Peak Factor in Random Vibration

What Is It?

The peak factor is the ratio of the expected maximum response to the RMS response over a given duration. It converts the statistical RMS level into an estimate of the peak response. The peak factor depends on the duration (how many cycles occur), the bandwidth of the process (narrow-band or broad-band) and the amplitude distribution. It is the standard method for estimating peaks from RMS values in random vibration.

Why It Matters

The peak factor provides a more rigorous basis for peak estimation than simply using 3σ. Instead of a fixed screening level, the peak factor accounts for the actual duration and the bandwidth of the response. This produces a more accurate and more defensible peak estimate — one that is tied to the specific conditions rather than a generic assumption. Understanding peak factors is essential for moving beyond 3σ screening to duration-specific peak estimation.

The peak factor provides a duration-specific peak estimate rather than a fixed 3σ screening level. It accounts for the actual number of cycles and the bandwidth of the response.

Peak Factor Definition

The peak factor is defined as the expected maximum value divided by the RMS. For a zero-mean process, the RMS equals the standard deviation σ. The peak factor is therefore the expected maximum in units of σ.

Peak factor:

k_p  =  E[ x_max ] / x_rms  =  E[ x_max ] / σ

where:
k_p    =  peak factor
x_max  =  maximum value over the duration
x_rms  =  RMS  =  σ (for zero-mean)

E[ x_max ]  =  k_p · σ  =  k_p · x_rms

Bandwidth

The bandwidth of the response process affects the peak factor. A narrow-band process (dominated by a single resonance) has a well-defined cycle count and a predictable peak factor. A broad-band process (multiple modes, wide frequency content) has more irregular peaks and a different peak factor. The distinction matters because narrow-band and broad-band processes have different extreme value statistics.

Duration

The peak factor grows with duration because more cycles mean more opportunities for large peaks. For a narrow-band Gaussian process, the peak factor grows as √(2·ln(N)), where N is the number of cycles. This growth is slow — the peak factor for 100 cycles is about 3.0, for 10,000 cycles about 3.7, for 1 million cycles about 4.9. But the growth is real and means that no fixed peak factor is a true maximum.

Narrow-band Gaussian peak factor:

k_p  ≈  √( 2 · ln(N) )

where N = number of cycles = f_dominant × T

Examples:
  N = 100:       k_p ≈ 3.0
  N = 1,000:     k_p ≈ 3.3
  N = 10,000:    k_p ≈ 3.7
  N = 100,000:   k_p ≈ 4.1
  N = 1,000,000: k_p ≈ 4.4

Note: growth is slow but continuous.
No fixed peak factor is a true maximum.

Statistical Distribution

The peak factor is not a deterministic value — it is the expected (mean) value of the maximum over a duration. The actual maximum has a distribution around this expected value. For engineering purposes, a conservative peak factor may be used — for example, the expected value plus one standard deviation of the extreme value distribution. This provides a higher confidence that the actual peak will not exceed the estimate.

• Peak factor is the EXPECTED maximum, not a guaranteed maximum
• The actual maximum has a distribution around the expected value
• For higher confidence, use expected + 1 standard deviation
• The distribution depends on the bandwidth and amplitude distribution

Engineering Use

The peak factor is used to convert RMS response to expected peak response for design and assessment. It is more rigorous than 3σ screening because it accounts for the actual duration. For short durations, the peak factor may be close to 3σ. For long durations, it may be significantly higher than 3σ. Using the duration-specific peak factor produces a more defensible peak estimate.

Engineering use:

x_peak  =  k_p · x_rms

where k_p depends on duration and bandwidth

For narrow-band Gaussian:
  x_peak  ≈  √( 2 · ln(N) ) · x_rms

Example:
  x_rms = 5 g,  f = 200 Hz,  T = 60 s
  N = 200 × 60 = 12,000 cycles
  k_p = √(2 · ln(12000)) = √(2 · 9.39) = √18.78 = 4.33
  x_peak ≈ 4.33 × 5 = 21.7 g

Compare with 3σ = 15 g — the duration-specific
peak is significantly higher for this duration.

For long durations, the duration-specific peak factor can be significantly higher than 3σ. Using 3σ for a 1-hour assessment may be non-conservative. Use the peak factor for the actual duration.

Comparison with 3σ Screening

The 3σ screening level corresponds to a peak factor of 3, which is the expected peak for about 100 cycles. For shorter durations (fewer cycles), 3σ is conservative. For longer durations (more cycles), the duration-specific peak factor exceeds 3 and 3σ is non-conservative. The table below shows the relationship.

Key Takeaways

• Peak factor = expected maximum / RMS — converts RMS to duration-specific peak estimate
• For narrow-band Gaussian: k_p ≈ √(2·ln(N)) — grows slowly with number of cycles
• 3σ corresponds to about 100 cycles — conservative for fewer, non-conservative for more
• Peak factor accounts for both duration and bandwidth — more rigorous than 3σ
• For long durations, the peak factor can be significantly higher than 3σ