Random Vibration Checks: What Miles' Equation Gets Right and Misses
Miles' equation is the quick random vibration check most engineers reach for, and on a simple bracket it holds within a few percent. We run four generic bracket cases under the NASA GEVS component levels to show where it holds, why a stiffer bracket can push more acceleration into the electronics, and how a second mode can shake the free end of a bracket to 29.42 Grms while the unit itself sees 5.43.
Read the full article below for the methodology, the data, and the engineering decisions behind it.
Most random vibration checks on a bracket start the same way. Find the first natural frequency, read the input level at that frequency, put both into Miles' equation with a Q of 10, and multiply the answer by three. It is a good habit and often a good answer. This note looks at when it is, using the public GEVS component levels from NASA Goddard and a small worked check on a generic bracket carrying a 0.5 kg electronics unit.
1. The levels the bracket has to see
GSFC-STD-7000B, the General Environmental Verification Standard (Revision B, April 2021), gives two random vibration tables for components that are useful reference points even outside spaceflight. Table 2.4-4 is the minimum workmanship level, which the standard says all components shall see as a workmanship screen, mainly aimed at electrical, electronic and electromechanical hardware of 45.4 kg or less. Table 2.4-3 gives generalized qualification and acceptance levels for components of 22.7 kg or less, with a weight-based reduction above that.
Level (GSFC-STD-7000B) | Shape | Plateau, g²/Hz | Overall stated | Our integration, 20 to 2000 Hz |
|---|---|---|---|---|
Workmanship, Table 2.4-4 | 0.01 at 20 Hz, +3 dB/oct, flat 80 to 500 Hz, -3 dB/oct to 0.01 at 2000 Hz | 0.04 | 6.8 Grms, one minute per axis | 6.78 Grms |
Qualification, Table 2.4-3 | 0.026 at 20 Hz, +6 dB/oct, flat 50 to 800 Hz, -6 dB/oct to 0.026 at 2000 Hz | 0.16 | 14.1 Grms | 14.14 Grms |
Acceptance, Table 2.4-3 | 0.013 at 20 Hz, +6 dB/oct, flat 50 to 800 Hz, -6 dB/oct to 0.013 at 2000 Hz | 0.08 | 10.0 Grms | 10.00 Grms |
Integrating the breakpoints ourselves gives back the stated overall levels, which is a cheap check worth doing on any test specification before it goes into a model. The overall Grms tells you very little about a bracket on its own. What matters is the level at the bracket's own frequencies.

GEVS component levels on log axes with the first natural frequency of each bracket case from the worked check. Case A sits on the rising slope, B on the plateau, C near the qualification roll-off. Case D has a second mode at 441 Hz, inside the plateau.
2. Miles' equation and what it assumes
For a single resonance excited by a broadband base input, the rms acceleration of the mass is approximately the square root of (π/2 × fn × Q × ASD at fn). The equation assumes one dominant mode and an input spectrum that is roughly flat across the resonance, and it needs the amplification Q as an input. Q is usually assumed rather than measured, and a value of 10 is common for bolted metal structures. Response at one point also says nothing about another point on the same part.
3. A worked check on a generic bracket
We modelled a 6061-T6 aluminium cantilever strip, 40 mm wide, carrying a 0.5 kg unit, in scikit-fem with Euler-Bernoulli beam elements (E 68.9 GPa, density 2,700 kg/m³). The base was driven by the GEVS spectra, Q was taken as 10 (5% damping) in every mode, and the response was summed over the first 8 modes from 20 to 2000 Hz. For the three tip-mounted cases the first-mode frequency agrees with the closed-form cantilever with tip mass to the decimal shown, and the case B result did not change between 10 and 120 elements. The peak rms bending stress sits at the root in all four cases, which is where the table quotes it.
Case | Bracket | f1, Hz | Workmanship: unit Grms, Miles / modal | Root stress rms (3σ), MPa | Qualification: unit Grms, modal | Root stress 3σ, MPa |
|---|---|---|---|---|---|---|
A | 100 × 4 mm, unit at tip | 46.8 | 4.15 / 4.14 | 19.3 (58.0) | 9.92 | 139.0 |
B | 100 × 8 mm, unit at tip | 131.0 | 9.07 / 9.22 | 11.0 (33.1) | 18.50 | 66.5 |
C | 50 × 12 mm, unit at tip | 684.3 | 17.72 / 18.34 | 4.8 (14.5) | 41.99 | 33.2 |
D | 150 × 4 mm, unit at mid-span | 65.6 (mode 2: 441.0) | 5.81 / 5.43 | 21.7 (65.0) | 11.95 | 143.7 |
For the three tip-mounted cases Miles is close. The first mode carries 93% or more of the mass, and the single-mode estimate of unit acceleration lands within 3.3% of the modal sum under the workmanship level. Root stress worked out from Miles, by scaling the static 1 g moment, is within 5.1% across both levels.
The stiffening ladder from A to C is the more useful lesson. Moving f1 from 46.8 Hz to 684.3 Hz cuts the root stress from 19.3 to 4.8 MPa rms under the workmanship level, while the unit's own acceleration rises from 4.14 to 18.34 Grms. A stiffer bracket protects itself and hands more of the input to the electronics. Whether that is a good trade depends on what is inside the box, so the bracket and the unit have to be judged together.

Left: response spectra for case D under the workmanship level. The unit at mid-span responds mainly in mode 1; the free end of the same bracket picks up a large second peak at 441 Hz. Right: unit acceleration from Miles' equation against the 8-mode sum for all four cases, with the free end of case D added.
4. Where the single-mode answer misses
Case D moves the unit to mid-span on a longer, thinner strip, a layout that turns up when a bracket is extended to carry a connector or a cable clamp beyond the box. Miles still does reasonably for the unit itself, 5.81 Grms against 5.43 from the modal sum, about 7% on the safe side. The free end of the same bracket, though, reaches 29.42 Grms under workmanship and 60.31 Grms under qualification, because the second mode at 441 Hz sits on the plateau and the end of the strip moves freely in it. A check that only reports the unit acceleration from mode 1 never shows this, yet a connector or clamp fixed at that end sees the full motion.
The Q assumption is the other soft spot. Halving the damping to Q 20 raises case B's unit response from 9.22 to 13.07 Grms and the root stress from 11.0 to 15.7 MPa rms, an increase of about 42%, which is close to the square-root-of-Q scaling Miles predicts. Where a low-level sine sweep on the real part is possible, the measured Q should replace the default.
5. What a random vibration report should show
We ask for the input spectrum with its breakpoints and the integrated Grms, so the reader can confirm it is the right table and the right weight band. Then the modal effective mass for each mode in the band, and the damping actually used with its source. Results should be given at the unit and at any overhang as well as at the root, and stresses should state whether they are rms or 3σ. Where a single-mode estimate is used, the report should say why one mode is enough.
Where Shirsh fits
At Shirsh Technosolutions we run modal and random vibration analyses in ANSYS Mechanical for brackets, enclosures, electronics units and mounted equipment, and we check the results against simple estimates like the one in this note before anything goes to a shaker table. If you have a vibration specification to meet and want to know which part of your assembly will feel it, please get in touch.
References
NASA Goddard Space Flight Center, GSFC-STD-7000B, General Environmental Verification Standard (GEVS), Revision B, 28 April 2021: section 2.4.2.5 (workmanship screen, one minute per axis), Table 2.4-3 (generalized component levels, 22.7 kg or less, 14.1 and 10.0 Grms), Table 2.4-4 (minimum workmanship levels, 45.4 kg or less, 6.8 Grms).
J. W. Miles, "On structural fatigue under random loading", Journal of the Aeronautical Sciences, 1954 (single-degree-of-freedom response to broadband random input).
Worked check computed for this note on 16 November 2026: scikit-fem Euler-Bernoulli beam, generic 6061-T6 strip 40 mm wide with a 0.5 kg unit, Q 10 in all modes, 8 modes, 20 to 2000 Hz. Not any project.
Click to view full size
Related case studies


Want to Apply This to Your Product?
Our engineers can help you implement these techniques on your specific components and assemblies.
Get insights like this in your inbox
Engineering deep-dives, simulation tips, and industry analysis. No spam, unsubscribe any time.
