Investigation of the Hydraulic Unit Operation Features Based on Vibration System Data Mining

Abstract

The software and hardware level used for on-line monitoring of the hydropower equipment functioning parameters as well as the large amounts of stored data create the necessary conditions for application of innovative methods and technologies of data analysis in the tasks of analytical assessment of equipment condition. This paper presents the results of detecting the operational patterns of the hydraulic unit in various modes and functioning conditions based on the data mining techniques – principal component analysis and cluster analysis – applied to the monitoring data of the vibration control system. In multidimensional data space, two principal components have been selected and interpreted taking into account the contribution of the data attributes to the principal components. On the plane of the first two principal components, a five-cluster structure has been constructed to define the moments of time when the system demonstrates a characteristic behaviour. In addition, the monitored parameters have been analysed in terms of time series at characteristic moments of time. As a result, the comprehensive multidimensional analysis of monitoring data has allowed us to discover the hydraulic unit operation patterns and dependencies, determine the character of the influence induced by its constructive elements and work out the ratio between the ranges of key parameters in various modes of equipment operation.