Search academic texts

Information × Registration Number 0123U105198, ( 0224U003063  0224U031423  0224U032198  ) R & D request Title Research on the Prospects of Applying Automated Control with a Learning Neuroregulator in Multifrequency Eddy Current Non-Destructive Testing Head Bazhenov Viktor H., Кандидат технічних наук Registration Date 14-12-2023 Organization National Technscal University of Ukraine "Kiev Polytechnic Institute". popup.description1 The objective of the research is to investigate the use of regression artificial neural networks for determining the hardness of steel samples using an eddy current-based structureoscope with the higher harmonics method. The goals include assessing the feasibility of employing neural networks for processing structureoscope measurement results, conducting an analysis of scientific achievements in the research domain, and specifying the architecture, loss function, and optimizer for the neural networks. Describing the model of the eddy current-based structureoscope, identifying the challenge of using angles as input parameters, and proposing a method to overcome this issue is also part of the research. The study involves exploring fundamental types of neural networks for regression, selecting the feedforward type, justifying the choice of mean squared error as the loss function, and opting for the Adam optimizer. Training and testing of various neural network architectures on real steel samples are conducted. The research concludes with the formulation of recommendations for the further application of the developed method. popup.nrat_date 2024-12-09 Close