Phase distribution and properties identification of heterogeneous materials: A data-driven approach

This paper presents a new methodology to extend the Data-Driven Identification (DDI) to heterogeneous samples made of multiple elastic materials. By using the Correspondence Analysis (CA) technique to post-process DDI, we are able to identify multiple material databases representative of the material behavior of each phase. Simultaneously, we localize the different phases (matrix and inclusions) in the sample. For different contrasts between phases, the method is tested on synthetically generated data and a parametric study is performed. Furthermore, we show that it is possible to iterate between DDI and CA in order to improve the method's predictions when it is limited by scarce input data. In all the cases studied, the methodology proves to be effective for estimating stresses, as well as for identifying the different phases in the sample.(c) 2021 Elsevier B.V. All rights reserved.

Automated summary

This paper presents a new methodology to extend the Data-Driven Identification (DDI) to heterogeneous samples made of multiple elastic materials. By using the Correspondence Analysis (CA) technique to post-process DDI, the study is able to identify representative material databases for each phase and localize the different phases in the sample. It shows that the method is effective for estimating stresses and identifying different phases in the sample, and it can be improved by iterating between DDI and CA when input data is limited.