U. L. Mishigdorzhiyn, B. A. Dyshenov, A. P. Semenov, N. S. Ulakhanov, B. E. Markhadayev
{"title":"Prediction of the Thickness of a Boroaluminized Layer Using an Artificial Neural Network","authors":"U. L. Mishigdorzhiyn, B. A. Dyshenov, A. P. Semenov, N. S. Ulakhanov, B. E. Markhadayev","doi":"10.1134/S1027451024020344","DOIUrl":null,"url":null,"abstract":"<p>The application of mathematical models and artificial neural networks for predicting the properties of diffusion coatings created by thermal–chemical treatment based on the boroaluminizing process is considered. The formalization and analysis of forecasting experimental results are conducted. Building computer models for prediction based on experimental data of the boroaluminizing process with high accuracy is a solvable task when using artificial neural networks such as a multilayer perceptron. Testing the number of hidden layers and the number of neurons in them revealed the highest correlation coefficient <i>R</i> = 0.99993 for an artificial neural network using two hidden layers with ten and six neurons, respectively. The highest efficiency can be achieved using the hyperbolic tangent activation function.</p>","PeriodicalId":671,"journal":{"name":"Journal of Surface Investigation: X-ray, Synchrotron and Neutron Techniques","volume":"18 2","pages":"466 - 473"},"PeriodicalIF":0.5000,"publicationDate":"2024-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Surface Investigation: X-ray, Synchrotron and Neutron Techniques","FirstCategoryId":"1085","ListUrlMain":"https://link.springer.com/article/10.1134/S1027451024020344","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"PHYSICS, CONDENSED MATTER","Score":null,"Total":0}
引用次数: 0
Abstract
The application of mathematical models and artificial neural networks for predicting the properties of diffusion coatings created by thermal–chemical treatment based on the boroaluminizing process is considered. The formalization and analysis of forecasting experimental results are conducted. Building computer models for prediction based on experimental data of the boroaluminizing process with high accuracy is a solvable task when using artificial neural networks such as a multilayer perceptron. Testing the number of hidden layers and the number of neurons in them revealed the highest correlation coefficient R = 0.99993 for an artificial neural network using two hidden layers with ten and six neurons, respectively. The highest efficiency can be achieved using the hyperbolic tangent activation function.
期刊介绍：
Journal of Surface Investigation: X-ray, Synchrotron and Neutron Techniques publishes original articles on the topical problems of solid-state physics, materials science, experimental techniques, condensed media, nanostructures, surfaces of thin films, and phase boundaries: geometric and energetical structures of surfaces, the methods of computer simulations; physical and chemical properties and their changes upon radiation and other treatments; the methods of studies of films and surface layers of crystals (XRD, XPS, synchrotron radiation, neutron and electron diffraction, electron microscopic, scanning tunneling microscopic, atomic force microscopic studies, and other methods that provide data on the surfaces and thin films). Articles related to the methods and technics of structure studies are the focus of the journal. The journal accepts manuscripts of regular articles and reviews in English or Russian language from authors of all countries. All manuscripts are peer-reviewed.
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号
