Spectral Estimation by the Random Dec Technique

This paper contains an empirical study of the accuracy of the Random Dec (RDD) technique. Realizations of the response from a single-degree-of-freedom system loaded by white noise are simulated using an ARMA model. The Autocorrelation function is estimated using the RDD technique and the estimated functions are compared to an analytic solution. The RDD and FFT techniques are compared, and the influence og trig level, damping ratio, type og trig window, and selection og trig points is investigated. The RDD technique appears to be accurate and fast compared to the FFT technique, and the technique is accurate also at small damping ratios and when the estimated functions contain only few points. It is illustrated how bias might be introduced or removed by the choice of trig window and the selection of trig points.