Matching Items (15)
Filtering by
- All Subjects: Materials Science
- All Subjects: Machine Learning
- Status: Published
Description
Damage detection in heterogeneous material systems is a complex problem and requires an in-depth understanding of the material characteristics and response under varying load and environmental conditions. A significant amount of research has been conducted in this field to enhance the fidelity of damage assessment methodologies, using a wide range of sensors and detection techniques, for both metallic materials and composites. However, detecting damage at the microscale is not possible with commercially available sensors. A probable way to approach this problem is through accurate and efficient multiscale modeling techniques, which are capable of tracking damage initiation at the microscale and propagation across the length scales. The output from these models will provide an improved understanding of damage initiation; the knowledge can be used in conjunction with information from physical sensors to improve the size of detectable damage. In this research, effort has been dedicated to develop multiscale modeling approaches and associated damage criteria for the estimation of damage evolution across the relevant length scales. Important issues such as length and time scales, anisotropy and variability in material properties at the microscale, and response under mechanical and thermal loading are addressed. Two different material systems have been studied: metallic material and a novel stress-sensitive epoxy polymer.
For metallic material (Al 2024-T351), the methodology initiates at the microscale where extensive material characterization is conducted to capture the microstructural variability. A statistical volume element (SVE) model is constructed to represent the material properties. Geometric and crystallographic features including grain orientation, misorientation, size, shape, principal axis direction and aspect ratio are captured. This SVE model provides a computationally efficient alternative to traditional techniques using representative volume element (RVE) models while maintaining statistical accuracy. A physics based multiscale damage criterion is developed to simulate the fatigue crack initiation. The crack growth rate and probable directions are estimated simultaneously.
Mechanically sensitive materials that exhibit specific chemical reactions upon external loading are currently being investigated for self-sensing applications. The "smart" polymer modeled in this research consists of epoxy resin, hardener, and a stress-sensitive material called mechanophore The mechanophore activation is based on covalent bond-breaking induced by external stimuli; this feature can be used for material-level damage detections. In this work Tris-(Cinnamoyl oxymethyl)-Ethane (TCE) is used as the cyclobutane-based mechanophore (stress-sensitive) material in the polymer matrix. The TCE embedded polymers have shown promising results in early damage detection through mechanically induced fluorescence. A spring-bead based network model, which bridges nanoscale information to higher length scales, has been developed to model this material system. The material is partitioned into discrete mass beads which are linked using linear springs at the microscale. A series of MD simulations were performed to define the spring stiffness in the statistical network model. By integrating multiple spring-bead models a network model has been developed to represent the material properties at the mesoscale. The model captures the statistical distribution of crosslinking degree of the polymer to represent the heterogeneous material properties at the microscale. The developed multiscale methodology is computationally efficient and provides a possible means to bridge multiple length scales (from 10 nm in MD simulation to 10 mm in FE model) without significant loss of accuracy. Parametric studies have been conducted to investigate the influence of the crosslinking degree on the material behavior. The developed methodology has been used to evaluate damage evolution in the self-sensing polymer.
For metallic material (Al 2024-T351), the methodology initiates at the microscale where extensive material characterization is conducted to capture the microstructural variability. A statistical volume element (SVE) model is constructed to represent the material properties. Geometric and crystallographic features including grain orientation, misorientation, size, shape, principal axis direction and aspect ratio are captured. This SVE model provides a computationally efficient alternative to traditional techniques using representative volume element (RVE) models while maintaining statistical accuracy. A physics based multiscale damage criterion is developed to simulate the fatigue crack initiation. The crack growth rate and probable directions are estimated simultaneously.
Mechanically sensitive materials that exhibit specific chemical reactions upon external loading are currently being investigated for self-sensing applications. The "smart" polymer modeled in this research consists of epoxy resin, hardener, and a stress-sensitive material called mechanophore The mechanophore activation is based on covalent bond-breaking induced by external stimuli; this feature can be used for material-level damage detections. In this work Tris-(Cinnamoyl oxymethyl)-Ethane (TCE) is used as the cyclobutane-based mechanophore (stress-sensitive) material in the polymer matrix. The TCE embedded polymers have shown promising results in early damage detection through mechanically induced fluorescence. A spring-bead based network model, which bridges nanoscale information to higher length scales, has been developed to model this material system. The material is partitioned into discrete mass beads which are linked using linear springs at the microscale. A series of MD simulations were performed to define the spring stiffness in the statistical network model. By integrating multiple spring-bead models a network model has been developed to represent the material properties at the mesoscale. The model captures the statistical distribution of crosslinking degree of the polymer to represent the heterogeneous material properties at the microscale. The developed multiscale methodology is computationally efficient and provides a possible means to bridge multiple length scales (from 10 nm in MD simulation to 10 mm in FE model) without significant loss of accuracy. Parametric studies have been conducted to investigate the influence of the crosslinking degree on the material behavior. The developed methodology has been used to evaluate damage evolution in the self-sensing polymer.
ContributorsZhang, Jinjun (Author) / Chattopadhyay, Aditi (Thesis advisor) / Dai, Lenore (Committee member) / Jiang, Hanqing (Committee member) / Papandreou-Suppappola, Antonia (Committee member) / Rajadas, John (Committee member) / Arizona State University (Publisher)
Created2014
Description
Identification of early damage in polymer composite materials is of significant importance so that preventative measures can be taken before the materials reach catastrophic failure. Scientists have been developing damage detection technologies over many years and recently, mechanophore-based polymers, in which mechanical energy is translated to activate a chemical transformation, have received increasing attention. More specifically, the damage can be made detectable by mechanochromic polymers, which provide a visible color change upon the scission of covalent bonds under stress. This dissertation focuses on the study of a novel self-sensing framework for identifying early and in-situ damage by employing unique stress-sensing mechanophores. Two types of mechanophores, cyclobutane and cyclooctane, were utilized, and the former formed from cinnamoyl moeities and the latter formed from anthracene upon photodimerization. The effects on the thermal and mechanical properties with the addition of the cyclobutane-based polymers into epoxy matrices were investigated. The emergence of cracks was detected by fluorescent signals at a strain level right after the yield point of the polymer blends, and the fluorescence intensified with the accumulation of strain. Similar to the mechanism of fluorescence emission from the cleavage of cyclobutane, the cyclooctane moiety generated fluorescent emission with a higher quantum yield upon cleavage. The experimental results also demonstrated the success of employing the cyclooctane type mechanophore as a potential force sensor, as the fluorescence intensification was correlated with the strain increase.
ContributorsZou, Jin (Author) / Dai, Lenore L (Thesis advisor) / Chattopadhyay, Aditi (Thesis advisor) / Lind, Mary L (Committee member) / Mu, Bin (Committee member) / Yu, Hongyu (Committee member) / Arizona State University (Publisher)
Created2014
Description
Advanced composites are being widely used in aerospace applications due to their high stiffness, strength and energy absorption capabilities. However, the assurance of structural reliability is a critical issue because a damage event will compromise the integrity of composite structures and lead to ultimate failure. In this dissertation a novel homogenization based multiscale modeling framework using semi-analytical micromechanics is presented to simulate the response of textile composites. The novelty of this approach lies in the three scale homogenization/localization framework bridging between the constituent (micro), the fiber tow scale (meso), weave scale (macro), and the global response. The multiscale framework, named Multiscale Generalized Method of Cells (MSGMC), continuously bridges between the micro to the global scale as opposed to approaches that are top-down and bottom-up. This framework is fully generalized and capable of modeling several different weave and braids without reformulation. Particular emphasis in this dissertation is placed on modeling the nonlinearity and failure of both polymer matrix and ceramic matrix composites.
ContributorsLiu, Guang (Author) / Chattopadhyay, Aditi (Thesis advisor) / Mignolet, Marc (Committee member) / Jiang, Hanqing (Committee member) / Li, Jian (Committee member) / Rajadas, John (Committee member) / Arizona State University (Publisher)
Created2011
Description
Composite materials are increasingly being used in aircraft, automobiles, and other applications due to their high strength to weight and stiffness to weight ratios. However, the presence of damage, such as delamination or matrix cracks, can significantly compromise the performance of these materials and result in premature failure. Structural components are often manually inspected to detect the presence of damage. This technique, known as schedule based maintenance, however, is expensive, time-consuming, and often limited to easily accessible structural elements. Therefore, there is an increased demand for robust and efficient Structural Health Monitoring (SHM) techniques that can be used for Condition Based Monitoring, which is the method in which structural components are inspected based upon damage metrics as opposed to flight hours. SHM relies on in situ frameworks for detecting early signs of damage in exposed and unexposed structural elements, offering not only reduced number of schedule based inspections, but also providing better useful life estimates. SHM frameworks require the development of different sensing technologies, algorithms, and procedures to detect, localize, quantify, characterize, as well as assess overall damage in aerospace structures so that strong estimations in the remaining useful life can be determined. The use of piezoelectric transducers along with guided Lamb waves is a method that has received considerable attention due to the weight, cost, and function of the systems based on these elements. The research in this thesis investigates the ability of Lamb waves to detect damage in feature dense anisotropic composite panels. Most current research negates the effects of experimental variability by performing tests on structurally simple isotropic plates that are used as a baseline and damaged specimen. However, in actual applications, variability cannot be negated, and therefore there is a need to research the effects of complex sample geometries, environmental operating conditions, and the effects of variability in material properties. This research is based on experiments conducted on a single blade-stiffened anisotropic composite panel that localizes delamination damage caused by impact. The overall goal was to utilize a correlative approach that used only the damage feature produced by the delamination as the damage index. This approach was adopted because it offered a simplistic way to determine the existence and location of damage without having to conduct a more complex wave propagation analysis or having to take into account the geometric complexities of the test specimen. Results showed that even in a complex structure, if the damage feature can be extracted and measured, then an appropriate damage index can be associated to it and the location of the damage can be inferred using a dense sensor array. The second experiment presented in this research studies the effects of temperature on damage detection when using one test specimen for a benchmark data set and another for damage data collection. This expands the previous experiment into exploring not only the effects of variable temperature, but also the effects of high experimental variability. Results from this work show that the damage feature in the data is not only extractable at higher temperatures, but that the data from one panel at one temperature can be directly compared to another panel at another temperature for baseline comparison due to linearity of the collected data.
ContributorsVizzini, Anthony James, II (Author) / Chattopadhyay, Aditi (Thesis advisor) / Fard, Masoud (Committee member) / Papandreou-Suppappola, Antonia (Committee member) / Arizona State University (Publisher)
Created2012
Description
With the maturity of advanced composites as feasible structural materials for various applications there is a critical need to solve the challenge of designing these material systems for optimal performance. However, determining superior design methods requires a deep understanding of the material-structure properties at various length scales. Due to the length-scale dependent behavior of advanced composites, multiscale modeling techniques may be used to describe the dominant mechanisms of damage and failure in these material systems. With polymer matrix fiber composites and nanocomposites it becomes essential to include even the atomic length scale, where the resin-hardener-nanofiller molecules interact, in the multiscale modeling framework. Additionally, sources of variability are also critical to be included in these models due to the important role of uncertainty in advance composite behavior. Such a methodology should be able to describe length scale dependent mechanisms in a computationally efficient manner for the analysis of practical composite structures.
In the research presented in this dissertation, a comprehensive nano to macro multiscale framework is developed for the mechanical and multifunctional analysis of advanced composite materials and structures. An atomistically informed statistical multiscale model is developed for linear problems, to estimate and scale elastic properties of carbon fiber reinforced polymer composites (CFRPs) and carbon nanotube (CNT) enhanced CFRPs using information from molecular dynamics simulation of the resin-hardener-nanofiller nanoscale system. For modeling inelastic processes, an atomistically informed coupled damage-plasticity model is developed using the framework of continuum damage mechanics, where fundamental nanoscale covalent bond disassociation information is scaled up as a continuum scale damage identifying parameter. This damage model is coupled with a nanocomposite microstructure generation algorithm to study the sub-microscale damage mechanisms in CNT/CFRP microstructures. It is further integrated in a generalized method of cells (GMC) micromechanics model to create a low-fidelity computationally efficient nonlinear multiscale method with imperfect interfaces between the fiber and matrix, where the interface behavior is adopted from nanoscale MD simulations. This algorithm is used to understand damage mechanisms in adhesively bonded composite joints as a case study for the comprehensive nano to macroscale structural analysis of practical composites structures. At each length scale sources of variability are identified, characterized, and included in the multiscale modeling framework.
In the research presented in this dissertation, a comprehensive nano to macro multiscale framework is developed for the mechanical and multifunctional analysis of advanced composite materials and structures. An atomistically informed statistical multiscale model is developed for linear problems, to estimate and scale elastic properties of carbon fiber reinforced polymer composites (CFRPs) and carbon nanotube (CNT) enhanced CFRPs using information from molecular dynamics simulation of the resin-hardener-nanofiller nanoscale system. For modeling inelastic processes, an atomistically informed coupled damage-plasticity model is developed using the framework of continuum damage mechanics, where fundamental nanoscale covalent bond disassociation information is scaled up as a continuum scale damage identifying parameter. This damage model is coupled with a nanocomposite microstructure generation algorithm to study the sub-microscale damage mechanisms in CNT/CFRP microstructures. It is further integrated in a generalized method of cells (GMC) micromechanics model to create a low-fidelity computationally efficient nonlinear multiscale method with imperfect interfaces between the fiber and matrix, where the interface behavior is adopted from nanoscale MD simulations. This algorithm is used to understand damage mechanisms in adhesively bonded composite joints as a case study for the comprehensive nano to macroscale structural analysis of practical composites structures. At each length scale sources of variability are identified, characterized, and included in the multiscale modeling framework.
ContributorsRai, Ashwin (Author) / Chattopadhyay, Aditi (Thesis advisor) / Dai, Lenore (Committee member) / Jiang, Hanqing (Committee member) / Rajadas, John (Committee member) / Fard, Masoud Yekani (Committee member) / Arizona State University (Publisher)
Created2018
Description
This dissertation presents the development of structural health monitoring and prognostic health management methodologies for complex structures and systems in the field of mechanical engineering. To overcome various challenges historically associated with complex structures and systems such as complicated sensing mechanisms, noisy information, and large-size datasets, a hybrid monitoring framework comprising of solid mechanics concepts and data mining technologies is developed. In such a framework, the solid mechanics simulations provide additional intuitions to data mining techniques reducing the dependence of accuracy on the training set, while the data mining approaches fuse and interpret information from the targeted system enabling the capability for real-time monitoring with efficient computation.
In the case of structural health monitoring, ultrasonic guided waves are utilized for damage identification and localization in complex composite structures. Signal processing and data mining techniques are integrated into the damage localization framework, and the converted wave modes, which are induced by the thickness variation due to the presence of delamination, are used as damage indicators. This framework has been validated through experiments and has shown sufficient accuracy in locating delamination in X-COR sandwich composites without the need of baseline information. Besides the localization of internal damage, the Gaussian process machine learning technique is integrated with finite element method as an online-offline prediction model to predict crack propagation with overloads under biaxial loading conditions; such a probabilistic prognosis model, with limited number of training examples, has shown increased accuracy over state-of-the-art techniques in predicting crack retardation behaviors induced by overloads. In the case of system level management, a monitoring framework built using a multivariate Gaussian model as basis is developed to evaluate the anomalous condition of commercial aircrafts. This method has been validated using commercial airline data and has shown high sensitivity to variations in aircraft dynamics and pilot operations. Moreover, this framework was also tested on simulated aircraft faults and its feasibility for real-time monitoring was demonstrated with sufficient computation efficiency.
This research is expected to serve as a practical addition to the existing literature while possessing the potential to be adopted in realistic engineering applications.
In the case of structural health monitoring, ultrasonic guided waves are utilized for damage identification and localization in complex composite structures. Signal processing and data mining techniques are integrated into the damage localization framework, and the converted wave modes, which are induced by the thickness variation due to the presence of delamination, are used as damage indicators. This framework has been validated through experiments and has shown sufficient accuracy in locating delamination in X-COR sandwich composites without the need of baseline information. Besides the localization of internal damage, the Gaussian process machine learning technique is integrated with finite element method as an online-offline prediction model to predict crack propagation with overloads under biaxial loading conditions; such a probabilistic prognosis model, with limited number of training examples, has shown increased accuracy over state-of-the-art techniques in predicting crack retardation behaviors induced by overloads. In the case of system level management, a monitoring framework built using a multivariate Gaussian model as basis is developed to evaluate the anomalous condition of commercial aircrafts. This method has been validated using commercial airline data and has shown high sensitivity to variations in aircraft dynamics and pilot operations. Moreover, this framework was also tested on simulated aircraft faults and its feasibility for real-time monitoring was demonstrated with sufficient computation efficiency.
This research is expected to serve as a practical addition to the existing literature while possessing the potential to be adopted in realistic engineering applications.
ContributorsLi, Guoyi (Ph.D.) (Author) / Chattopadhyay, Aditi (Thesis advisor) / Mignolet, Marc (Committee member) / Papandreou-Suppappola, Antonia (Committee member) / Yekani Fard, Masoud (Committee member) / Jiang, Hanqing (Committee member) / Arizona State University (Publisher)
Created2019
Description
All structures suffer wear and tear because of impact, excessive load, fatigue, corrosion, etc. in addition to inherent defects during their manufacturing processes and their exposure to various environmental effects. These structural degradations are often imperceptible, but they can severely affect the structural performance of a component, thereby severely decreasing its service life. Although previous studies of Structural Health Monitoring (SHM) have revealed extensive prior knowledge on the parts of SHM processes, such as the operational evaluation, data processing, and feature extraction, few studies have been conducted from a systematical perspective, the statistical model development.
The first part of this dissertation, the characteristics of inverse scattering problems, such as ill-posedness and nonlinearity, reviews ultrasonic guided wave-based structural health monitoring problems. The distinctive features and the selection of the domain analysis are investigated by analytically searching the conditions of the uniqueness solutions for ill-posedness and are validated experimentally.
Based on the distinctive features, a novel wave packet tracing (WPT) method for damage localization and size quantification is presented. This method involves creating time-space representations of the guided Lamb waves (GLWs), collected at a series of locations, with a spatially dense distribution along paths at pre-selected angles with respect to the direction, normal to the direction of wave propagation. The fringe patterns due to wave dispersion, which depends on the phase velocity, are selected as the primary features that carry information, regarding the wave propagation and scattering.
The following part of this dissertation presents a novel damage-localization framework, using a fully automated process. In order to construct the statistical model for autonomous damage localization deep-learning techniques, such as restricted Boltzmann machine and deep belief network, are trained and utilized to interpret nonlinear far-field wave patterns.
Next, a novel bridge scour estimation approach that comprises advantages of both empirical and data-driven models is developed. Two field datasets from the literature are used, and a Support Vector Machine (SVM), a machine-learning algorithm, is used to fuse the field data samples and classify the data with physical phenomena. The Fast Non-dominated Sorting Genetic Algorithm (NSGA-II) is evaluated on the model performance objective functions to search for Pareto optimal fronts.
The first part of this dissertation, the characteristics of inverse scattering problems, such as ill-posedness and nonlinearity, reviews ultrasonic guided wave-based structural health monitoring problems. The distinctive features and the selection of the domain analysis are investigated by analytically searching the conditions of the uniqueness solutions for ill-posedness and are validated experimentally.
Based on the distinctive features, a novel wave packet tracing (WPT) method for damage localization and size quantification is presented. This method involves creating time-space representations of the guided Lamb waves (GLWs), collected at a series of locations, with a spatially dense distribution along paths at pre-selected angles with respect to the direction, normal to the direction of wave propagation. The fringe patterns due to wave dispersion, which depends on the phase velocity, are selected as the primary features that carry information, regarding the wave propagation and scattering.
The following part of this dissertation presents a novel damage-localization framework, using a fully automated process. In order to construct the statistical model for autonomous damage localization deep-learning techniques, such as restricted Boltzmann machine and deep belief network, are trained and utilized to interpret nonlinear far-field wave patterns.
Next, a novel bridge scour estimation approach that comprises advantages of both empirical and data-driven models is developed. Two field datasets from the literature are used, and a Support Vector Machine (SVM), a machine-learning algorithm, is used to fuse the field data samples and classify the data with physical phenomena. The Fast Non-dominated Sorting Genetic Algorithm (NSGA-II) is evaluated on the model performance objective functions to search for Pareto optimal fronts.
ContributorsKim, Inho (Author) / Chattopadhyay, Aditi (Thesis advisor) / Jiang, Hanqing (Committee member) / Liu, Yongming (Committee member) / Mignolet, Marc (Committee member) / Rajadas, John (Committee member) / Arizona State University (Publisher)
Created2016
Description
There is a concerted effort in developing robust systems health monitoring/management (SHM) technology as a means to reduce the life cycle costs, improve availability, extend life and minimize downtime of various platforms including aerospace and civil infrastructure. The implementation of a robust SHM system requires a collaborative effort in a variety of areas such as sensor development, damage detection and localization, physics based models, and prognosis models for residual useful life (RUL) estimation. Damage localization and prediction is further complicated by geometric, material, loading, and environmental variabilities. Therefore, it is essential to develop robust SHM methodologies by taking into account such uncertainties. In this research, damage localization and RUL estimation of two different physical systems are addressed: (i) fatigue crack propagation in metallic materials under complex multiaxial loading and (ii) temporal scour prediction near bridge piers. With little modifications, the methodologies developed can be applied to other systems.
Current practice in fatigue life prediction is based on either physics based modeling or data-driven methods, and is limited to predicting RUL for simple geometries under uniaxial loading conditions. In this research, crack initiation and propagation behavior under uniaxial and complex biaxial fatigue loading is addressed. The crack propagation behavior is studied by performing extensive material characterization and fatigue testing under in-plane biaxial loading, both in-phase and out-of-phase, with different biaxiality ratios. A hybrid prognosis model, which combines machine learning with physics based modeling, is developed to account for the uncertainties in crack propagation and fatigue life prediction due to variabilities in material microstructural characteristics, crack localization information and environmental changes. The methodology iteratively combines localization information with hybrid prognosis models using sequential Bayesian techniques. The results show significant improvements in the localization and prediction accuracy under varying temperature.
For civil infrastructure, especially bridges, pier scour is a major failure mechanism. Currently available techniques are developed from a design perspective and provide highly conservative scour estimates. In this research, a fully probabilistic scour prediction methodology is developed using machine learning to accurately predict scour in real-time under varying flow conditions.
Current practice in fatigue life prediction is based on either physics based modeling or data-driven methods, and is limited to predicting RUL for simple geometries under uniaxial loading conditions. In this research, crack initiation and propagation behavior under uniaxial and complex biaxial fatigue loading is addressed. The crack propagation behavior is studied by performing extensive material characterization and fatigue testing under in-plane biaxial loading, both in-phase and out-of-phase, with different biaxiality ratios. A hybrid prognosis model, which combines machine learning with physics based modeling, is developed to account for the uncertainties in crack propagation and fatigue life prediction due to variabilities in material microstructural characteristics, crack localization information and environmental changes. The methodology iteratively combines localization information with hybrid prognosis models using sequential Bayesian techniques. The results show significant improvements in the localization and prediction accuracy under varying temperature.
For civil infrastructure, especially bridges, pier scour is a major failure mechanism. Currently available techniques are developed from a design perspective and provide highly conservative scour estimates. In this research, a fully probabilistic scour prediction methodology is developed using machine learning to accurately predict scour in real-time under varying flow conditions.
ContributorsNeerukatti, Rajesh Kumar (Author) / Chattopadhyay, Aditi (Thesis advisor) / Jiang, Hanqing (Committee member) / Liu, Yongming (Committee member) / Rajadas, John (Committee member) / Yekani Fard, Masoud (Committee member) / Arizona State University (Publisher)
Created2016
Stochastic multiscale modeling and statistical characterization of complex polymer matrix composites
Description
There are many applications for polymer matrix composite materials in a variety of different industries, but designing and modeling with these materials remains a challenge due to the intricate architecture and damage modes. Multiscale modeling techniques of composite structures subjected to complex loadings are needed in order to address the scale-dependent behavior and failure. The rate dependency and nonlinearity of polymer matrix composite materials further complicates the modeling. Additionally, variability in the material constituents plays an important role in the material behavior and damage. The systematic consideration of uncertainties is as important as having the appropriate structural model, especially during model validation where the total error between physical observation and model prediction must be characterized. It is necessary to quantify the effects of uncertainties at every length scale in order to fully understand their impact on the structural response. Material variability may include variations in fiber volume fraction, fiber dimensions, fiber waviness, pure resin pockets, and void distributions. Therefore, a stochastic modeling framework with scale dependent constitutive laws and an appropriate failure theory is required to simulate the behavior and failure of polymer matrix composite structures subjected to complex loadings. Additionally, the variations in environmental conditions for aerospace applications and the effect of these conditions on the polymer matrix composite material need to be considered. The research presented in this dissertation provides the framework for stochastic multiscale modeling of composites and the characterization data needed to determine the effect of different environmental conditions on the material properties. The developed models extend sectional micromechanics techniques by incorporating 3D progressive damage theories and multiscale failure criteria. The mechanical testing of composites under various environmental conditions demonstrates the degrading effect these conditions have on the elastic and failure properties of the material. The methodologies presented in this research represent substantial progress toward understanding the failure and effect of variability for complex polymer matrix composites.
ContributorsJohnston, Joel Philip (Author) / Chattopadhyay, Aditi (Thesis advisor) / Liu, Yongming (Committee member) / Jiang, Hanqing (Committee member) / Dai, Lenore (Committee member) / Rajadas, John (Committee member) / Arizona State University (Publisher)
Created2016
Description
Uncertainty quantification is critical for engineering design and analysis. Determining appropriate ways of dealing with uncertainties has been a constant challenge in engineering. Statistical methods provide a powerful aid to describe and understand uncertainties. This work focuses on applying Bayesian methods and machine learning in uncertainty quantification and prognostics among all the statistical methods. This study focuses on the mechanical properties of materials, both static and fatigue, the main engineering field on which this study focuses. This work can be summarized in the following items: First, maintaining the safety of vintage pipelines requires accurately estimating the strength. The objective is to predict the reliability-based strength using nondestructive multimodality surface information. Bayesian model averaging (BMA) is implemented for fusing multimodality non-destructive testing results for gas pipeline strength estimation. Several incremental improvements are proposed in the algorithm implementation. Second, the objective is to develop a statistical uncertainty quantification method for fatigue stress-life (S-N) curves with sparse data.Hierarchical Bayesian data augmentation (HBDA) is proposed to integrate hierarchical Bayesian modeling (HBM) and Bayesian data augmentation (BDA) to deal with sparse data problems for fatigue S-N curves. The third objective is to develop a physics-guided machine learning model to overcome limitations in parametric regression models and classical machine learning models for fatigue data analysis. A Probabilistic Physics-guided Neural Network (PPgNN) is proposed for probabilistic fatigue S-N curve estimation. This model is further developed for missing data and arbitrary output distribution problems. Fourth, multi-fidelity modeling combines the advantages of low- and high-fidelity models to achieve a required accuracy at a reasonable computation cost. The fourth objective is to develop a neural network approach for multi-fidelity modeling by learning the correlation between low- and high-fidelity models. Finally, conclusions are drawn, and future work is outlined based on the current study.
ContributorsChen, Jie (Author) / Liu, Yongming (Thesis advisor) / Chattopadhyay, Aditi (Committee member) / Mignolet, Marc (Committee member) / Ren, Yi (Committee member) / Yan, Hao (Committee member) / Arizona State University (Publisher)
Created2022