Interval approach to identification of the describing function of experimental noised process
S.I. Kumkov · AIP conference proceedings · 2019
In practice, experimentalist obtains information on the investigated process under conditions of uncertainty of probabilistic characteristics of errors in measurements, the measurement sample is short, errors are supposed to be bounded (on modulus) by some approximate magnitude. So, the experimentalist is hampered in the choice of appropriate type of function for describing the process. Under such conditions, formal application of standard statistical methods (for example, the least squares mean) can give, as a rule, only point-wise estimates of the describing function. In contrast, approach on the basis of Interval Analysis allows one to estimate reliably the type of the process describing function on the basis of its information set of parameters.