HAMPEL FILTERING OF POWER SPECTRAL DENSITIES TO REMOVE SPURIOUS SINE TONES IN RANDOM VIBRATION DATA

Collecting accurate data during vibration testing requires clean data.One of the first steps in a vibration test is the characterization of the noise floor of the measurement chain.This step is important so that artifacts of the measurement chain do not pollute the measured data, particularly at low signal to noise ratios.During a recent transportation vibration test conducted by Sandia National Laboratories, the data acquisition system introduced narrow band spurious signals that were large enough to be observable in the ASDs of the accelerations collected during the test.The spurious noise tones had to be removed from the ASDs collected during the driving phase of the test.A novel approach based on frequency domain median filtering of the ASDs was used to clean the measured data.Three methods were developed, and all were able to remove the sine tones.The results show that frequency domain median filtering is a viable, straightforward way to remove certain types of measurement noise artifacts without biasing the data.