538 research outputs found
Electrostatic monitoring of wind turbine gearbox on oil-lubricated system
No full textThe electrostatic sensing technique has been verified to be a viable method for tribo-contact monitoring under laboratory conditions in previous investigations. This paper reports on the evolution of electrostatic monitoring on a real oil-lubricated wind turbine gearbox, using a modified oil-line sensor. In a nominal test and a ramp-up test, features were extracted and the presence of debris can be detected. The permutation entropy was further introduced in an accelerated life test. It can accurately reflect the wear condition of the gearboxes and detect early faults earlier than conventional techniques, which also has a better sensitivity and performance degradation trend than time-domain features
An inexact l<sub>2</sub>-norm penalty method for cardinality constrained portfolio optimization
No full textWe analyze and solve a single-period portfolio optimization problem with non-convex constraints, which address practical concerns of investment such as the active share weights of sectors and the number of stocks held in a portfolio. We reformulate the problem to simplify the computation and propose an inexact l2-norm penalty method to solve the problem.</p
RF micro-systems for 5G front-end signal processing
No full textThe fifth-generation (5G) communication has sparked great research interest in developing the next generation radio frequency (RF) front ends for more stringent requirements on performance, power consumption, and spectral utilization efficiency. More parallel RF bands and paths are added in the same form factor, along with which come more components and tighter integration. Designing portable systems faces the new challenge of reducing component size while still operating at RF, where the path attenuation is low and fading is readily manageable. Satisfactory size reduction is particularly difficult for passive components that rely on the principle of waveguiding and thus scale with electromagnetic (EM) wavelength at RF (typical ~10s cm). Hence, radical size reduction by several orders of magnitude can only be attained by resorting to a physical domain other than EM, namely acoustic waves with wavelengths 4~5 orders of magnitude smaller. In fact, acoustic devices at RF, such as surface or bulk acoustic wave devices, have been widely used for mobile phone applications. Other acoustic elements, such as couplers, correlators, and impedance matching networks, have also shown promising potential to outperform the state-of-the-art EM counterparts. However, the past developments often battled the challenge of efficiently accessing the acoustics over a sufficiently wide bandwidth and subsequently producing application-worthy performance, because of the fundamental limitations from the lack of high electromechanical coupling (k2) and low damping piezoelectric platforms. Recently, thanks to the advances in materials, design, and fabrication, LiNbO3 thin-film has been proved as a promising low-loss, wideband, and frequency-diverse acoustic platform for novel functions toward high-performance 5G front-end signal processing.
Based on acoustic devices in LiNbO3 thin films, this thesis aims to design and demonstrate several classes of novel RF microsystems that can enable conventional signal processing functions with better performance or new tasks for emerging applications. First, the acoustic systems are used as passive signal processing elements for the Internet of Things (IoT) applications. The high figure of merit (FoM) LiNbO3 resonator array is used as the impedance matching element for interfacing with the high impedance CMOS rectifiers in the IoT-inspired wake-up radio. The high FoM, adequately large static capacitance and spurious free performance collectively contribute to a high voltage gain over 20. Another type of microsystem for IoT applications is 1-dB IL acoustic delay lines (ADLs) on the S0 mode in thin-film LiNbO3, showing record-breaking low IL over a larger bandwidth, opening new horizons for low-power RF acoustic signal processing. Second, the miniature nonreciprocal component based on switched high-performance delay elements is demonstrated for full duplex radio. The wideband and long delay featured by the high-performance LiNbO3 ADLs significantly benefit the performance, including the dynamic switching induced IL and intra-modulations, while relaxing the requirements for synthesizing and synchronizing the control signals. The demonstrated 4 port circulator shows a highly symmetric performance across the 4-ports with 18.8 dB nonreciprocal contrast between the IL (6.6 dB) and isolation (25.4 dB) over an FBW of 8.8% at a center frequency 155 MHz, all of which are accomplished with a record low switching frequency of 877.22 kHz. Upon further optimizations, such circulators can potentially outperform ferrite-based devices in loss, bandwidth, and isolation while offering more compact size and reconfigurable operation. Third, low-loss wideband GHz S0 mode ADLs are explored for self-interference cancellation in full-duplex radio. The fabricated miniature acoustic delay lines show a fractional bandwidth of 4% and a minimum IL of 3.2 dB at a center frequency of 0.96 GHz. Various delays ranging from 20 ns to 900 ns have been obtained for digitally addressable delay synthesis. Multiple acoustic delay lines with center frequencies from 0.9 to 2 GHz have been demonstrated. The demonstrated ADLs can potentially provide wide-range and high-resolution reconfigurable delays for future SIC applications. Finally, design and measurement of 5 GHz antisymmetric mode acoustic delay lines for 5G enhanced mobile broadband (eMBB) applications are presented; the demonstrated ADLs significantly surpass the state of the art with similar feature sizes in center frequency. The implemented ADLs at 5 GHz show a minimum insertion loss of 7.94 dB and a fractional bandwidth around 4%. In addition to the remarkable device performance, these designs also point out the opportunities to advance the operation frequencies of acoustic devices toward the wideband and high-frequency signal processing functions required for future 5G applications.
RF acoustic microsystems demonstrated in this thesis have shown promising prospects for 5G front-end signal processing applications. Thanks to the simultaneously low damping and wideband performance at RF, acoustic devices based on LiNbO3 thin films are auspicious candidates to provide the design flexibilities and high performance required for various 5G application scenarios. Further development in high-performance RF acoustic devices may put on the horizon an RF front-end synthesized either purely or predominantly from an RF acoustic component kit.Submission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2021-12-01The student, Ruochen Lu, accepted the attached license on 2019-09-12 at 09:43.The student, Ruochen Lu, submitted this Dissertation for approval on 2019-09-12 at 10:03.This Dissertation was approved for publication on 2019-09-13 at 16:52.DSpace SAF Submission Ingestion Package generated from Vireo submission #14451 on 2020-02-28 at 17:35:17Made available in DSpace on 2020-03-02T22:38:37Z (GMT). No. of bitstreams: 4
LU-DISSERTATION-2019.pdf: 41135901 bytes, checksum: d53219a4453f36fb646113ee6c675878 (MD5)
20190912_thesis_version04_format_check_4.docx: 86478459 bytes, checksum: 9f010d92462f29b038d1c4cfc2694ffe (MD5)
LICENSE.txt: 4207 bytes, checksum: 1ad070f4f17db180272819881c06afdb (MD5)
PROQUEST_LICENSE.txt: 4553 bytes, checksum: 98e0eafa7b6c5cf3d66f91e35aef082c (MD5)
Previous issue date: 2019-09-13Embargo set by: Seth Robbins for item 113969
Lift date: 2022-03-02T22:39:04Z
Reason: Author requested closed access (OA after 2yrs) in Vireo ETD systemLimited Restriction Lifted for Item 113969 on 2022-03-03T10:15:08Z
Genetic characterization of moyamoya disease suggests the involvement of immune cells
No full textMoyamoya disease (MMD) is a rare cerebrovascular disease characterized by progressive stenosis of the internal carotid artery (ICA). Previous studies have found association of the gene variant RNF213 p.R4859K in East Asian population. However, other genetic factors involved in the disease and affected cell types remain to be discovered. In this thesis, I conducted genetic characterization of moyamoya disease by whole exome sequencing. In our cohort, 25 out of all patients were found to carry inherited RNF213 p.R4859K variants. Further analysis suggests that STAT1 is a potential transcription factor of RNF213 in endothelial cells. Analysis on rare germline SNPs suggests the involvement of the HLA system in the disease. Taking advantage of trio blood samples of the patients and their parents, a total of 106 de novo germline mutations were identified, with a validation rate of 91% by Sanger sequencing. Next, an in-house cell type gene enrichment analysis pipeline based on single cell RNA sequencing data from PanglaoDB was applied. It is found that red pulp macrophages followed by endothelial cells are most likely affected by our list of detected DNMs in coding regions. We also developed a network based pipeline for cell type enrichment analysis and the result suggests phagocytes as most likely affected by the DNMs which is compatible with previous results. I also constructed a pipeline for identifying compound heterozygous variants potentially contributing to the development of the disease. As a result, 5 genes were found to be affected by potentially pathogenic compound heterozygous variants. To summarize, this thesis studied genetic factors potentially associated with the disease. Interestingly, cell type gene enrichment analysis revealed a potential role of immune cells. This finding could revolutionize our understanding of MMD and the pipeline could help with studies of diseases with unknown cell types.</p
Thin-Film Lithium Niobate Acoustic Delay Line Oscillators
No full textIn this work, thin-film lithium niobate (LiNbO3) acoustic delay line (ADL) based oscillators are experimentally investigated for the first time for the application of single-mode oscillators and frequency comb generation. The design space for the ADL-based oscillator is first analyzed, illustrating that the key to low phase noise lies in high center frequency (fo), large delay (τ G), and low insertion loss (IL) of the delay. Therefore, two self-sustained oscillators employing low noise amplifiers (LNA) and a low IL, long delay (fo=157MHz, IL =2.9dB, τG= 200-440ns) SH0 mode ADLs are designed for a case study. The two SH0 ADL oscillators show measured phase noise of -109 dBc/Hz and -127 dBc/Hz at 10-kHz offset while consuming 16 mA and 48 mA supply currents, respectively. Although the carrier power of the proposed oscillator is lower than published state-of-the-art ADL oscillators, competitive phase noise performance is still attained thanks to the low IL. Finally, frequency comb generation is also demonstrated with the same delay line and a commercial RF feedback amplifier, showing a comb spacing of 3.4 MHz that matches the open-loop characterization.Accepted Author ManuscriptDynamics of Micro and Nano System
Recommended from our members
Towards efficient, effective, and robust Neural Architecture Search methods
No full textRecently, Neural Architecture Search (NAS) has attracted lots of attention for its potential to democratize deep learning. For a practical end-to-end deep learning platform, NAS plays a crucial role in discovering task-specific architecture depending on users' configurations (e.g., dataset, evaluation metric, etc.). Among various search paradigms, Differentiable Neural Architecture Search is one of the most popular NAS methods for its search efficiency and simplicity, accomplished by jointly optimizing the model weight and architecture parameters in a weight-sharing supernet via gradient-based algorithms. At the end of the search phase, the operations with the largest architecture parameters will be selected to form the final architecture, with the implicit assumption that the values of architecture parameters reflect the operation strength. Despite the search efficiency, the weight-sharing supernet also shows a tendency towards non-parametric operations, resulting in shallow architectures with degenerated performance. We provide both theoretical and empirical analysis of the poor generalization observed in Differentiable NAS, which links this issue to the failure of the magnitude-based selection. Following this inspiration, we discuss two lines of methods that greatly improve the effectiveness and robustness of Differentiable NAS: The first line proposes an alternative perturbation-based architecture selection that is shown to identify better architectures in the search space, whereas the second line aligns the architecture parameter with the strength of underlying operations. To complete the picture, an alternative paradigm to the differential architecture search (predictor-based NAS) is also presented
Study of Mode Switch of the Hydraulically Interconnected Inerter-spring-damper Suspension System
No full textRecommended from our members
Automated Machine Learning in the Era of Large Foundation Models
No full textIntelligence, one of the most profound phenomena on Earth, has evolved over 600 million years, transforming from simple neural systems into human cognition capable of unraveling universal mysteries and creating silicon-based intelligence. This evolutionary process, with its inherent drive towards increasing complexity, seemingly defies thermodynamic principles, suggesting the existence of self-evolving mechanisms in life. Recreating such mechanisms within artificial systems is a crucial milestone on the path toward Artificial General Intelligence (AGI).Automated Machine Learning (AutoML) represents a significant step in this direction. By enabling AI systems to optimize their own design processes, AutoML reformulates machine learning pipeline construction as a search problem, automating the selection of architectures, optimizers, hyperparameters, and even reasoning paths from an expansive search space. Despite its current limitations, AutoML has already demonstrated remarkable success in advancing machine learning across diverse applications.This thesis delves into the synergistic interplay between AutoML and large foundation models, particularly the recent breakthroughs in large-scale generative models like large language models (LLMs) and diffusion models. These models exhibit emergent behaviors, highlighting machine learning systems as complex entities where scaling produces unpredictable and potentially transformative capabilities.We emphasize the application of AutoML in automating the design of training and inference processes, crucial for the continued advancement of these models.Ultimately, we aspire that this research contributes a meaningful step towards the ambitious pursuit of Artificial General Intelligence (AGI)
Targeted Priority Mechanisms in Organ Transplantation
No full textThe persistent shortage of transplantable organs, compounded by high rates of organ underutilization, necessitates innovative allocation mechanisms. This dissertation develops and analyzes targeted priority mechanisms, voluntary incentive-based programs designed to enhance access for disadvantaged patient groups and improve organ-recipient matching. Using a rigorous queueing-theoretic framework, I characterize patients' equilibrium participation strategies, identifying conditions under which no-, full-, and mixed-participation equilibria emerge. I further establish the necessary and sufficient conditions for their existence and uniqueness, highlighting how careful mechanism design can align individual incentives with socially optimal outcomes.
The study extends the analysis to class-separating allocations, demonstrating the feasibility of equilibria that improve social welfare while safeguarding non-participating patients' access to high-quality organs. A clinically detailed simulation of the U.S. kidney allocation system, focusing on elderly patients, illustrates the potential benefits: a targeted threshold of 84% KDPI yields approximately 220 additional annual transplants, reduces the waiting list by more than 450 patients, and prevents over 60 pre-transplant deaths annually, with minimal impact on graft survival rates. Overall, the findings provide both theoretical and practical guidance for the design of efficient, implementable allocation mechanisms.Doctor of PhilosophyEvery year, thousands of patients die while waiting for a life-saving organ transplant, even though many donated organs never get used. This research looks at how to make organ allocation fairer and more effective by introducing what are called targeted priority programs. These programs give certain groups of patients, such as older adults, earlier access to organs that are less likely to be used otherwise. In return, these patients give up priority for the highest-quality organs, making it easier to match every organ with the patient who can benefit most.
The study shows how patients might respond to such programs, and how the rules can be designed to encourage participation without disadvantaging those who are not eligible. A detailed computer model of the U.S. kidney transplant system suggests that, with the right design, targeted priority programs could prevent more than 60 deaths each year, reduce the waiting list by over 450 patients, and add about 220 extra transplants annually—all while keeping transplant success rates nearly the same.
Although no policy is perfect, this approach offers a promising way to make better use of donated organs and give more patients a second chance at life. By carefully choosing which patient groups to include and how to balance trade-offs, policymakers could create a system that is both more efficient and more equitable
Across the boundary: addressing segregation along transportation infrastructure
No full textAcross the United States, we can see examples of cities where highways and railways pass between two ethnic communities or through communities of color and are used to further divide and segregate cities. Increasingly we are seeing a new typology of landscape architecture projects to transform, redesign, or remove transportation infrastructure to help ameliorate the negative impact of transportation infrastructure on neighboring communities and support public uses. However, these projects often lead to gentrification, whereby surrounding housing prices rise, and the original residents are displaced or lose their sense of belonging. The goal of my research is to understand how landscape architects are redesigning transportation infrastructure to address divisions within cities, as well as possible ways to prevent and address issues of gentrification. Eventually making benefits to community members and enhances their understanding and communication between each other. After study of examples of landscape projects reusing transportation infrastructure in several North American cities, design principles, guidelines and toolkits are developed to guide the similar types of design. Finally, these are teste on a site of Providence city which has a long segregation history due to the construction of the I-95 corridor. New community gardens, terrace lawn, gathering plaza, small business programs are working together to make a strong link for the long divided surrounding neighborhoods
- …
