Information Value Measure for Optimization of Flexible Diagnosis Programs of Technical Objects

Abstract

Existing methods of calculating of the value of diagnostic information circulating in the automated systems of monitoring of technical condition of objects do not take into account "losses" ("gains") resulting from making “wrong” decisions when identifying this state. The purpose of the work is to develop an algorithm that allows to solve the problem of recognizing the technical state of the object being analyzed by means of dynamic programming, the value of the diagnostic information as an optimized indicator being used. The solution to the optimization problem of a diagnostic procedure is based on the use of a measure of the information value proposed by R. L. Stratonovich. It is modified according to the subject area of the technical diagnostics and in the case when the diagnostic features presented in the form of intervals on the real numerical axis are used. The maximum value of the diagnostic information is achieved by minimizing the average "losses" (maximizing the average "gains") obtained when performing tests of diagnostic signs in the process of recognizing the technical condition of an object. To solve the problem, a recurrent expression possessing a scientific novelty has been proposed. It allows to calculate the value of the information obtained when performing tests of diagnostic signs in each of the analyzed information states of the diagnostic process. In the process of the diagnostics program implementation when recognizing the technical condition of the object both “losses” and “winnings” are possible. The difference between their a priori and a posteriori means values characterizes the value of the diagnostic information numerically. The magnitude of the information value indication depends on the probabilities of the results of the diagnostic signs checks and is proportional to the difference between the a posteriori and a priori probabilities of achieving the diagnostic goal. By using the proposed solution, it is possible to synthesize the flexible diagnostics program that is optimal according to the maximum value of diagnostic information in the form of a oriented graph or sets of tests in proper sequence of their execution. This is necessary in order to recognize the specific technical state in which the object is located. The implementation of the algorithm developed is possible in the software and algorithmic support of the automated systems for monitoring the state of complex technical objects.