks before competition day. After freeze, the team runs only validation tests — confirming that performance is stable and repeatable. Late design changes introduce untested configurations and risk introducing new failure modes under competition conditions.

The best teams arrive at competition with a vehicle they have run 50+ times across multiple sessions. They know its quirks, its consistent behavior, and its failure thresholds. That knowledge is worth more than any last-minute tweak.

9.2 Measuring Run Performance

Why Measurement Consistency Is Non-Negotiable

Data is only as good as the measurement process that produced it. In hovercraft testing, two teams might test the "same" vehicle configuration and record completely different run times — not because the vehicle behaved differently, but because one team eyeballed distances while the other used tape marks, or one timed from button press while the other timed from craft motion.

For data to drive useful design decisions, measurements must be consistent across all tests. This section establishes the standards and tools that make run performance data trustworthy.

Setting Up a Repeatable Test Track

The ideal test track mimics competition conditions as closely as possible. Key setup steps:

Surface: Use a hard, flat, clean surface. Gymnasium floors, smooth concrete, or polished tile are best. Carpet introduces variable drag depending on pile direction, humidity, and surface age — avoid it unless the competition venue is carpeted. Sweep the track before every session.

Distance markers: Place cones or tape marks at fixed intervals: for example, at 0.5 m, 1.0 m, 1.5 m, 2.0 m, 2.5 m, and 3.0 m from the launch line. The launch line itself must be a fixed physical reference (a strip of tape, not a person's foot).

Starting position: The front edge of the craft should be placed consistently at the launch line for every run. A small tape notch on the hull can help the operator align quickly without measuring each time.

Wind and doors: Close doors and windows to minimize ambient airflow. Even a gentle breeze (1–2 m/s) can add or subtract measurable time on a low-drag hovercraft. Record any unusual air conditions in the test log.

Measuring Distance

Distance measurement has two practical approaches:

1. Final rest position: After the craft stops, measure from the launch line to the nearest edge of the craft. This method captures total run distance but is sensitive to any post-stop sliding on a very smooth surface.

2. Last cone cleared: Record which distance cone the craft passed before stopping. This is less precise but faster and less prone to post-run movement errors.

For scientific rigor, use a steel tape measure anchored at the launch line. Record to the nearest centimeter. If the craft rotates or veers off-axis, note the deviation angle and measure along the original travel axis.

Measuring Time

Time measurement should capture two intervals:

• T₁ (Lift time): Time from switch activation to first observable craft levitation. This separates lift system performance from thrust performance.
• T₂ (Run time): Time from first craft movement (translation across the track) to rest. This is the primary performance metric.

Use a digital stopwatch or smartphone timer. Reaction time introduces ~0.15–0.25 s error per timing event; across a 4-second run, this is roughly 4–6% relative error. To reduce reaction time error, use video capture: record runs at 60+ fps and measure frame-count intervals in post-processing for precision to ±0.016 s.

The Target Score Function

Competition scoring typically combines distance and time into a single function. A common form is:

Score = k₁ × Distance − k₂ × Time

where k₁ and k₂ are event-specific constants. Understanding this function helps teams prioritize: if k₂ (time penalty) is large, shaving 0.5 s off run time may be more valuable than adding 10 cm of distance. If k₁ (distance reward) is large, ensuring the craft reaches the target zone reliably outweighs speed.

During testing, plot your craft's current (distance, time) coordinates against iso-score contours of the scoring function. This reveals whether your current design is limited by distance (needs more thrust) or time (needs less drag or friction).

For a simplified example, if Score = 10 × Distance(m) − 2 × Time(s):

• A 2.0 m run in 3.0 s scores: 10(2.0) − 2(3.0) = 14.0
• A 2.5 m run in 5.0 s scores: 10(2.5) − 2(5.0) = 15.0
• A 3.0 m run in 8.0 s scores: 10(3.0) − 2(8.0) = 14.0

In this case, the 2.5 m at 5.0 s beats both extremes — illustrating why teams should optimize for the scoring function, not for either metric in isolation.

Tracking Metrics Across Test Sessions

Maintain a run log spreadsheet with columns for: date, run ID, configuration version, surface condition, battery SoC (state of charge) at start, T₁, T₂, distance, calculated score, and notes. This allows performance trends to be analyzed over time and across configuration changes.

Plot run time and distance as time-series over your test sessions. Improving performance should show a downward trend in run time and upward trend in distance. A sudden spike in run time in an otherwise stable series usually indicates a mechanical issue — skirt damage, battery degradation, or loose hardware — rather than random variability.

Competition Tip: Pre-Run Baseline

At competition, before your first official run attempt, request a "practice" or "scratch" run if event rules allow. Use it to calibrate your mental model of the specific floor surface, temperature, and lighting conditions. Competition floors are often harder and cleaner than practice surfaces, which can increase craft speed and distance. Teams that pre-calibrate adjust their fan speed settings accordingly instead of being surprised.

9.3 Controlled Variable Testing

The Principle of One-Variable Isolation

Controlled variable testing is the experimental analog of surgical precision: change one thing, measure the effect, then decide what to do next. This principle is foundational to all scientific inquiry, and it is just as essential in an engineering lab as in a chemistry experiment. For Science Olympiad hovercraft teams, it is the methodology that separates systematic improvement from random tinkering.

The formal framework is simple:

• Independent variable: The single element you deliberately change between tests.
• Dependent variable: The metric you measure to assess the effect (e.g., run time, lift-off time, drift angle).
• Controlled variables: Every other design parameter, test condition, and measurement method — held constant across the two tests you are comparing.

Violating the controlled-variable principle by changing two or more elements simultaneously means the test data cannot tell you which change caused the observed outcome. You might get lucky and improve performance, but you will not understand why, and you will not be able to reproduce or extend the improvement reliably.

Common Design Variables in Hovercraft Testing

Hovercraft design offers a rich space of testable independent variables. The following are the highest-leverage candidates for systematic experimentation:

Skirt type and geometry:

• Open plenum vs. segmented skirt vs. bag skirt
• Skirt attachment overlap (e.g., 5 mm vs. 10 mm)
• Number of segments in a segmented skirt
• Skirt material (polythene film vs. mylar vs. fabric)

Fan system:

• Fan speed setting (% throttle or discrete switch positions)
• Propeller pitch angle
• Number of lift fans vs. thrust fans
• Fan mounting angle relative to hull

Power system:

• Battery charge level at start of run (fully charged vs. 80% vs. 60%)
• Battery chemistry (NiMH vs. LiPo, if rules allow)
• Wire gauge and connector resistance (affects voltage sag under load)

Hull configuration:

• Total craft mass (add/remove ballast in controlled increments)
• Hull footprint area (affects required lift pressure: P = mg/A)
• Center of mass position (fore/aft balance affects pitch stability)

Each of these is a legitimate independent variable for a controlled test. Each test should isolate exactly one of them.

Designing the Test Matrix

For systematic experimentation, lay out a test matrix in advance: a table with configurations on rows and performance metrics on columns. Plan which variable to test in each session and what values to step through.

Example test matrix for skirt attachment overlap:

All other parameters (fan speed, battery SoC, surface) are controlled — locked to the same values for all four tests. The test matrix also serves as documentation: when you revisit data weeks later, you know exactly what you were testing and why.

Battery Charge Level: The Hidden Confound

One of the most commonly overlooked controlled variables is battery state of charge (SoC). A LiPo battery at 100% charge delivers approximately 4.2 V per cell; at 50% charge, it may deliver only 3.7 V per cell. For a 2S (2-cell) pack, this is the difference between 8.4 V and 7.4 V driving the fan motor — roughly a 12% reduction in motor speed, which significantly affects both lift and thrust.

If Test A is run on a fresh charge and Test B is run on a partially depleted pack, any performance difference is confounded by battery voltage, not the design variable being studied. Standard practice: recharge the battery (or replace it with an identical freshly charged pack) between every test configuration. Record the resting voltage before each run as a sanity check.

Handling Test-to-Test Variability

Even with all controlled variables locked, consecutive runs of the same configuration will not produce identical results. Sources of intra-configuration variability include:

• Small differences in craft starting position or alignment
• Micro-variations in floor debris or moisture
• Fan blade vibration modes that change with temperature
• Operator reaction time variability

Standard practice is to run 3–5 trials per configuration and report the mean and standard deviation. A result is meaningful only if the performance difference between two configurations exceeds the within-configuration variability. Formally:

Δ_mean > 2 × σ_pooled (a rough two-sigma criterion for practical significance)

If configuration A averages 3.2 s ± 0.3 s and configuration B averages 3.0 s ± 0.3 s, the 0.2 s difference is within noise. Do not draw conclusions. Run more trials or look for a larger effect to test.

Multi-Variable Changes Obscure Root Cause

Consider a team that simultaneously changes the skirt material (polythene → mylar) and the fan speed setting (60% → 70%). Run time improves from 4.5 s to 3.8 s. Did the better skirt material reduce air leakage? Did the higher fan speed produce more thrust? Did both contribute equally? The team has no way to know. They have improved the vehicle, but they have not learned anything about why it improved — and therefore cannot extrapolate to the next iteration.

This kind of shotgun experimentation is a common beginner trap. It feels productive because performance goes up, but it produces a dead end. Once the team needs to troubleshoot a regression or adapt to a different surface, they have no model to reason from.

Competition Tip: Test on the Correct Surface

If possible, find out the competition venue floor type in advance (gymnasium hardwood, polished concrete, rubber tile) and practice on the same material. A craft optimized for smooth tile may produce unexpected friction on gymnasium wood floors with a slightly rougher grain — which looks like a controlled-variable failure but is actually a surface mismatch. Always record surface type as a controlled variable in your test log.

9.4 Data Tables & Graphs

Data as the Engineer's Evidence Base

Engineers do not operate on intuition alone — they operate on evidence. In hovercraft design, the difference between a team that iterates intelligently and one that flails through dozens of changes without progress often comes down to whether they kept rigorous data and used graphs to understand their craft's behavior. Data tables and graphs are not administrative overhead; they are the primary tools for translating raw test results into actionable design insights.

Structuring a Run Data Table

A well-designed data table captures enough information to reconstruct the test conditions months later. The minimum recommended fields for a hovercraft run log are:

Avoiding the temptation to "remember" details from a run and write them down later — memory is unreliable and systematically optimistic. Record data immediately after each run.

Types of Graphs and What They Reveal

Raw tables are hard to interpret at a glance. Converting data to graphs reveals trends, nonlinearities, and anomalies that are invisible in rows of numbers.

Speed vs. Fan Setting Curve (Response Curve)

Plot average craft speed (total distance / run time, in cm/s) on the Y-axis against fan speed setting (% throttle) on the X-axis. This is the most informative single graph in hovercraft development. It reveals:

• The onset threshold: the minimum fan setting at which the craft reliably hovers (below this, lift is insufficient and the craft drags rather than glides).
• The linear region: the range over which speed increases proportionally with fan setting — this is the operating regime for fine-tuning.
• The saturation plateau: at high fan settings, speed stops increasing meaningfully. This identifies diminishing returns and battery waste.

A typical response curve starts flat (insufficient lift), rises steeply through a threshold, then gradually flattens as aerodynamic drag and surface friction limit speed gains.

Distance vs. Run Number (Repeatability Plot)

Plot run distance vs. sequential run number across a test session. This reveals battery degradation effects: if distance drops steadily over 10 consecutive runs without recharging, the battery voltage sag is affecting performance. The slope of the decline characterizes the magnitude of the effect and guides decisions about recharging frequency during competition.

Score Contour Plot

For the scoring formula Score = k₁ × D − k₂ × T, plot iso-score contours in (Distance, Time) space. Your current test results appear as scattered points. This visualization immediately shows whether current runs cluster near high-score contours or whether a shift in either axis would dramatically improve score. It also identifies tradeoff directions: sometimes a run that covers slightly less distance in significantly less time produces a higher score.

Skirt Leak Test: Pressure vs. Time Decay

For a static lift test, inflate the plenum by running the lift fan for 5 seconds, then kill power. Measure the time for the craft to settle fully onto the surface. Plot pressure-decay curves for different skirt configurations. A slower decay indicates lower leakage rate — quantifiable as a leak coefficient that can track improvements iteration by iteration.

Using Graphs to Predict Competition Settings

Once you have a well-characterized speed vs. fan setting curve, you can use it predictively. Suppose competition scoring rewards runs closest to a target speed of 35 cm/s. From your response curve, you can identify the exact fan setting that produces that speed on your test surface. You then apply a surface correction factor if the competition floor is different (estimated from the difference in friction coefficients — μ_comp vs. μ_test).

The correction is approximate:

T_comp ≈ T_test × (μ_comp / μ_test)^0.5

for runs dominated by sliding friction. This is an approximation valid for low-speed, friction-dominated regimes. Use it as a first-order estimate, then refine with any available data from the competition floor.

Graphing Best Practices

• Label all axes with units (seconds, centimeters, % throttle).
• Include error bars (±1 standard deviation) on mean values so the graph communicates reliability.
• Mark configuration changes with vertical lines or color breaks on time-series plots so improvements can be attributed to specific design decisions.
• Log the graph in your design notebook along with the date and the design question it was meant to answer.

Competition Tip: Pre-Competition Data Review

In the week before competition, print or organize your best response-curve graph and bring it as part of your allowed documentation. If you are permitted to use notes during the event window, you can reference your fan setting vs. predicted time chart rather than guessing. A simple lookup table derived from your graphs — "Fan setting 65%: expected 3.4 s run, 2.8 m distance" — is actionable intelligence that can guide run strategy in real time.

9.5 Failure Mode Analysis

Engineering Risk Management at Competition Scale

In professional engineering, Failure Mode and Effects Analysis (FMEA) is a structured methodology for identifying all the ways a system can fail, ranking those failures by severity and likelihood, and then implementing mitigations before deployment. For Science Olympiad Hovercraft, a simplified FMEA gives Division C teams a powerful mental framework for anticipating problems before they occur on competition day — when there is no time to improvise.

The FMEA process asks three questions for each potential failure:

1. What could go wrong? (Failure mode)
2. How bad is it? (Severity: does it prevent scoring entirely, or only reduce score?)
3. How likely is it? (Probability: has it happened before? Under what conditions?)

Risk Priority = Severity × Probability × Detectability. Failures that are severe, likely, and hard to detect before a run deserve the most mitigation investment.

Cataloging Hovercraft Failure Modes

The following are the most common failure categories observed in practice and competition:

1. Skirt Failure

• Modes: Tear at seam, delamination from hull, puncture from floor debris, fold-over during levitation causing air loss from one side.
• Symptoms: Craft tilts, fails to hover symmetrically, or drags on one edge.
• Severity: High — a major skirt tear often prevents hover entirely (score = 0 for that run).
• Likelihood: Medium-High — skirts are the most mechanically stressed component. Mylar and polythene are particularly vulnerable to crease-induced micro-tears at fold lines.
• Mitigation: Carry pre-cut replacement skirt material, tape, and glue at impound. Practice skirt replacement under 2-minute time pressure. Inspect seams with a flashlight at the start of the event window.

2. Insufficient Lift

• Modes: Fan speed set too low, battery depleted, plenum hole blocked by debris, skirt pinhole leak pattern.
• Symptoms: Craft moves sluggishly or skates with friction; audible floor-contact noise.
• Severity: High — insufficient lift means high friction, dramatically increasing run time or causing the craft to stop short.
• Likelihood: Medium — typically occurs when battery SoC is low or after skirt damage.
• Mitigation: Always start with a fully charged battery. Check plenum clearance before each run. Have a secondary skirt inspection protocol (see Topic 11.2).

3. Steering Drift

• Modes: Asymmetric skirt inflation, off-center propeller thrust, warped hull causing the craft to curve.
• Symptoms: Craft veers left or right rather than tracking straight; may exit the target lane.
• Severity: Medium — depending on competition layout, drifting off-axis may reduce distance score (craft travels on a diagonal) or invalidate the run.
• Likelihood: Medium — warped hulls from heat or pressure, and slightly misaligned propellers, are common causes.
• Mitigation: Test specifically for straightness — mark a center line on the test track and measure lateral deviation after a 3-meter run. Recenter the propeller mount if deviation exceeds ~5 cm. Carry a small level and feeler gauge for pre-run checks.

4. Battery Drain During Multiple Runs

• Modes: Voltage sag over successive runs without recharge, excessive current draw from a high-power fan.
• Symptoms: Progressive increase in run time across the event window; first run is fast, third run is noticeably slower.
• Severity: Medium — reduces score but usually does not eliminate scoring entirely.
• Likelihood: Medium-High — very common in teams using high-drain motors without characterizing battery endurance.
• Mitigation: Run a battery endurance test: conduct 5 consecutive runs without recharging and measure the performance drop. If third-run performance is >10% worse than first-run, either use a higher-capacity battery or reduce fan power to extend runtime.

5. Structural / Fastener Failure

• Modes: Fan detaches from mount, hull cracks at a stress concentration, loose wires cause intermittent motor cutoff.
• Symptoms: Motor suddenly stops, unusual vibration, visible damage.
• Severity: High — motor detachment typically prevents all further runs.
• Likelihood: Low-Medium — rare in well-built vehicles, but catastrophic when it occurs.
• Mitigation: Apply threadlocker or secondary tape/zip-tie to all fan mounts. Conduct a pre-competition "shake test": gently flex and tap all joints looking for movement or rattling.

Contingency Planning: The Competition Kit

For each high-priority failure mode, prepare a contingency response plan and a corresponding spare parts kit in the impound bag:

Practice each repair under time pressure. At competition, you may have only the remaining time in the 8-minute window to execute a fix and still complete another run.

Competition Tip: Failure Mode Priority at Competition

Division C teams competing at invitational and state levels should perform a formal FMEA review of their specific vehicle two weeks before competition. Rank failure modes by their actual observed frequency from your test log data — not just hypothetical risk. Failures that have never occurred in 50 practice runs are lower priority than failures that have occurred twice in the past week. Let data drive risk mitigation investment.

9.6 Practice Runs & Repeatability

The Repeatability Principle in Engineering

In manufacturing and quality engineering, repeatability refers to the ability of a process to produce consistent results across successive trials under identical conditions. A CNC machine that mills a part to ±0.01 mm on the first run and ±0.05 mm on the second run has poor repeatability — and poor repeatability makes quality assurance impossible.

The same concept applies directly to Science Olympiad hovercraft. A craft that achieves an exceptional run once but cannot reproduce it is, from a competition-strategy standpoint, nearly worthless. Every run at competition is an independent event: you cannot bank the exceptional run and spend it later. What you earn is a distribution of scores across your actual competition attempts, and the expectation (in the statistical sense) of your final score depends on both your mean performance and your variance.

Quantifying Repeatability

Repeatability is formally quantified by the coefficient of variation (CV): the ratio of the standard deviation to the mean, expressed as a percentage.

CV = (σ / μ) × 100%

For a hovercraft design aiming for competition-level reliability, target CV < 10% for run time across at least 10 consecutive test runs under consistent conditions. A CV above 20% indicates the craft is unreliable — something in the design is sensitive to small perturbations (battery state, starting position, skirt orientation) in a way that should be addressed before competition.

Example: A craft's run times over 10 trials are: 3.8, 4.1, 3.7, 5.2, 3.9, 4.0, 3.8, 3.6, 5.0, 3.9 seconds.

• Mean: 3.9 + (one outlier treatment needed)
• The two outliers (5.0, 5.2) indicate occasional mode-switching failure — perhaps the skirt folds over under certain initial conditions, dramatically increasing friction.
• CV including outliers ≈ 16% — borderline. The root cause of the outliers should be investigated and eliminated.

Why Moderate-But-Reliable Beats Peak-But-Erratic

Consider two hypothetical hovercraft competing under a scoring formula where a 2.5 m run in 3.5 s earns 18.0 points:

Craft A (Peak Performer): Run times: 2.8, 2.9, 8.0, 3.1, 7.5 seconds over 5 runs. Distances: 3.0, 3.0, 0.5, 2.8, 0.8 m. When it works, it's fast. But it fails catastrophically ~40% of the time.

Craft B (Reliable Performer): Run times: 3.8, 3.9, 4.0, 3.7, 4.1 seconds. Distances: 2.4, 2.5, 2.3, 2.5, 2.4 m. Never spectacular, never fails.

If competition allows 3 attempts and scores the best 2:

• Craft A: Best 2 from (5 runs sample): likely 2 good runs for ~35+ points, but a ~24% chance that both best runs fall in the failure regime.
• Craft B: Best 2 runs are consistently ~15.0–16.0 points each, total ~30–32 points, zero failure risk.

Depending on scoring specifics, Craft A has a higher ceiling but a significantly higher variance. In large competitions where placement matters more than raw score ceiling, reliability often wins. The team with Craft B knows exactly what score to expect; the team with Craft A is gambling.

Sources of Repeatability Problems

When CV is high, the root cause is usually one of the following:

Skirt instability: The skirt folds over or inflates asymmetrically depending on microscopic placement variations. Fix by adding skirt stiffeners or adjusting the attachment geometry.

Starting orientation sensitivity: The craft tracks differently depending on the exact angle at launch. Fix by adding a launching guide or practicing a standardized placement routine with physical alignment marks.

Battery voltage variability: If tests are conducted at different SoC levels, run time variance inflates. Fix by always starting at the same SoC (recharge between sessions, not between individual runs unless characterizing battery effects).

Fan oscillation modes: At certain speeds, fan blades can enter resonance modes that cause irregular thrust. Fix by testing at discrete fan speed increments and avoiding settings near resonance points.

Operator variability: Different operators (or the same operator on a bad day) launch the craft with slightly different force, position, or timing. Fix by standardizing the launch procedure and practicing under pressure.

Designing for Repeatability: Practical Strategies

Repeatability should be a design requirement, not an afterthought:

1. Robust skirt design: A segmented skirt with individual segment support is more repeatable than a single-piece bag skirt that can fold unpredictably.
2. Fixed-position starting guides: A launch notch or track guide that aligns the craft identically every run eliminates orientation variability.
3. Standardized startup procedure: Define a precise sequence — switch on lift, wait for full levitation (confirmed by visual or tactile check), then release — and practice it until it is automatic.
4. Battery management protocol: Test and document the performance decay curve for your battery. Use a discharge budget: start each competition run at ≥90% SoC if possible.

Competition Tip: The Repeatability Test Protocol

Two weeks before competition, run the repeatability protocol: 10 consecutive runs, same operator, same conditions, no adjustments between runs. Calculate mean and CV. If CV < 10%, the craft is competition-ready in terms of reliability. If CV > 15%, spend the remaining time diagnosing and resolving the root cause rather than trying to improve peak speed. At competition, reliability earns points; brilliance that misfires does not.