Using machine learning algorithms to detect wear on selected components of multi-piston pumps

The article presents the use of selected machine learning algorithms for the detection, classification and assessment of the degree of wear of the main components of an axial multi-piston positive displacement pump. The first part of the article describes current work on diagnostic systems for positive displacement pumps. The authors then present their own diagnostic tests of the object (multi-piston pump), equipped with a damage models of its main components. Based on the obtained signal matrices, the statistical characteristics of the signals are calculated. The next step is to use available machine learning classifiers to detect the type of damaged pump component and assess its degree of wear. Then, using Matlab programme, models are developed to classify the wear condition of the pump. Using the adopted criteria, their effectiveness in recognising the modelled pump damage is assessed.