One of the important factors to be considered in designing an engineering system is uncertainty, which emanates from natural randomness, limited data, or limited knowledge of systems. In this study, a robust design methodology is established in order to design multifunctional materials, employing multi-time and length scale analyses. The Robust Concept Exploration Method with Error Margin Index (RCEM-EMI) is proposed for design incorporating non-deterministic system behavior. The Inductive Design Exploration Method (IDEM) is proposed to facilitate distributed, robust decision-making under propagated uncertainty in a series of multiscale analyses or simulations. These methods are verified in the context of Design of Multifunctional Energetic Structural Materials (MESM). The MESM is being developed to replace the large amount of steel reinforcement in a missile penetrator for light weight, high energy release, and sound structural integrity. In this example, the methods facilitate following state-of-the-art design capabilities, robust MESM design under (a) random microstructure changes and (b) propagated uncertainty in a multiscale analysis chain. The methods are designed to facilitate effective and efficient materials design; however, they are generalized to be applicable to any complex engineering systems design that incorporates computationally intensive simulations or expensive experiments, non-deterministic models, accumulated uncertainty in multidisciplinary analyses, and distributed, collaborative decision-making.

This thesis focuses on the micromechanical modeling of particulate viscoelastic composite materials in the quasi-static frequency domain to approximate macroscopic damping behavior and has two main objectives. The first objective is the development of a robust frequency dependent multiscale model. For this purpose, the self-consistent (SC) mean-field micromechanical model introduced by Cherkaoui et al [J. Eng. Mater. Technol. 116, 274-278 (1994)] is extended to include frequency dependence via the viscoelastic correspondence principal. The quasi-static model is then generalized using dilute strain concentration tensor formulation and validated by comparison with complex bounds from literature, acoustic and static experimental data, and established models. The second objective is SC model implementation as a tool for the design of high loss materials. This objective is met by integrating the SC model into a Compromise Decision Support Protocol (CDSP) to explore the microstructural design space of an automobile windshield. The integrated SC-CDSP design space exploration results definitively indicate that one microstructural variable dominates structure level acoustic isolation and rigidity: negative stiffness. The work concludes with a detailed description of the fundamental mechanisms leading to negative stiffness behavior and proposes two negative stiffness inclusion designs.

The play materials designed for special education are commonly used by children with various disabilities. This study overviews these materials from the perspectives of special needs, special user groups and special education. Furthermore, the effects of different disabilities on the interactions of the children with the same toys are investigated, within the light of those special cases. The base of the study is constructed with a literature review on toys, play and special education, followed by observations that are performed in a special education centre with six subjects having different disabilities, namely Rett&rsquo / s syndrome, mental retardation, hyperactivity, Down&rsquo / s syndrome, cerebral palsy and autism. The study is concluded with compared results of the observations and design implications derived from the interactions between the toys and children.

In the design of novel materials, tunable properties are parameters such as composition or structure that may be adjusted in order to enhance a desired chemical or material property. Trends in tunable properties can be accurately predicted using computational and combinatorial chemistry tools in order to optimize a desired property. I present a study of tunable properties in materials and employ a variety of algorithms that ranges from simple screening to machine learning. In the case of tuning a nanocomposite membrane for olefin/paraffin separations, I demonstrate a rational design approach based on statistical modeling followed by ab initio modeling of the interaction of olefins with various nanoparticles. My simplified model of gases diffusing on a heterogeneous lattice identifies the conditions necessary for optimal selectivity of olefins over paraffins. The ab initio modeling is then applied to identify realistic nanomaterials that will produce such conditions. The second case, [alpha]-Fe₂O₃, commonly known as hematite, is potential solar cell material. I demonstrate the use of a screened search through chemical compound space in order to identify doped hematite-based materials with an ideal band gap for maximum solar absorption. The electronic structure of hematite is poorly treated by standard density functional theory and requires the application of Hartree-Fock exchange in order to reproduce the experimental band gap. Using this approach, several potential solar cell materials are identified based on the behavior of the dopants within the overall hematite structure. The final aspect of this work is a new method for identifying low-energy chemical processes in condensed phase materials. The gap between timescales that are attainable with standard molecular dynamics and the processes that evolve on a human timescale presents a challenge for modeling the behavior of materials. This problem is particularly severe in the case of condensed phase systems where the reaction mechanisms may be highly complicated or completely unknown. I demonstrate the use of support vector machines, a machine-learning technique, to create transition state theory dividing surfaces without a priori information about the reaction coordinate. This method can be applied to modeling the stability of novel materials. / text

No description available.

Ultra-High-Performance Concretes (UHPCs) are a promising class of cementitious materials possessing mechanical properties superior to those of Normal Strength Concretes (NSCs). However, UHPCs have been slow to transition from laboratory testing to insertion in new applications, partly due to an intuitive trial-and-error materials development process. This research seeks to addresses this problem by implementing a materials design process for the design of UHPC materials and structures subject to blast loads with specific impulses between 1.25- and 1.5-MPa-ms and impact loads resulting from the impact of a 0.50-caliber bullet travelling between 900 and 1,000 m/s. The implemented materials design process consists of simultaneous bottom-up deductive mappings and top-down inductive decision paths through a set of process-structure-property-performance (PSPP) relations identified for this purpose. The bottom-up deductive mappings are constructed from a combination of analytical models adopted from the literature and two hierarchical multiscale models developed to simulate the blast performance of a 1,626-mm tall by 864-mm wide UHPC panel and the impact performance of a 305-mm tall by 305-mm wide UHPC panel. Both multiscale models employ models at three length scales – single fiber, multiple fiber, and structural – to quantify deductive relations in terms of fiber pitch (6-36 mm/revolution), fiber volume fraction (0-2%), uniaxial tensile strength of matrix (5-12 MPa), quasi-static tensile strength of fiber-reinforced matrix (10-20 MPa), and dissipated energy density (20-100 kJ/m²). The inductive decision path is formulated within the Inductive Design Exploration Method (IDEM), which determines robust combinations of properties, structures, and processing steps that satisfy the performance requirements. Subsequently, the preferred material and structural designs are determined by rank order of results of objective functions, defined in terms of mass and costs of the UHPC panel.

In this study, the templating role of bimolecular triple hydrogen bonded honeycomb network consisting perylene-3,4,9,10-tetracarboxydi-imide and melamine is investigated, using scanning tunneling microscopy. Although the stability of the network upon modification is a major obstacle toward higher complexity, three different approaches in this work highlight formation of successful architectures in a sequential way. 1. Insertion of pore modifier star shaped molecules based on tri(phenylene ethynylene)benzene core in the pores to construct a new template. 2. Insertion of iodinated molecules in the pores to study the network as a nanoreactor. 3. Electrochemical deposition of metals in the pores. Self-assembly monolayer of four different molecules based on tri(phenylene ethynylene)benzene core on uniform gold surface revealed different structures. The degree of the order within the structures depends highly on the symmetry of the molecules, and hence asymmetric molecule formed disordered structure. Upon insertion into the pores of the network, one of molecules did not match the pores size, while others fitted and illustrated rotation depending on the strength of their interaction with the network components and the substrate. The rotation is significantly reduced by modifying the molecules. These new architectures are used as templets hosting C₆₀ molecules which resulted in isolated single C₆₀ molecules. Self-assembly of iodinated molecule under different conditions on uniform gold surface leads to formation of different structures including monomers and dimers. Upon thermal treatment on the uniform surface oligomers are formed, whereas for the molecules confined in the pores of the network, the covalent bond formation was limited to dimerisation. Electrochemical copper deposition into the pores of the network under acidic condition (pH = 1 - 2) is not possible because of the stability of the network. However, by increasing pH of the electrolyte (pH = 5 – 7), a bilayer of Cu and anion is formed in the pores of the network, confirmed by scanning tunneling microscopy and X-ray photoelectron spectroscopy.

No description available.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第24235号 / 工博第5063号 / 新制||工||1790(附属図書館) / 京都大学大学院工学研究科物質エネルギー化学専攻 / (主査)教授 陰山 洋, 教授 藤田 晃司, 教授 作花 哲夫 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

<p>The design and discovery of novel materials are difficult not only due to expensive and time- consuming calculation and measurements of their properties, but also thanks to the infinite search spaces. With the increasingly abundant data from experiments and simulations, learning from data has the potential of bypassing complex physics-based simulations and experiments and providing fast approximations of the solution. Deep learning models are helpful in the design process that requires prohibitively expensive iterative computations. In addition, as efficient and accurate sur- rogate models, trained deep networks can incorporate techniques, such as sensitivity analysis and active learning, to provide guidance in searching promising candidates. Moreover, deep learning models need to account for the material structural information, such as molecule and atom align- ments, chemical bonds, and grain-level interactions, as it plays an important role in determining the macroscopic properties. In this thesis, we start with developing two standard deep learning model- based materials design frameworks for lithium-ion batteries and thermoelectric materials, and we then investigate the feasibility of standard deep learning models on data with graph-structured in- formation and identify the challenges. Finally, we propose a deep graph operator network that effectively capture the spatial dependency encoded in the graph structure to solve networked dy- namical systems.</p> <p><br></p> <p>In the first half of the thesis, we propose a hybrid convolutional neural network to infer lithium- ion battery microstructure properties, Bruggeman’s exponent and shape factor, given its voltage vs. capacity curves. The trained model accurately predicts the microstructural properties on both experimental and simulation data, and it can readily accelerate the processing-properties- performance and degradation characteristics of the existing and emerging chemistries of lithium- ion batteries. Also, we develop a AI-guided framework to discover and design thermoelectric materials, where we train classifiers based on the materials chemical and structural information embeddings and combine with variance-based sensitivity analysis to suggest candidates and con- duct fast screening.</p> <p><br></p> <p>In the second half of the thesis, we build a data-centric framework with a recurrent neural network-based classifier to achieve traffic incident detection on highway networks. We incorporate weak supervised learning and design labeling functions to create large amount of training data with probabilistic labels. The trained deep ensemble accurately detects incidents with predictive uncertainty. To capture the structural information in the network, we then propose a deep graph operator network that maps the input graph state function to the output graph state function. The proposed model enables resolution-independence and zero-shot transfer, where we do not require a set of fixed sensors to encode the graph trajectory and can use the trained model directly on larger graphs with high accuracy. We utilize the proposed model to solve power grid transient stability prediction and traffic forecasting problems.</p>