<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>

<div>Accelerating the development of novel materials is one of the central goals of the Materials Genome Initiative and improving the predictive power of computational</div><div>material science methods is critical to attain this goal. Molecular dynamics (MD) is one such computational technique that has been used to study a wide range of materials since its invention in the 1950s. In this work we explore some examples of using and increasing the predictive power of MD simulations to understand materials phenomena and provide guidelines to design tailored materials. We first demonstrate the use of MD simulations as a tool to explore the design space of shape memory alloys, using simple interatomic models to identify characteristics of an integrated coherent second phase that will modify the transformation characteristics of the base shape memory alloy to our desire. Our approach provides guidelines to identify potential coherent phases that will achieve tailored transformation temperatures and hysteresis. </div><div><br></div><div>We subsequently explore ideas to enhance the length and time scales accessible via MD simulations. We first discuss the use of kinetic Monte Carlo methods in MD simulations to predict the microstructure evolution of carbon fibers. We ?find our approach to accurately predict the transverse microstructures of carbon fibers, additionally predicting the transverse modulus of these fibers, a quantity difficult to measure via experiments. Another avenue to increase length and time scales accessible via MD simulations is to explore novel implementations of algorithms involved in machine-learned interatomic models to extract performance portability. Our approach here results in significant speedups and an efficient utilization of increasingly common CPU-GPU hybrid architectures.</div><div><br></div><div>We finally explore the use of machine learning methods in molecular dynamics, specifically developing machine learning methods to discover interpretable laws directly from data. As examples, we demonstrate the discovery of integration schemes for MD simulations, and the discovery of melting laws for perovskites and single elements. Overall, this work attempts to illustrate how improving the predictive capabilities of molecular dynamics simulations and incorporating machine learning ideas can help us design novel materials, in line with the goals of the Materials Genome Initiative.</div>