Precursors of carbon fibers include rayon, pitch, and polyacrylonitrile fibers that can be heat-treated for high-strength or high-modulus carbon fibers. Among them, polyacrylonitrile has been used most frequently due to its low viscosity for easy processing and excellent performance for high-end applications. To further explore polyacrylonitrile-based fibers for better precursors, in this study, carbon nanofillers were introduced in the polymer matrix to examine their reinforcement effects and influences on carbon fiber performance. Two-dimensional graphene nanoplatelets were mainly used for the polymer reinforcement and one-dimensional carbon nanotubes were also incorporated in polyacrylonitrile as a comparison. Dry-jet wet spinning was used to fabricate the composite fibers. Hot-stage drawing and heat-treatment were used to evolve the physical microstructures and molecular morphologies of precursor and carbon fibers. As compared to traditionally used random dispersions, selective placement of nanofillers was effective in improving composite fiber properties and enhancing mechanical and functional behaviors of carbon fibers. The particular position of reinforcement fillers with polymer layers was enabled by the in-house developed spinneret used for fiber spinning. The preferential alignment of graphitic planes contributed to the enhanced mechanical and functional behaviors than those of dispersed nanoparticles in polyacrylonitrile composites. The high in-plane modulus of graphene and the induction to polyacrylonitrile molecular carbonization/graphitization were the motivation for selectively placing graphene nanoplatelets between polyacrylonitrile layers. Mechanical tests, scanning electron microscopy, thermal, and electrical properties were characterized. Applications such as volatile organic compound sensing and pressure sensing were demonstrated.

Corrosion fatigue has been of prime concern in railways, aerospace, construction industries and so on. Even in the case of many medical equipment, corrosion fatigue is considered to be a major challenge. The fact that even high strength materials have lower resistance to corrosion fatigue makes it an interesting area for research. The analysis of propagation of fatigue crack growth under environmental interaction and the life prediction is significant to reduce the maintenance costs and assure structural integrity. Without proper investigation of the crack extension under corrosion fatigue, the scenario can lead to catastrophic disasters due to premature failure of a structure. An attempt has been made in this study to predict the corrosion fatigue crack growth with reasonable accuracy. Models that have been developed so far predict the crack propagation for constant amplitude loading (CAL). However, most of the industrial applications encounter random loading. Hence there is a need to develop models based on time scale. An existing time scale model that can predict the fatigue crack growth for constant and variable amplitude loading (VAL) in the Paris region is initially modified to extend the prediction to near threshold and unstable crack growth region. Extensive data collection was carried out to calibrate the model for corrosion fatigue crack growth (CFCG) based on the experimental data. The time scale model is improved to incorporate the effect of corrosive environments such as NaCl and dry hydrogen in the fatigue crack growth (FCG) by investigation of the trend in change of the crack growth. The time scale model gives the advantage of coupling the time phenomenon stress corrosion cracking which is suggested as a future work in this paper.

This thesis intends to cover the experimental investigation of the propagation of laser-generated optoacoustic waves in structural materials and how they can be utilized for damage detection. Firstly, a system for scanning a rectangular patch on the sample is designed. This is achieved with the help of xy stages which are connected to the laser head and allow it to move on a plane. Next, a parametric study was designed to determine the optimum testing parameters of the laser. The parameters so selected were then used in a series of tests which helped in discerning how the Ultrasound Waves behave when damage is induced in the sample (in the form of addition of masses). The first test was of increasing the mases in the sample. The second test was a scan of a rectangular area of the sample with and without damage to find the effect of the added masses. Finally, the data collected in such a manner is processed with the help of the Hilbert-Huang transform to determine the time of arrival. The major benefits from this study are the fact that this is a Non-Destructive imaging technique and thus can be used as a new method for detection of defects and is fairly cheap as well.

Machine learning has demonstrated great potential across a wide range of applications such as computer vision, robotics, speech recognition, drug discovery, material science, and physics simulation. Despite its current success, however, there are still two major challenges for machine learning algorithms: limited robustness and generalizability.

The robustness of a neural network is defined as the stability of the network output under small input perturbations. It has been shown that neural networks are very sensitive to input perturbations, and the prediction from convolutional neural networks can be totally different for input images that are visually indistinguishable to human eyes. Based on such property, hackers can reversely engineer the input to trick machine learning systems in targeted ways. These adversarial attacks have shown to be surprisingly effective, which has raised serious concerns over safety-critical applications like autonomous driving. In the meantime, many established defense mechanisms have shown to be vulnerable under more advanced attacks proposed later, and how to improve the robustness of neural networks is still an open question.

The generalizability of neural networks refers to the ability of networks to perform well on unseen data rather than just the data that they were trained on. Neural networks often fail to carry out reliable generalizations when the testing data is of different distribution compared with the training one, which will make autonomous driving systems risky under new environment. The generalizability of neural networks can also be limited whenever there is a scarcity of training data, while it can be expensive to acquire large datasets either experimentally or numerically for engineering applications, such as material and chemical design.

In this dissertation, we are thus motivated to improve the robustness and generalizability of neural networks. Firstly, unlike traditional bottom-up classifiers, we use a pre-trained generative model to perform top-down reasoning and infer the label information. The proposed generative classifier has shown to be promising in handling input distribution shifts. Secondly, we focus on improving the network robustness and propose an extension to adversarial training by considering the transformation invariance. Proposed method improves the robustness over state-of-the-art methods by 2.5% on MNIST and 3.7% on CIFAR-10. Thirdly, we focus on designing networks that generalize well at predicting physics response. Our physics prior knowledge is used to guide the designing of the network architecture, which enables efficient learning and inference. Proposed network is able to generalize well even when it is trained with a single image pair.

Non-Destructive Testing (NDT) is a branch of scientific methods and techniques

used to evaluate the defects and irregularities in engineering materials. These methods

conduct testing without destroying or altering material’s structure and functionality. Most

of these defects are subsurface making them difficult to detect and access.

SONIC INFRARED (IR) is a relatively new and emerging vibrothermography

method under the category of NDT methods. This is a fast NDT inspection method that

uses an ultrasonic generator to pass an ultrasonic pulse through the test specimen which

results in a temperature variation in the test specimen. The temperature increase around

the area of the defect is more because of frictional heating due to the vibration of the

specimen. This temperature variation can be observed using a thermal camera.

In this research study, the temperature variation in the composite laminate during

the SONIC IR experimentation using an infrared thermal camera. These recorded data are

used to determine the location, dimension and depth of defects through SONIC IR NDT

method using existing defect detection algorithms. Probability of detection analysis is

used to determine the probability of detection under specific experimental conditions for

two different types of composite laminates. Lastly, the effect of the process parameters

such as number of pulses, pulse duration and time delay between pulses of this technique

on the detectability and probability of detection is studied in detail.

An orthotropic elasto-plastic damage material model (OEPDMM) suitable for impact simulations has been developed through a joint research project funded by the Federal Aviation Administration (FAA) and the National Aeronautics and Space Administration (NASA). Development of the model includes derivation of the theoretical details, implementation of the theory into LS-DYNA®, a commercially available nonlinear transient dynamic finite element code, as material model MAT 213, and verification and validation of the model. The material model is comprised of three major components: deformation, damage, and failure. The deformation sub-model is used to capture both linear and nonlinear deformations through a classical plasticity formulation. The damage sub-model is used to account for the reduction of elastic stiffness of the material as the degree of plastic strain is increased. Finally, the failure sub-model is used to predict the onset of loss of load carrying capacity in the material. OEPDMM is driven completely by tabulated experimental data obtained through physically meaningful material characterization tests, through high fidelity virtual tests, or both. The tabulated data includes stress-strain curves at different temperatures and strain rates to drive the deformation sub-model, damage parameter-total strain curves to drive the damage sub-model, and the failure sub-model can be driven by the data required for different failure theories implemented in the computer code. The work presented herein focuses on the experiments used to obtain the data necessary to drive as well as validate the material model, development and implementation of the damage model, verification of the deformation and damage models through single element (SE) and multi-element (ME) finite element simulations, development and implementation of experimental procedure for modeling delamination, and finally validation of the material model through low speed impact simulations and high speed impact simulations.

This investigation focuses on the development of uncertainty modeling methods applicable to both the structural and thermal models of heated structures as part of an effort to enable the design under uncertainty of hypersonic vehicles. The maximum entropy-based nonparametric stochastic modeling approach is used within the context of coupled structural-thermal Reduced Order Models (ROMs). Not only does this strategy allow for a computationally efficient generation of samples of the structural and thermal responses but the maximum entropy approach allows to introduce both aleatoric and some epistemic uncertainty into the system.

While the nonparametric approach has a long history of applications to structural models, the present investigation was the first one to consider it for the heat conduction problem. In this process, it was recognized that the nonparametric approach had to be modified to maintain the localization of the temperature near the heat source, which was successfully achieved.

The introduction of uncertainty in coupled structural-thermal ROMs of heated structures was addressed next. It was first recognized that the structural stiffness coefficients (linear, quadratic, and cubic) and the parameters quantifying the effects of the temperature distribution on the structural response can be regrouped into a matrix that is symmetric and positive definite. The nonparametric approach was then applied to this matrix allowing the assessment of the effects of uncertainty on the resulting temperature distributions and structural response.

The third part of this document focuses on introducing uncertainty using the Maximum Entropy Method at the level of finite element by randomizing elemental matrices, for instance, elemental stiffness, mass and conductance matrices. This approach brings some epistemic uncertainty not present in the parametric approach (e.g., by randomizing the elasticity tensor) while retaining more local character than the operation in ROM level.

The last part of this document focuses on the development of “reduced ROMs” (RROMs) which are reduced order models with small bases constructed in a data-driven process from a “full” ROM with a much larger basis. The development of the RROM methodology is motivated by the desire to optimally reduce the computational cost especially in multi-physics situations where a lack of prior understanding/knowledge of the solution typically leads to the selection of ROM bases that are excessively broad to ensure the necessary accuracy in representing the response. It is additionally emphasized that the ROM reduction process can be carried out adaptively, i.e., differently over different ranges of loading conditions.

In this research, a new cutting edge wear estimator for micro-endmilling is developed and the reliabillity of the estimator is evaluated. The main concept of this estimator is the minimum chip thickness effect. This estimator predicts the cutting edge radius by detecting the drop in the chip production rate as the cutting edge of a micro- endmill slips over the workpiece when the minimum chip thickness becomes larger than the uncut chip thickness, thus transitioning from the shearing to the ploughing dominant regime. The chip production rate is investigated through simulation and experiment. The simulation and the experiment show that the chip production rate decreases when the minimum chip thickness becomes larger than the uncut chip thickness. Also, the reliability of this estimator is evaluated. The probability of correct estimation of the cutting edge radius is more than 80%. This cutting edge wear estimator could be applied to an online tool wear estimation system. Then, a large number of cutting edge wear data could be obtained. From the data, a cutting edge wear model could be developed in terms of the machine control parameters so that the optimum control parameters could be applied to increase the tool life and the machining quality as well by minimizing the cutting edge wear rate.

In addition, in order to find the stable condition of the machining, the stabillity lobe of the system is created by measuring the dynamic parameters. This process is needed prior to the cutting edge wear estimation since the chatter would affect the cutting edge wear and the chip production rate. In this research, a new experimental set-up for measuring the dynamic parameters is developed by using a high speed camera with microscope lens and a loadcell. The loadcell is used to measure the stiffness of the tool-holder assembly of the machine and the high speed camera is used to measure the natural frequency and the damping ratio. From the measured data, a stability lobe is created. Even though this new method needs further research, it could be more cost-effective than the conventional methods in the future.

Aging-related damage and failure in structures, such as fatigue cracking, corrosion, and delamination, are critical for structural integrity. Most engineering structures have embedded defects such as voids, cracks, inclusions from manufacturing. The properties and locations of embedded defects are generally unknown and hard to detect in complex engineering structures. Therefore, early detection of damage is beneficial for prognosis and risk management of aging infrastructure system.

Non-destructive testing (NDT) and structural health monitoring (SHM) are widely used for this purpose. Different types of NDT techniques have been proposed for the damage detection, such as optical image, ultrasound wave, thermography, eddy current, and microwave. The focus in this study is on the wave-based detection method, which is grouped into two major categories: feature-based damage detection and model-assisted damage detection. Both damage detection approaches have their own pros and cons. Feature-based damage detection is usually very fast and doesn’t involve in the solution of the physical model. The key idea is the dimension reduction of signals to achieve efficient damage detection. The disadvantage is that the loss of information due to the feature extraction can induce significant uncertainties and reduces the resolution. The resolution of the feature-based approach highly depends on the sensing path density. Model-assisted damage detection is on the opposite side. Model-assisted damage detection has the ability for high resolution imaging with limited number of sensing paths since the entire signal histories are used for damage identification. Model-based methods are time-consuming due to the requirement for the inverse wave propagation solution, which is especially true for the large 3D structures.

The motivation of the proposed method is to develop efficient and accurate model-based damage imaging technique with limited data. The special focus is on the efficiency of the damage imaging algorithm as it is the major bottleneck of the model-assisted approach. The computational efficiency is achieved by two complimentary components. First, a fast forward wave propagation solver is developed, which is verified with the classical Finite Element(FEM) solution and the speed is 10-20 times faster. Next, efficient inverse wave propagation algorithms is proposed. Classical gradient-based optimization algorithms usually require finite difference method for gradient calculation, which is prohibitively expensive for large degree of freedoms. An adjoint method-based optimization algorithms is proposed, which avoids the repetitive finite difference calculations for every imaging variables. Thus, superior computational efficiency can be achieved by combining these two methods together for the damage imaging. A coupled Piezoelectric (PZT) damage imaging model is proposed to include the interaction between PZT and host structure. Following the formulation of the framework, experimental validation is performed on isotropic and anisotropic material with defects such as cracks, delamination, and voids. The results show that the proposed method can detect and reconstruct multiple damage simultaneously and efficiently, which is promising to be applied to complex large-scale engineering structures.

Advanced material systems refer to materials that are comprised of multiple traditional constituents but complex microstructure morphologies, which lead to their superior properties over conventional materials. This dissertation is motivated by the grand challenge in accelerating the design of advanced material systems through systematic optimization with respect to material microstructures or processing settings. While optimization techniques have mature applications to a large range of engineering systems, their application to material design meets unique challenges due to the high dimensionality of microstructures and the high costs in computing process-structure-property (PSP) mappings. The key to addressing these challenges is the learning of material representations and predictive PSP mappings while managing a small data acquisition budget. This dissertation thus focuses on developing learning mechanisms that leverage context-specific meta-data and physics-based theories. Two research tasks will be conducted: In the first, we develop a statistical generative model that learns to characterize high-dimensional microstructure samples using low-dimensional features. We improve the data efficiency of a variational autoencoder by introducing a morphology loss to the training. We demonstrate that the resultant microstructure generator is morphology-aware when trained on a small set of material samples, and can effectively constrain the microstructure space during material design. In the second task, we investigate an active learning mechanism where new samples are acquired based on their violation to a theory-driven constraint on the physics-based model. We demonstrate using a topology optimization case that while data acquisition through the physics-based model is often expensive (e.g., obtaining microstructures through simulation or optimization processes), the evaluation of the constraint can be far more affordable (e.g., checking whether a solution is optimal or equilibrium). We show that this theory-driven learning algorithm can lead to much improved learning efficiency and generalization performance when such constraints can be derived. The outcomes of this research is a better understanding of how physics knowledge about material systems can be integrated into machine learning frameworks, in order to achieve more cost-effective and reliable learning of material representations and predictive models, which are essential to accelerate computational material design.