An Explainable Deep Learning Model Based on Multi-scale Microstructure Information for Establishing Composition–Microstructure–Property Relationship of Aluminum Alloys
{"title":"An Explainable Deep Learning Model Based on Multi-scale Microstructure Information for Establishing Composition–Microstructure–Property Relationship of Aluminum Alloys","authors":"Jiale Ma, Wenchao Zhang, Zhiqiang Han, Qingyan Xu, Haidong Zhao","doi":"10.1007/s40192-024-00374-2","DOIUrl":null,"url":null,"abstract":"<p>Establishing a quantitative composition–microstructure–property relationship is crucial in material design and process optimization. With the advent of big data technology, deep learning models, as a machine learning method that can automatically extract information from images, have been widely used in microstructure image identification and property prediction. However, most deep learning models only use single-scale images for property prediction, ignoring the multi-scale microstructure information of materials. In this study, an explainable deep learning model was developed based on a multi-modal and multi-scale dataset for predicting the tensile properties of aluminum alloys. Three different kinds of aluminum alloys, each incorporating various trace elements, were prepared to evaluate the adaptation of the model. The predicted results demonstrate that the integration of multi-scale microstructure information significantly improves the model’s prediction ability. Furthermore, the intrinsic mechanisms of the deep learning model were elucidated through the application of a visualization technique, greatly improving the explicability of the model. In addition, the effect of data redundancy on model performance was analyzed. The proposed deep learning model breaks the traditional deep learning strategy with the single-scale image as input and effectively establishes the composition–microstructure–property relationship.</p>","PeriodicalId":13604,"journal":{"name":"Integrating Materials and Manufacturing Innovation","volume":"25 1","pages":""},"PeriodicalIF":2.4000,"publicationDate":"2024-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Integrating Materials and Manufacturing Innovation","FirstCategoryId":"88","ListUrlMain":"https://doi.org/10.1007/s40192-024-00374-2","RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, MANUFACTURING","Score":null,"Total":0}
引用次数: 0
Abstract
Establishing a quantitative composition–microstructure–property relationship is crucial in material design and process optimization. With the advent of big data technology, deep learning models, as a machine learning method that can automatically extract information from images, have been widely used in microstructure image identification and property prediction. However, most deep learning models only use single-scale images for property prediction, ignoring the multi-scale microstructure information of materials. In this study, an explainable deep learning model was developed based on a multi-modal and multi-scale dataset for predicting the tensile properties of aluminum alloys. Three different kinds of aluminum alloys, each incorporating various trace elements, were prepared to evaluate the adaptation of the model. The predicted results demonstrate that the integration of multi-scale microstructure information significantly improves the model’s prediction ability. Furthermore, the intrinsic mechanisms of the deep learning model were elucidated through the application of a visualization technique, greatly improving the explicability of the model. In addition, the effect of data redundancy on model performance was analyzed. The proposed deep learning model breaks the traditional deep learning strategy with the single-scale image as input and effectively establishes the composition–microstructure–property relationship.
期刊介绍:
The journal will publish: Research that supports building a model-based definition of materials and processes that is compatible with model-based engineering design processes and multidisciplinary design optimization; Descriptions of novel experimental or computational tools or data analysis techniques, and their application, that are to be used for ICME; Best practices in verification and validation of computational tools, sensitivity analysis, uncertainty quantification, and data management, as well as standards and protocols for software integration and exchange of data; In-depth descriptions of data, databases, and database tools; Detailed case studies on efforts, and their impact, that integrate experiment and computation to solve an enduring engineering problem in materials and manufacturing.