How Researchers Analyze Vibration Energy Harvesters: Sweep Tests vs. PSD Methods

PSD and spectral analysis appear in a different context: when the excitation is random (broadband white noise, colored noise, or ambient vibrations) rather than a single-frequency sweep. For example, Halvorsen (2014) explicitly presents a "power-spectral-density based approach for estimating the maximum power that can be obtained using a resonant inertial power harvester from a random (aperiodic) vibration source" . That work targets broadband random inputs, not sine sweeps.

Key Distinction: Steady-State vs. Non-Stationary Sweep

When you apply Welch PSD to a frequency sweep, you are treating a non-stationary signal (the frequency changes over time) as if it were statistically stationary within each window. This is technically valid only if the sweep rate is slow enough relative to the FFT window length. Most energy harvesting papers avoid this complication by using the simpler "sweep-to-frequency mapping" method: record the RMS voltage at each time, map time to frequency using the known sweep law, and plot power as

Vrms^2 / R

How to Position Your Method

For a paper or thesis, a clear and honest framing is:

"Frequency-sweep tests were used to characterize the broadband harvesting performance, while PSD-based spectral power analysis was used as a complementary frequency-domain energy metric . Since the sweep excitation was treated as a nonstationary test condition in this analysis, the PSD-based result was interpreted as a spectral power distribution rather than a strict steady-state power-frequency response."

This avoids overclaiming that the PSD method is a mainstream standard for sweep data while still acknowledging its value as an additional lens on the data, especially for highlighting bandwidth, nonlinear effects, or high-frequency content from snap-through or bifurcations .

Practical Recommendation

• Primary method: Use sweep mapping to obtain P(f). This is the gold standard.
• Secondary or auxiliary method: Apply Welch PSD normalization as an extra metric (e.g., input-normalized spectral power distribution) but clearly label it as such.
• Do not claim the PSD method is the standard approach for sweep data. It is not.

References

Bi-stable piezoelectric energy harvester under sweeping frequency and harmonic excitations, NSF-PAR (2026).

PWL harvesters with effective frequency range broadening for up-sweep excitation, Royal Society Open Science (2025).

Halvorsen, E. Limits to inertial vibration power harvesting: power-spectral-density based approach, arXiv:1410.4734 (2014).

Monte Carlo simulations of Duffing nonlinearities under white and colored noise, University of Liverpool (2012).

Low-frequency meandering piezoelectric vibration energy harvester, Purdue University (2012).

Improving performance of vibration energy harvesting from weak excitations by a lever-type mechanism, Mechanical Systems and Signal Processing (2024).

A study of harvested power and energy harvesting efficiency using frequency response analyses of power variables, Mechanical Systems and Signal Processing (2019).