Slowly changing low-dynamic character of signals is a prominent feature of technical objects working in modes of normal operation. The traditional algorithms based on recurrent computing procedures lose serviceability because of the occurrence of singular situations. Results of identification become unreliable because of unpredictable surges of errors. Therefore a new class of more noise-tolerant algorithms has been developed on the basis of a new mathematical approach of their symbolical imitation modelling. It has allowed obtaining a number of new theoretical results in the field of identification and creating efficient software for systems of automatic control and diagnosis of technical objects.