Matching Items (41)
Description
This thesis encompasses a comprehensive research effort dedicated to overcoming the critical bottlenecks that hinder the current generation of neural networks, thereby significantly advancing their reliability and performance. Deep neural networks, with their millions of parameters, suffer from over-parameterization and lack of constraints, leading to limited generalization capabilities. In other words, the complex architecture and millions of parameters present challenges in finding the right balance between capturing useful patterns and avoiding noise in the data. To address these issues, this thesis explores novel solutions based on knowledge distillation, enabling the learning of robust representations. Leveraging the capabilities of large-scale networks, effective learning strategies are developed. Moreover, the limitations of dependency on external networks in the distillation process, which often require large-scale models, are effectively overcome by proposing a self-distillation strategy. The proposed approach empowers the model to generate high-level knowledge within a single network, pushing the boundaries of knowledge distillation. The effectiveness of the proposed method is not only demonstrated across diverse applications, including image classification, object detection, and semantic segmentation but also explored in practical considerations such as handling data scarcity and assessing the transferability of the model to other learning tasks. Another major obstacle hindering the development of reliable and robust models lies in their black-box nature, impeding clear insights into the contributions toward the final predictions and yielding uninterpretable feature representations. To address this challenge, this thesis introduces techniques that incorporate simple yet powerful deep constraints rooted in Riemannian geometry. These constraints confer geometric qualities upon the latent representation, thereby fostering a more interpretable and insightful representation. In addition to its primary focus on general tasks like image classification and activity recognition, this strategy offers significant benefits in real-world applications where data scarcity is prevalent. Moreover, its robustness in feature removal showcases its potential for edge applications. By successfully tackling these challenges, this research contributes to advancing the field of machine learning and provides a foundation for building more reliable and robust systems across various application domains.
ContributorsChoi, Hongjun (Author) / Turaga, Pavan (Thesis advisor) / Jayasuriya, Suren (Committee member) / Li, Wenwen (Committee member) / Fazli, Pooyan (Committee member) / Arizona State University (Publisher)
Created2023
Description
Semiconductor devices often face reliability issues due to their operational con-
ditions causing performance degradation over time. One of the root causes of such
degradation is due to point defect dynamics and time dependent changes in their
chemical nature. Previously developed Unified Solver was successful in explaining
the copper (Cu) metastability issues in cadmium telluride (CdTe) solar cells. The
point defect formalism employed there could not be extended to chlorine or arsenic
due to numerical instabilities with the dopant chemical reactions. To overcome these
shortcomings, an advanced version of the Unified Solver called PVRD-FASP tool was
developed. This dissertation presents details about PVRD-FASP tool, the theoretical
framework for point defect chemical formalism, challenges faced with numerical al-
gorithms, improvements for the user interface, application and/or validation of the
tool with carefully chosen simulations, and open source availability of the tool for the
scientific community.
Treating point defects and charge carriers on an equal footing in the new formalism
allows to incorporate chemical reaction rate term as generation-recombination(G-R)
term in continuity equation. Due to the stiff differential equations involved, a reaction
solver based on forward Euler method with Newton step is proposed in this work.
The Jacobian required for Newton step is analytically calculated in an elegant way
improving speed, stability and accuracy of the tool. A novel non-linear correction
scheme is proposed and implemented to resolve charge conservation issue.
The proposed formalism is validated in 0-D with time evolution of free carriers
simulation and with doping limits of Cu in CdTe simulation. Excellent agreement of
light JV curves calculated with PVRD-FASP and Silvaco Atlas tool for a 1-D CdTe
solar cell validates reaction formalism and tool accuracy. A closer match with the Cu
SIMS profiles of Cu activated CdTe samples at four different anneal recipes to the
simulation results show practical applicability. A 1D simulation of full stack CdTe
device with Cu activation at 350C 3min anneal recipe and light JV curve simulation
demonstrates the tool capabilities in performing process and device simulations. CdTe
device simulation for understanding differences between traps and recombination
centers in grain boundaries demonstrate 2D capabilities.
ditions causing performance degradation over time. One of the root causes of such
degradation is due to point defect dynamics and time dependent changes in their
chemical nature. Previously developed Unified Solver was successful in explaining
the copper (Cu) metastability issues in cadmium telluride (CdTe) solar cells. The
point defect formalism employed there could not be extended to chlorine or arsenic
due to numerical instabilities with the dopant chemical reactions. To overcome these
shortcomings, an advanced version of the Unified Solver called PVRD-FASP tool was
developed. This dissertation presents details about PVRD-FASP tool, the theoretical
framework for point defect chemical formalism, challenges faced with numerical al-
gorithms, improvements for the user interface, application and/or validation of the
tool with carefully chosen simulations, and open source availability of the tool for the
scientific community.
Treating point defects and charge carriers on an equal footing in the new formalism
allows to incorporate chemical reaction rate term as generation-recombination(G-R)
term in continuity equation. Due to the stiff differential equations involved, a reaction
solver based on forward Euler method with Newton step is proposed in this work.
The Jacobian required for Newton step is analytically calculated in an elegant way
improving speed, stability and accuracy of the tool. A novel non-linear correction
scheme is proposed and implemented to resolve charge conservation issue.
The proposed formalism is validated in 0-D with time evolution of free carriers
simulation and with doping limits of Cu in CdTe simulation. Excellent agreement of
light JV curves calculated with PVRD-FASP and Silvaco Atlas tool for a 1-D CdTe
solar cell validates reaction formalism and tool accuracy. A closer match with the Cu
SIMS profiles of Cu activated CdTe samples at four different anneal recipes to the
simulation results show practical applicability. A 1D simulation of full stack CdTe
device with Cu activation at 350C 3min anneal recipe and light JV curve simulation
demonstrates the tool capabilities in performing process and device simulations. CdTe
device simulation for understanding differences between traps and recombination
centers in grain boundaries demonstrate 2D capabilities.
ContributorsShaik, Abdul Rawoof (Author) / Vasileska, Dragica (Thesis advisor) / Ringhofer, Christian (Committee member) / Sankin, Igor (Committee member) / Brinkman, Daniel (Committee member) / Goodnick, Stephen (Committee member) / Bertoni, Mariana (Committee member) / Arizona State University (Publisher)
Created2019
Description
Models that learn from data are widely and rapidly being deployed today for real-world use, and have become an integral and embedded part of human lives. While these technological advances are exciting and impactful, such data-driven computer vision systems often fail in inscrutable ways. This dissertation seeks to study and improve the reliability of machine learning models from several perspectives including the development of robust training algorithms to mitigate the risks of such failures, construction of new datasets that provide a new perspective on capabilities of vision models, and the design of evaluation metrics for re-calibrating the perception of performance improvements. I will first address distribution shift in image classification with the following contributions: (1) two methods for improving the robustness of image classifiers to distribution shift by leveraging the classifier's failures into an adversarial data transformation pipeline guided by domain knowledge, (2) an interpolation-based technique for flagging out-of-distribution samples, and (3) an intriguing trade-off between distributional and adversarial robustness resulting from data modification strategies. I will then explore reliability considerations for \textit{semantic vision} models that learn from both visual and natural language data; I will discuss how logical and semantic sentence transformations affect the performance of vision--language models and my contributions towards developing knowledge-guided learning algorithms to mitigate these failures. Finally, I will describe the effort towards building and evaluating complex reasoning capabilities of vision--language models towards the long-term goal of robust and reliable computer vision models that can communicate, collaborate, and reason with humans.
ContributorsGokhale, Tejas (Author) / Yang, Yezhou (Thesis advisor) / Baral, Chitta (Thesis advisor) / Ben Amor, Heni (Committee member) / Anirudh, Rushil (Committee member) / Arizona State University (Publisher)
Created2023
Description
The past decade witnessed the success of deep learning models in various applications of computer vision and natural language processing. This success can be predominantly attributed to the (i) availability of large amounts of training data; (ii) access of domain aware knowledge; (iii) i.i.d assumption between the train and target distributions and (iv) belief on existing metrics as reliable indicators of performance. When any of these assumptions are violated, the models exhibit brittleness producing adversely varied behavior. This dissertation focuses on methods for accurate model design and characterization that enhance process reliability when certain assumptions are not met. With the need to safely adopt artificial intelligence tools in practice, it is vital to build reliable failure detectors that indicate regimes where the model must not be invoked. To that end, an error predictor trained with a self-calibration objective is developed to estimate loss consistent with the underlying model. The properties of the error predictor are described and their utility in supporting introspection via feature importances and counterfactual explanations is elucidated. While such an approach can signal data regime changes, it is critical to calibrate models using regimes of inlier (training) and outlier data to prevent under- and over-generalization in models i.e., incorrectly identifying inliers as outliers and vice-versa. By identifying the space for specifying inliers and outliers, an anomaly detector that can effectively flag data of varying semantic complexities in medical imaging is next developed. Uncertainty quantification in deep learning models involves identifying sources of failure and characterizing model confidence to enable actionability. A training strategy is developed that allows the accurate estimation of model uncertainties and its benefits are demonstrated for active learning and generalization gap prediction. This helps identify insufficiently sampled regimes and representation insufficiency in models. In addition, the task of deep inversion under data scarce scenarios is considered, which in practice requires a prior to control the optimization. By identifying limitations in existing work, data priors powered by generative models and deep model priors are designed for audio restoration. With relevant empirical studies on a variety of benchmarks, the need for such design strategies is demonstrated.
ContributorsNarayanaswamy, Vivek Sivaraman (Author) / Spanias, Andreas (Thesis advisor) / J. Thiagarajan, Jayaraman (Committee member) / Berisha, Visar (Committee member) / Tepedelenlioğlu, Cihan (Committee member) / Arizona State University (Publisher)
Created2023
Description
This study introduces a new outdoor accelerated testing method called “Field Accelerated Stress Testing (FAST)” for photovoltaic (PV) modules performed at two
different climatic sites in Arizona (hot-dry) and Florida (hot-humid). FAST is a combined
accelerated test methodology that simultaneously accounts for all the field-specific stresses
and accelerates only key stresses, such as temperature, to forecast the failure modes by 2-
7 times in advance depending on the activation energy of the degradation mechanism (i.e.,
10th year reliability issues can potentially be predicted in the 2nd year itself for an
acceleration factor of 5). In this outdoor combined accelerated stress study, the
temperatures of test modules were increased (by 16-19℃ compared to control modules)
using thermal insulations on the back of the modules. All other conditions (ambient
temperature, humidity, natural sunlight, wind speed, wind direction, and tilt angle) were
left constant for both test modules (with back thermal insulation) and control modules
(without thermal insulation). In this study, a total of sixteen 4-cell modules with two different construction types (glass/glass [GG] and glass/backsheet [GB]) and two different encapsulant types (ethylene
vinyl acetate [EVA] and polyolefin elastomer [POE]), were investigated at both sites with
eight modules at each site (four insulated and four non-insulated modules at each site). All
the modules were extensively characterized before installation in the field and after field
exposure over two years. The methods used for characterizing the devices included I-V
(current-voltage curves), EL (electroluminescence), UVF (ultraviolet fluorescence), and
reflectance. The key findings of this study are: i) the GG modules tend to operate at a higher temperature (1-3℃) than the GB modules at both sites of Arizona and Florida (a lower
lifetime is expected for GG modules compared to GB modules); ii) the GG modules tend
to experience a higher level of encapsulant discoloration and grid finger degradation than
the GB modules at both sites (a higher level of the degradation rate is expected in GG
modules compared to GB modules); and, iii) the EVA-based modules tend to have a higher
level of discoloration and finger degradation compared to the POE-based modules at both
sites.
ContributorsThayumanavan, Rishi Gokul (Author) / Tamizhmani, Govindasamy (Thesis advisor) / Phelan, Patrick (Thesis advisor) / Calhoun, Ronald (Committee member) / Arizona State University (Publisher)
Created2023
Description
Stress-related failure such as cracking are an important photovoltaic (PV) reliability issue since it accounts for a high percentage of power losses in the midlife-failure and wear-out failure regimes. Cell cracking can only be correlated with module degradation when cracks are of detectable size and detrimental to the performance. Several techniques have been explored to access the deflection and stress status on solar cell, but they have disadvantages such as high surface sensitivity.
This dissertation presents a new and non-destructive method for mapping the deflection on encapsulated solar cells using X-ray topography (XRT). This method is based on Bragg diffraction imaging, where only the areas that meet diffraction conditions will present contrast. By taking XRT images of the solar cell at various sample positions and applying an in-house developed algorithm framework, the cell‘s deflection map is obtained. Error analysis has demonstrated that the errors from the experiment and the data processing are below 4.4 and 3.3%.
Von Karman plate theory has been applied to access the stress state of the solar cells. Under the assumptions that the samples experience pure bending and plain stress conditions, the principal stresses are obtained from the cell deflection data. Results from a statistical analysis using a Weibull distribution suggest that 0.1% of the data points can contribute to critical failure. Both the soldering and lamination processes put large amounts of stress on solar cells. Even though glass/glass packaging symmetry is preferred over glass/backsheet, the solar cells inside the glass/glass packaging experience significantly more stress. Through a series of in-situ four-point bending test, the assumptions behind Von Karman theory are validated for cases where the neutral plane is displaced by the tensile and compressive stresses.
The deflection and stress mapping method is applied to two next generation PV concepts named Flex-circuit and PVMirror. The Flex-circuit module concept replaces traditional metal ribbons with Al foils for electrical contact and PVMirror concept utilizes a curved PV module design with a dichroic film for thermal storage and electrical output. The XRT framework proposed in this dissertation successfully characterized the impact of various novel interconnection and packaging solutions.
This dissertation presents a new and non-destructive method for mapping the deflection on encapsulated solar cells using X-ray topography (XRT). This method is based on Bragg diffraction imaging, where only the areas that meet diffraction conditions will present contrast. By taking XRT images of the solar cell at various sample positions and applying an in-house developed algorithm framework, the cell‘s deflection map is obtained. Error analysis has demonstrated that the errors from the experiment and the data processing are below 4.4 and 3.3%.
Von Karman plate theory has been applied to access the stress state of the solar cells. Under the assumptions that the samples experience pure bending and plain stress conditions, the principal stresses are obtained from the cell deflection data. Results from a statistical analysis using a Weibull distribution suggest that 0.1% of the data points can contribute to critical failure. Both the soldering and lamination processes put large amounts of stress on solar cells. Even though glass/glass packaging symmetry is preferred over glass/backsheet, the solar cells inside the glass/glass packaging experience significantly more stress. Through a series of in-situ four-point bending test, the assumptions behind Von Karman theory are validated for cases where the neutral plane is displaced by the tensile and compressive stresses.
The deflection and stress mapping method is applied to two next generation PV concepts named Flex-circuit and PVMirror. The Flex-circuit module concept replaces traditional metal ribbons with Al foils for electrical contact and PVMirror concept utilizes a curved PV module design with a dichroic film for thermal storage and electrical output. The XRT framework proposed in this dissertation successfully characterized the impact of various novel interconnection and packaging solutions.
ContributorsMeng, Xiaodong (Author) / Bertoni, Marian I (Thesis advisor) / Meier, Rico (Committee member) / Holman, Zachary C (Committee member) / Alford, Terry (Committee member) / Arizona State University (Publisher)
Created2019
Description
An ongoing effort in the photovoltaic (PV) industry is to reduce the major manufacturing cost components of solar cells, the great majority of which are based on crystalline silicon (c-Si). This includes the substitution of screenprinted silver (Ag) cell contacts with alternative copper (Cu)-based contacts, usually applied with plating. Plated Cu contact schemes have been under study for many years with only minor traction in industrial production. One of the more commonly-cited barriers to the adoption of Cu-based contacts for photovoltaics is long-term reliability, as Cu is a significant contaminant in c-Si, forming precipitates that degrade performance via degradation of diode character and reduction of minority carrier lifetime. Cu contamination from contacts might cause degradation during field deployment if Cu is able to ingress into c-Si. Furthermore, Cu contamination is also known to cause a form of light-induced degradation (LID) which further degrades carrier lifetime when cells are exposed to light.
Prior literature on Cu-contact reliability tended to focus on accelerated testing at the cell and wafer level that may not be entirely replicative of real-world environmental stresses in PV modules. This thesis is aimed at advancing the understanding of Cu-contact reliability from the perspective of quasi-commercial modules under more realistic stresses. In this thesis, c-Si solar cells with Cu-plated contacts are fabricated, made into PV modules, and subjected to environmental stress in an attempt to induce hypothesized failure modes and understand any new vulnerabilities that Cu contacts might introduce. In particular, damp heat stress is applied to conventional, p-type c-Si modules and high efficiency, n-type c-Si heterojunction modules. I present evidence of Cu-induced diode degradation that also depends on PV module materials, as well as degradation unrelated to Cu, and in either case suggest engineering solutions to the observed degradation. In a forensic search for degradation mechanisms, I present novel evidence of Cu outdiffusion from contact layers and encapsulant-driven contact corrosion as potential key factors. Finally, outdoor exposures to light uncover peculiarities in Cu-plated samples, but do not point to especially serious vulnerabilities.
Prior literature on Cu-contact reliability tended to focus on accelerated testing at the cell and wafer level that may not be entirely replicative of real-world environmental stresses in PV modules. This thesis is aimed at advancing the understanding of Cu-contact reliability from the perspective of quasi-commercial modules under more realistic stresses. In this thesis, c-Si solar cells with Cu-plated contacts are fabricated, made into PV modules, and subjected to environmental stress in an attempt to induce hypothesized failure modes and understand any new vulnerabilities that Cu contacts might introduce. In particular, damp heat stress is applied to conventional, p-type c-Si modules and high efficiency, n-type c-Si heterojunction modules. I present evidence of Cu-induced diode degradation that also depends on PV module materials, as well as degradation unrelated to Cu, and in either case suggest engineering solutions to the observed degradation. In a forensic search for degradation mechanisms, I present novel evidence of Cu outdiffusion from contact layers and encapsulant-driven contact corrosion as potential key factors. Finally, outdoor exposures to light uncover peculiarities in Cu-plated samples, but do not point to especially serious vulnerabilities.
ContributorsKaras, Joseph (Author) / Bowden, Stuart (Thesis advisor) / Alford, Terry (Thesis advisor) / Tamizhmani, Govindasamy (Committee member) / Michaelson, Lynne (Committee member) / Arizona State University (Publisher)
Created2020
Description
Information exists in various forms and a better utilization of the available information can benefit the system awareness and response predictions. The focus of this dissertation is on the fusion of different types of information using Bayesian-Entropy method. The Maximum Entropy method in information theory introduces a unique way of handling information in the form of constraints. The Bayesian-Entropy (BE) principle is proposed to integrate the Bayes’ theorem and Maximum Entropy method to encode extra information. The posterior distribution in Bayesian-Entropy method has a Bayesian part to handle point observation data, and an Entropy part that encodes constraints, such as statistical moment information, range information and general function between variables. The proposed method is then extended to its network format as Bayesian Entropy Network (BEN), which serves as a generalized information fusion tool for diagnostics, prognostics, and surrogate modeling.
The proposed BEN is demonstrated and validated with extensive engineering applications. The BEN method is first demonstrated for diagnostics of gas pipelines and metal/composite plates for damage diagnostics. Both empirical knowledge and physics model are integrated with direct observations to improve the accuracy for diagnostics and to reduce the training samples. Next, the BEN is demonstrated in prognostics and safety assessment in air traffic management system. Various information types, such as human concepts, variable correlation functions, physical constraints, and tendency data, are fused in BEN to enhance the safety assessment and risk prediction in the National Airspace System (NAS). Following this, the BE principle is applied in surrogate modeling. Multiple algorithms are proposed based on different type of information encoding, such as Bayesian-Entropy Linear Regression (BELR), Bayesian-Entropy Semiparametric Gaussian Process (BESGP), and Bayesian-Entropy Gaussian Process (BEGP) are demonstrated with numerical toy problems and practical engineering analysis. The results show that the major benefits are the superior prediction/extrapolation performance and significant reduction of training samples by using additional physics/knowledge as constraints. The proposed BEN offers a systematic and rigorous way to incorporate various information sources. Several major conclusions are drawn based on the proposed study.
The proposed BEN is demonstrated and validated with extensive engineering applications. The BEN method is first demonstrated for diagnostics of gas pipelines and metal/composite plates for damage diagnostics. Both empirical knowledge and physics model are integrated with direct observations to improve the accuracy for diagnostics and to reduce the training samples. Next, the BEN is demonstrated in prognostics and safety assessment in air traffic management system. Various information types, such as human concepts, variable correlation functions, physical constraints, and tendency data, are fused in BEN to enhance the safety assessment and risk prediction in the National Airspace System (NAS). Following this, the BE principle is applied in surrogate modeling. Multiple algorithms are proposed based on different type of information encoding, such as Bayesian-Entropy Linear Regression (BELR), Bayesian-Entropy Semiparametric Gaussian Process (BESGP), and Bayesian-Entropy Gaussian Process (BEGP) are demonstrated with numerical toy problems and practical engineering analysis. The results show that the major benefits are the superior prediction/extrapolation performance and significant reduction of training samples by using additional physics/knowledge as constraints. The proposed BEN offers a systematic and rigorous way to incorporate various information sources. Several major conclusions are drawn based on the proposed study.
ContributorsWang, Yuhao (Author) / Liu, Yongming (Thesis advisor) / Chattopadhyay, Aditi (Committee member) / Mignolet, Marc (Committee member) / Yan, Hao (Committee member) / Ren, Yi (Committee member) / Arizona State University (Publisher)
Created2020
Description
Over the past decade, machine learning research has made great strides and significant impact in several fields. Its success is greatly attributed to the development of effective machine learning algorithms like deep neural networks (a.k.a. deep learning), availability of large-scale databases and access to specialized hardware like Graphic Processing Units. When designing and training machine learning systems, researchers often assume access to large quantities of data that capture different possible variations. Variations in the data is needed to incorporate desired invariance and robustness properties in the machine learning system, especially in the case of deep learning algorithms. However, it is very difficult to gather such data in a real-world setting. For example, in certain medical/healthcare applications, it is very challenging to have access to data from all possible scenarios or with the necessary amount of variations as required to train the system. Additionally, the over-parameterized and unconstrained nature of deep neural networks can cause them to be poorly trained and in many cases over-confident which, in turn, can hamper their reliability and generalizability. This dissertation is a compendium of my research efforts to address the above challenges. I propose building invariant feature representations by wedding concepts from topological data analysis and Riemannian geometry, that automatically incorporate the desired invariance properties for different computer vision applications. I discuss how deep learning can be used to address some of the common challenges faced when working with topological data analysis methods. I describe alternative learning strategies based on unsupervised learning and transfer learning to address issues like dataset shifts and limited training data. Finally, I discuss my preliminary work on applying simple orthogonal constraints on deep learning feature representations to help develop more reliable and better calibrated models.
ContributorsSom, Anirudh (Author) / Turaga, Pavan (Thesis advisor) / Krishnamurthi, Narayanan (Committee member) / Spanias, Andreas (Committee member) / Li, Baoxin (Committee member) / Arizona State University (Publisher)
Created2020
Description
The popularity of solar photovoltaic (PV) energy is growing across the globe with more than 500 GW installed in 2018 with a capacity of 640 GW in 2019. Improved PV module reliability minimizes the levelized cost of energy. Studying and accelerating encapsulant browning and solder bond degradation—two of the most commonly observed degradation modes in the field—in a lab requires replicating the stress conditions that induce the same field degradation modes in a controlled accelerated environment to reduce testing time.
Accelerated testing is vital in learning about the reliability of solar PV modules. The unique streamlined approach taken saves time and resources with a statistically significant number of samples being tested in one chamber under multiple experimental stress conditions that closely mirror field conditions that induce encapsulant browning and solder bond degradation. With short circuit current (Isc) and series resistance (Rs) degradation data sets at multiple temperatures, the activation energies (Ea) for encapsulant browning and solder bond degradation was calculated.
Regular degradation was replaced by the wear-out stages of encapsulant browning and solder bond degradation by subjecting two types of field-aged modules to further accelerated testing. For browning, the Ea calculated through the Arrhenius model was 0.37 ± 0.17 eV and 0.71 ± 0.07 eV. For solder bond degradation, the Arrhenius model was used to calculate an Ea of 0.12 ± 0.05 eV for solder with 2wt% Ag and 0.35 ± 0.04 eV for Sn60Pb40 solder.
To study the effect of types of encapsulant, backsheet, and solder on encapsulant browning and solder bond degradation, 9-cut-cell samples maximizing available data points while minimizing resources underwent accelerated tests described for modules. A ring-like browning feature was observed in samples with UV pass EVA above and UV cut EVA below the cells. The backsheet permeability influences the extent of oxygen photo-bleaching. In samples with solder bond degradation, increased bright spots and cell darkening resulted in increased Rs. Combining image processing with fluorescence imaging and electroluminescence imaging would yield great insight into the two degradation modes.
Accelerated testing is vital in learning about the reliability of solar PV modules. The unique streamlined approach taken saves time and resources with a statistically significant number of samples being tested in one chamber under multiple experimental stress conditions that closely mirror field conditions that induce encapsulant browning and solder bond degradation. With short circuit current (Isc) and series resistance (Rs) degradation data sets at multiple temperatures, the activation energies (Ea) for encapsulant browning and solder bond degradation was calculated.
Regular degradation was replaced by the wear-out stages of encapsulant browning and solder bond degradation by subjecting two types of field-aged modules to further accelerated testing. For browning, the Ea calculated through the Arrhenius model was 0.37 ± 0.17 eV and 0.71 ± 0.07 eV. For solder bond degradation, the Arrhenius model was used to calculate an Ea of 0.12 ± 0.05 eV for solder with 2wt% Ag and 0.35 ± 0.04 eV for Sn60Pb40 solder.
To study the effect of types of encapsulant, backsheet, and solder on encapsulant browning and solder bond degradation, 9-cut-cell samples maximizing available data points while minimizing resources underwent accelerated tests described for modules. A ring-like browning feature was observed in samples with UV pass EVA above and UV cut EVA below the cells. The backsheet permeability influences the extent of oxygen photo-bleaching. In samples with solder bond degradation, increased bright spots and cell darkening resulted in increased Rs. Combining image processing with fluorescence imaging and electroluminescence imaging would yield great insight into the two degradation modes.
ContributorsGopalakrishna, Hamsini (Author) / Tamizhmani, Govindasamy (Thesis advisor) / Rogers, Bradley (Committee member) / Hacke, Peter (Committee member) / Arizona State University (Publisher)
Created2020