Evaluating the Reliability of Groove Turning for Piston Rings in Combustion Engines with the Use of Neural Networks

Abstract

Abstract The article describes a method of evaluating the reliability of groove turning for piston rings in combustion engines. Parameters representing the roughness of a machined surface, Ra and Rz, were selected for use in evaluation. At present, evaluation of surface roughness is performed manually by operators and recorded on measurement sheets. The authors studied a method for evaluation of the surface roughness parameters Ra and Rz using multi-layered perceptron with error back-propagation (MLP) and Kohonen neural networks. Many neural network models were developed, and the best of them were chosen on the basis of the effectiveness of measurement evaluation. Experiments were carried out on real data from a production company, obtained from several machine tools. In this way it becomes possible to assess machines in terms of the reliability evaluation of turning.