Explicit expressions of the saturation flux density and thermal stability in Fe-based metallic glasses based on Lasso regression. (December 2021)

• Record Type: Journal Article
• Authors: Li, Zhuang; Long, Zhilin; Lei, Shan; Yang, Lingming; Zhang, Wei; Zhang, Ting
• Is Part Of: Intermetallics. Volume 139(2021)
• Publication Date: 2021-12
• Subjects: Metallic glass -- Machine learning -- Lasso regression -- Soft magnetic property -- Thermal stability; Intermetallic compounds -- Metallography -- Periodicals; Metallic glasses -- Periodicals; Composés intermétalliques -- Métallographie -- Périodiques; 669.94
• DOI: 10.1016/j.intermet.2021.107361
• Languages: English
• ISSNs: 0966-9795

Abstract

Abstract: Fe-based soft magnetic metallic glasses (MGs) have great potential for a wide variety of applications due to their low material cost and excellent soft magnetic properties. At present, a satisfactory combination of the saturation flux density ( B s ) and thermal stability ( T x ) in the above-mentioned MGs is of great challenge if the conventional trial-and-error method is adopted. In order to explore explicit expressions for predicting B s and T x of Fe-based soft magnetic MGs instead of implicit and unexplained machine learning (ML) models, herein, Lasso regression approach was employed and the predictive performance of the developed explicit expressions based on different input descriptors were analyzed. The obtained results show that there exists a highly linear relationship among composition, structure and property in Fe-based MGs. The studied explicit expressions of B s and T x exhibit good prediction efficiency with the high R 2 scores of 0.952 and 0.968, respectively, which are both superior to the previously reported corresponding R 2 values in the literature. This work suggests that Lasso regression possesses a great potential for assessing the quantitative composition-structure-property relationship, and thus might provide some hints to discover new Fe-based soft magnetic MGs. Highlights: There exist highly linear relationships among composition, structure and property in Fe-based MGs. Explicit expressions of B s and T x were developed based on Lasso regression. The predictive performance of developed explicit expressions based on different input descriptors were investigated. The influence of constituent elements and atomic scale physical parameters on B s and T x were analyzed. The formulas of B s and T x exhibit higher accuracy than previous work, with high R 2 of 0.952 and 0.968, respectively.