Modelling of mechanical characteristics of deformable flat specimens under creep by machine learning methods

Авторы: Shevchenko E.C., Teraud W.V.
Журнал: Journal of Mechanical Engineering and Sciences
Год издания: 2021
DOI: 10.15282/jmes.14.1.2020.15.0500
Аннотация: The paper considers the mechanical process of stretching flat aluminum specimens under high-temperature (400o C) creep. The obtained mechanical characteristics were modeled by gradient boosting and neuron networks from the photos of deformation process. The photos contained 24 series corresponding to different experiments and totaling 1500 photos. The length of the specimen, its deformation and the time of strain localisation were modeled. Gradient boosting modelling showed that the algorithm lost generalization ability already on a reduced data set, therefore it was not considered any further. Neural network analysis of longitudinal strain produced a general error of ~10 %. To further reduce the error, linear regression was constructed on the predictions of three neural networks with the best sets of hyperparameters. That allowed reducing the modelling error to 3.6% for the entire sample. Localisation times were modeled by an artificial neural network for dimensional and dimensionless values. There was used a neural network with four outputs and four separate neural networks with one output, which showed a better result of 1.5% for the entire data sample.
Добавил в систему: Терауд Валентин Викторович

№	Имя	Описание	Имя файла	Размер	Добавлен
1.	Полный текст		JMES1906455_Revised_V2.5.pdf	604,8 КБ	29 ноября 2021 [ldrnww]
2.	Подтверждение принятия статьи к публикации		JMES_Editor_Decision.docx	14,9 КБ	29 ноября 2021 [Lokoshchenko]

	ИСТИНА	Войти в систему Регистрация
	Интеллектуальная Система Тематического Исследования НАукометрических данных
	Главная Поиск Статистика О проекте Помощь

ИСТИНА