Diagnostic Data Fusion Collected from Railway Automatics and Telemechanics Devices on the Basis of Soft Computing Technologies

Abstract

When diagnosing devices of railway automatics and telemechanics (RAT), a large amount of semi-structured data collected from a variety of different sensors is used in technical diagnostics and monitoring systems. In order to improve efficiency and develop scientifically based management actions aimed at detecting faults in RAT devices under conditions of uncertainty, it is necessary to implement a process of continuous data and knowledge fusion into technical diagnostics and monitoring systems. This paper considers a generalized scheme for heterogeneous data fusion showing the features of this process. The data fusion scheme is universal as it is easy to use for any RA device diagnostics, as well as to adapt to different structures of data representation. In addition, it provides correct operation under heterogeneous data, protects against incorrect data, and makes it possible to easily increase the number of methods for data fusion and input parameters. An approach based on the JDL data fusion model and soft computing technologies has been proposed. The model is a composition of consecutive interconnected stages of the data fusion process and their functions. The proposed approach makes it possible to increase the efficiency of making diagnostic decisions under the conditions of heterogeneous data collected from a variety of different types of sensors. The paper is structured as follows. Part 2 deals with the current problems of the diagnostic data fusion obtained in real time from many different sensors and ways to solve them. In part 3 the authors discuss the diagnostic data representation structure and provide a detailed description of a generalized scheme for merging the diagnostic heterogeneous data. This scheme is the basis of the proposed approach which uses the JDL data fusion model and soft computing technology. In part 4 a model for troubleshooting in RAT devices with an example illustrating the use of the developed model is proposed. This model provides searching for faults in RAT devices, possible problems, as well as predicting all kinds of situations related to device malfunctions.