A power spectrum characterizes frequency content and resonances within a system. Because degradation usually causes changes in the spectral signature, spectral behavior provides a rich source of information for feature generation.
Select from parametric and nonparametric algorithms. For the parametric methods, Diagnostic Feature Designer fits a parametric model to the signal. The software then uses this model to compute the spectral density. You can choose an auto-regressive (AR) model or a state-space model.
Signal — Source signal for the power spectrum.
Frequency — Frequency settings for the frequency axis. To set these values manually, clear Select automatically and update the parameters for the frequency vector generation.
Auto-regressive model — The app fits an AR model to the
signal and uses this model to compute the spectral density. For information on setting
model order, approach, and windowing method, see ar.
State-space model — The app fits a state-space model to the signal and uses this model to compute the spectral density.
Model Order — Specify the model order directly, or specify a range of orders for automatic order selection. With automatic order selection, the software automatically selects the smallest order that leads to a good fit to the data.
Improve results using nonlinear least squares
search — Selecting this option improves estimation results for
specific scenarios, at the cost of additional computational time. For more
information, see the 'SearchMethod' option in ssestOptions.
Maximum number of iterations — Increase the number of iterations to improve result accuracy. Decrease the number to improve computational speed.
For more information on state-space modeling, see ssest.
Welch's method — The app calculates the power spectrum
from the source signal using Welch's method. For information on setting window
parameters, see pwelch.
The software stores the results of the computation in a new variable. The new variable
name includes the source signal name with the suffix ps.