Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Improving the processes of training in the diagnostic models of complex technical systems V.I. SENCHENKOV, E.V. SHISHKIN Evolve the training methods in the development of diagnostic models of complex technical systems with limitation and non-uniformity of a priori information, in particular, reveal new aspects construction of images – of the mathematical description of types of technical condition of the system. Formed saturation conditions of the training process, introduced the concept of the rate of convergence of the training in the construction of images, offered quantitative characteristic change of the rate of convergence. On this basis, an improved version of the procedure of the grouping of training images and ranking groups is received as compared with those offered in previous works. This procedure allows to significantly speed up the convergence of the training process. Development of procedure is based on the application in the computing circuit, implementing the algorithm Robbins-Monroe, of orthogonal trigonometric basis in the space of continuous functions, square integrable on Riemann, as well as on the use of topology and properties of finite-dimensional Euclidean spaces. Examples are given of the training process without the application of the procedure of grouping and ranking, as well as with its application. The comparison is performed the two presented methods by the convergence rate. Keywords: diagnostic model, technical condition, training process, procedure of grouping and ranking. Contacts: E-mail: svi9@rambler.ru Pp. 33-44.