Scholarlink

<p>Metal production requires accurate control of oxide inclusion content which significantly affects mechanical and operational properties of materials. Traditional analysis methods do not always allow for unambiguous determination of composition and distribution of oxides, so that development of new approaches to interpreting fractional gas analysis (FGA) data becomes relevant. In this work, we propose an automated method for determining metal’s oxide composition based on FGA data. The method is based on a&#xa0;mathematical model describing the process of carbon monoxide (CO) evolution from oxide inclusions during non-isothermal melting of a&#xa0;sample. In order to solve the problem of optimizing the model parameters, we used the modified gradient descent method (RMSProp) with automatic differentiation, which allows for effective minimization of the loss function which takes into account both the accuracy of experimental data approximation and the physical constraints of the model. The method has been tested on real data obtained from analysis of samples of grade ShKh-15 steel. The results agreed quite well with the expert assessment of oxide content, which confirms the applicability of the method in industrial conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A method for determining the metal’s oxide composition, based on non-Isothermal gas analysis data

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • D. S. Tokar,
  • K. V. Grigorovich,
  • Y. K. Feldman,
  • A. S. Krylov

摘要

Metal production requires accurate control of oxide inclusion content which significantly affects mechanical and operational properties of materials. Traditional analysis methods do not always allow for unambiguous determination of composition and distribution of oxides, so that development of new approaches to interpreting fractional gas analysis (FGA) data becomes relevant. In this work, we propose an automated method for determining metal’s oxide composition based on FGA data. The method is based on a mathematical model describing the process of carbon monoxide (CO) evolution from oxide inclusions during non-isothermal melting of a sample. In order to solve the problem of optimizing the model parameters, we used the modified gradient descent method (RMSProp) with automatic differentiation, which allows for effective minimization of the loss function which takes into account both the accuracy of experimental data approximation and the physical constraints of the model. The method has been tested on real data obtained from analysis of samples of grade ShKh-15 steel. The results agreed quite well with the expert assessment of oxide content, which confirms the applicability of the method in industrial conditions.