AN FFT SPECTRAL ANALYSIS FOR SYSTEM IDENTIFICATION

Abstract

With the advent of, larger am more complex systems being used in industry the requirement for more accurate estimates of the system operating parameters becomes greater. One method of obtaining these operating parameters is by the use of the information contained in the spectral density functions of a test signal applied to the system. This thesis reviews the basic relationships between a system transfer function am the spectral density functions of the system input signal am output signal. The method of obtaining the spectral density functions by means of the fast Fourier transformation algorithm is discussed in terms of test procedure design, and accuracy considerations. The use· or the coherence function to obtain additional information as to the accuracy or the transfer function estimate is discussed and demonstrated. This method of system identification was implemented on a medium sized digital computer using fixed -- point arithmetic. The experimental results of the tests conducted on several first and second order systems demonstrate the ability of this method to provide useful transfer function estimates.