Prediction of Piston Ring Wear in Marine Main Engine Using Neural Networks

The article shows the results of using a neural network to approximate the data of experimental studies of the wear of the ceramic-metal coating of the piston rings of ship engines. A structural diagram is proposed that qualitatively describes the ambiguous nature of the relationship between wear and a set of operational parameters. The coating of piston rings of operating ship engines 6S50ME-C-GI with electronic control was measured. The condition of the piston rings of the marine main engine determines a number of parameters that determine the operating conditions and reliability of the engine. Factor analysis of the effect of structural and technological parameters on the wear of piston rings was carried out. The specific consumption of cylinder oil and engine power are selected as the main factors affecting wear. A neural network was used to process the accumulated data. Taking into account the size of the training data, the network structure is shown, which describes the functional relationship between fuel consumption and engine power with wear, data arrays for its operation are determined and the network is trained. As a result of the operation of the neural network, an approximating surface of wear of rings and a combination of significant parameters is obtained, at which wear is maximum. The obtained results can be applied in practice to develop an algorithm that predicts the calendar date of overhaul of the cylinder, taking into account wear factors.