Validation

How close are
the predictions?

We compare ThrustLab’s predictions with wind-tunnel and motor-bench measurements, then publish the differences.

Published benchmark · PROM 5.3 · prom-rs/0.4.0

Measured on the production solver server on September 8, 2026. Results apply to this model and engine build.

75propellers

APC Thin Electric, Sport and Slow Flyer. Compared with university wind-tunnel measurements.

95motors

T-Motor models in the published bench envelope. Compared with measured electrical and mechanical performance.

These are the full benchmark counts. Each plot identifies its selected measurements and conditions.

Across all 9,842 propeller measurements, mean absolute coefficient error is Cₜ 0.0138 and Cₚ 0.0091. These are coefficient differences, not percentages. The plots below separate families and operating conditions.

Loading the saved production measurements…

03 / Motors

Current draw.
And full-throttle RPM.

Current is checked at measured shaft speed and load. The speed check asks how fast the motor can turn at full throttle.

Separate performance tests

The original performance test group, rerun on this release. These are measurements of current and speed, not simulated reference values.

Current draw
7.2%
average difference · 2,209 measurements
Full-throttle RPM
6.1%
average difference · 175 measurements

Separate bench dataset

A different motor-bench dataset in the same campaign. We keep it separate so a combined average cannot hide differences between the two sets.

Current draw
9.6%
average difference · 570 measurements
Full-throttle RPM
14.8%
average difference · 24 measurements

Using the results

Know what the evidence covers.

Measured at the component level

Propeller thrust and shaft power, plus motor current, speed and efficiency. The report gives the tested sizes, RPM ranges and flight conditions.

See the tested operating ranges ↗

Still needs system-level testing

This dataset does not validate battery behavior, temperatures, complete powertrains, aircraft flight time or autopilot missions. Coaxial and angled-flow predictions also lack measured validation here.

Use the model to narrow your choices.

Compare candidate components, leave room for the measured error, and bench-test the final setup. An average error is useful context for a design decision; it is not a guaranteed error limit for your aircraft.

Using a custom component? These error figures come from known components. They do not apply to user-supplied geometry or specifications.