Polymer nanocomposites have attracted great interest in recent years because of their potential as tailored materials with enhanced properties. Recent experiments have shown that polymer nanocomposites are able to achieve significant improvement in dielectrical, thermal, mechanical and other physical properties compared with their parent polymer systems. More importantly, these outstanding properties can be achieved at low filler loadings such that the polymer system does not sacrifice the advantages of easy processability. Despite the excellent performance of advanced materials including polymer nanocomposites, the time frame of development to application of those materials in industry is long, typically over 10 years. Therefore, a top priority of researchers and engineers is to reduce the time required to bring advanced materials from laborites to the market. The Material Genome Initiative (MGI), proposed by the White House, is one of the major efforts aimed at addressing this challenge. In order to better understand the behavior of these materials and further design materials with targeted properties, researchers have relied on the materials science paradigm of processing-structure-property (PSP) linkage. Data-driven approaches founded on PSP linkage have much attention recently and have become a hot topic in many areas of materials research. Data-driven approaches combine materials science, computer science and statistics with the goal of expediating material discovery by utilizing past research data to draw new fundamental understanding. To facilitate the development of data archiving, sharing and development of data-driven approaches, efforts have been made to create online databases for fast queries and reference. A data resource, NanoMine, has been developed to accelerate material design for polymer nanocomposites. NanoMine allows fast data queries, visualization, and sharing, as well as a number of tools for analysis including microstructure descriptor identification and reconstruction tools. There are three critical requirements needed to expand the functionality of NanoMine for usage by the broader nanocomoposite community. First, more analysis and module tools that better model the behavior of the material must be built. Second, case studies highlighting the capabilities of NanoMine must be developed using curated data to quantify PSP relationships and elucidate material mechanisms and physics. Finally, it is essential to put continuous efforts toward robust data curation, which expands the size of database and enables development systematic studies. In relation to the three needs above, this dissertation first presents a combined finite element analysis (FEA) and optimization approach to accelerate the identification of interphase properties given experimental data of bulk nanocomposite properties. This approach is tested on both simulations of dielectric and viscoelastic properties in nanocomposites. Our work provides insight into identifying interphase properties for polymer nanocomposites using adaptive optimization and demonstrates the potential of data-driven approaches for achieving a deeper understanding of interphase properties that have proven difficult to directly characterize experimentally. Secondly, we present a novel deep learning approach that probes the structure-property relationships in polymer nanocomposites. Analysis of archived experimental data motivates the development of a computational model that allows for demonstration of this approach and gives flexibility to sufficiently explore a wide range of structures. Lastly, to facilitate the data curation process from literature sources, by applying recent machine learning and natural language processing methods, an end-to-end framework to extract material processing and synthesis information from full-length journal articles is developed. The proposed models and methods are shown herein to be a powerful tool to supplement labor-intensive manual curation and improve the efficiency of data input for the database.

Creator

• Wang, Yixing

DOI

• https://doi.org/10.21985/n2-55bj-jk05

Subject

• Mechanical engineering

Language

• en

Alternate Identifier

• etdadmin_upload_685582
• http://dissertations.umi.com/northwestern:14811

Keyword

• polymer nanocomposite
• simulation
• machine learning
• optimization
• database
• data mining

Date created

• 2019-01-01

Resource type

• Dissertation

Rights statement

• In Copyright