Nonstationary Spectral Analysis for Linear Dynamic Systems.
MASSACHUSETTS INST OF TECH LEXINGTON LINCOLN LAB
Pagination or Media Count:
R REPRESENTATIONS OF SECOND-ORDER STATISTICS OF NONSTATIONARY PROCESSES ARE STUDIED WITH EMPHASIS ON INSTANTANEOUS AND EVOLUTIONARY SPECTRAL DENSITIES. An integrated procedure for the time dependent spectral response calculation for linear systems with digital analysis of input data is presented for excitation processes which are formed from the product of deterministic envelopes and stationary random processes. Linear response is approximated using finite sum Fourier series for the input envelope, the Fourier coefficients being estimated directly from input time records. Bounds on the spectral density of the underlying stationary excitation process are calculated using discrete Fourier transform techniques which take advantage of the fast Fourier transform algorithm. Both the data analysis and linear response relations offer advantages over previous methods in simplicity and computational effort. Examples using pyrotechnic shock time histories and artificial earthquake motions illustrate the procedure. Author
- Statistics and Probability