Mean Square Measurements of Nonstationary Random Processes

Three techniques for estimating mean square values of nonstationary random processes are analyzed and compared. These include ensemble averaging, orthogonal function approximation, and short time averaging. It is shown that ensemble averaging is useful only when the number of records available is large because of the estimation errors. The orthogonal function approximation technique is shown to be better than ensemble averaging, although more difficult to mechanize. It is also shown that short time averaging generally produces biased estimates. Finally, a brief discussion is presented on the selection of the best technique to implement for particular applications.