INFORMATIONAL ENTROPY IN SIMULATION OF ONE-DIMENSIONAL RANDOM FIELDS

Abstract

The entropy H of a continuous distribution with probability density function f(•) is defined as a function of the number of nodes (n) in a one-dimensional scalar random field. For the second order theory this entropy is expressed by the determinants of the covariance matrices and simulated for several types of correlation functions. In the numerical example the propagation of the entropy for the static response of linear elastic, randomly loaded beam has been considered. Two unexpected results have been observed:
• function H(n) is entirely different for differentiable (m. s.) and non-differentiable fields, with the same parameters in the correlation functions,
• in some cases, the greater randomness at the input (measured by the entropy) does not lead to the greater randomness at the output.