RMS Response, Peaks & Statistical Interpretation

Technical provenance

Applicable standards / specifications

• IEC 60068-2-64 (2008+AMD1:2019) — Environmental testing — Part 2-64: Tests — Test Fh: Vibration, broadband random and guidance
• GSFC-STD-7000B (B (2021)) — General Environmental Verification Standard for GSFC Flight Programs and Projects — Applicable to relevant GSFC flight programmes; not a universal random-vibration specification.

References

• Newland, D. E. — An Introduction to Random Vibrations, Spectral & Wavelet Analysis — Background reference for PSD-based random processes and structural response.
• NASA GSFC-STD-7000B — General Environmental Verification Standard (GEVS) — Space-hardware environmental verification reference; project-specific environments govern.

What Is It?

Random vibration response is a statistical quantity. The RMS gives the overall level. The standard deviation describes the spread. Peak responses are estimated statistically using peak factors or screening levels like 3σ. Understanding what these quantities mean — and what they do not mean — is essential for interpreting random vibration results and for converting them into engineering decisions.

Mean, Variance, Standard Deviation and RMS

For a random signal, the mean is the average value. The variance is the mean-square deviation from the mean. The standard deviation is the square root of variance. The RMS (root mean square) is the square root of the mean-square value. For a zero-mean signal — which is the usual case in vibration — the RMS equals the standard deviation.

Mean:           μ  =  E[x]
Variance:       σ²  =  E[(x − μ)²]
Standard dev:   σ   =  √variance
RMS:            x_rms  =  √( E[x²] )

For zero-mean signal:  RMS  =  σ  =  standard deviation

1σ, 2σ, 3σ for Gaussian Response

For a Gaussian (normal) random process, the probability of the instantaneous value exceeding a given number of standard deviations is well defined. The 1σ level is exceeded approximately 32% of the time. The 2σ level is exceeded approximately 4.6% of the time. The 3σ level is exceeded approximately 0.27% of the time. These probabilities refer to the instantaneous value, not to the peak over a duration.

Three-Sigma Is a Statistical Level, Not a Hard Physical Limit

The 3σ level is widely used as a screening response level in random vibration analysis. It is a convenient and conservative screening value. However, it is not a hard physical maximum. Over a sufficiently long duration, the response will exceed 3σ. The 3σ level is exceeded approximately 0.27% of the time for a Gaussian process — which means for a long-duration test or service life, there will be excursions above 3σ. Treating 3σ as a deterministic maximum is a statistical error.

Three-sigma is a statistical level, not a hard physical limit. Over a long enough duration, the response will exceed 3σ. It is a screening level, not a maximum.

Duration and Number of Cycles Affect Expected Peaks

The expected peak response over a duration depends on how long the excitation lasts and how many cycles occur. A longer duration means more opportunities for large peaks. The expected maximum value of a Gaussian process over N cycles increases with N. For a narrowband process, the expected peak is approximately √(2 ln(N)) · σ, where N is the number of cycles. This means the expected peak grows slowly with duration — but it does grow.

Expected peak (narrowband, Gaussian, N cycles):

E[peak]  ≈  σ · √( 2 · ln(N) )

where:
σ  =  standard deviation (RMS for zero-mean)
N  =  number of cycles in the duration

Note: this is an expected value; actual peaks may be higher or lower.

Peak Factor

The peak factor is the ratio of the expected maximum response to the RMS. It depends on the duration, the frequency content and the bandwidth of the process. For a narrowband Gaussian process, the peak factor grows approximately as √(2 ln(N)). For a broadband process, the peak factor may be different. Peak factors are used to convert RMS response to expected maximum response for design screening.

• Peak factor = expected peak / RMS
• Depends on duration, frequency content and bandwidth
• For narrowband Gaussian: approximately √(2 ln(N)) for N cycles
• Higher peak factors for longer durations (more cycles)
• Peak factors are statistical estimates, not deterministic guarantees

Extreme Response Estimates

Estimating the maximum response over a service life is more complex than applying 3σ. It requires knowledge of the duration, the frequency content, the bandwidth and the amplitude distribution. For qualification, screening levels (3σ or a specified peak factor) are often used with documented assumptions. For detailed assessment, extreme value statistics may be applied. In all cases, the estimate should be accompanied by its statistical basis — what probability of exceedance it corresponds to and over what duration.

STATISTICAL INTERPRETATION: When reporting a peak response from random vibration analysis, state its statistical basis — whether it is a 3σ screening level, an expected peak over a specified duration, or an extreme value estimate with a specified probability of exceedance.

Why 3 × RMS May Be Used but Is Not Universal

The 3 × RMS (3σ) level is widely used as a screening response for several reasons: it is simple, it is conservative for many engineering purposes, and it corresponds to a low instantaneous exceedance probability. However, it is not a universal maximum response. For short-duration events, the actual peak may be well below 3σ. For long-duration service life, the actual peak may exceed 3σ. Using 3σ is a practical engineering convention, not a physical law.

Spatial and Frequency Correlation

In a multi-degree-of-freedom structure under random excitation, the response at different locations is correlated — the same input drives them all. The stress at a particular point depends on the relative motion of surrounding points, which is determined by the modal response. Similarly, the response at different frequencies is correlated through the modal transfer function. These correlations matter for combining responses, for assessing relative displacement and for computing interface loads. They are handled automatically in a random vibration FEA but should be understood conceptually.

Common Mistakes

• Treating 3σ as a deterministic maximum — it is a statistical level that can be exceeded
• Comparing RMS stress directly against a static or fatigue allowable
• Reporting a peak response without stating its statistical basis
• Assuming the peak factor is the same for all durations and bandwidths
• Ignoring the correlation between responses at different locations when assessing relative motion

Key Takeaways

• For zero-mean Gaussian response, RMS equals standard deviation
• 3σ is a statistical screening level — exceeded 0.27% of the time, not a hard maximum
• Expected peak response increases with duration and number of cycles
• Peak factor converts RMS to expected maximum — it depends on duration, bandwidth and frequency content
• Always state the statistical basis when reporting a peak response from random vibration analysis