Diagnostic Signal Informativeness Increase

Abstract

Modern requirements for the operation of cars lead to the need to improve the efficiency of their maintenance. Diagnosis as an element of the maintenance process allows you to provide information about the technical condition of a particular element, which allows you to respond in a timely manner to the technical condition change of a diagnostic object with minimal resources. In this paper, we consider the way improving the diagnostic signal quality. It is known that a diagnostic signal must meet several requirements. The most important requirement is informativeness, which shows the decrease of uncertainty about the technical condition of an object, represented by a priori entropy after information application from this diagnostic signal, measured during the diagnosis. There are the methods for a diagnostic signal conversion, which allow to get rid of the noise entering it to a different degree or present it in such a way to facilitate the signal analysis process. Three methods are considered in the work: direct spectrum obtaining, signal envelope spectrum obtaining, and adaptive filtering. The analysis of these methods led to the conclusion that adaptive filtering has the greatest efficiency potential. We have proposed the method that is based on adaptive filtering, but with additional operations. In the course of the diagnostic signal studies and the adaptive filtering algorithm, we found that it is possible to set the function to be detected as a variable, as well as several parameters that affect the result quality. Based on this, a new method for a useful signal extraction was proposed. The results of the work were checked on a signal simulating a car gearbox signal. The results show that the method allows you to obtain the necessary knowledge about a defect, which can be used in the diagnosis. The developed method allows to increase the information content of the diagnostic signal by suppressing its other components. The results of the proposed method correlate with the results of other methods for general cases, i.e., when the ratio of the useful signal to noise is such that high sensitivity of the method is not required to identify the useful signal.