For more information visit:

Recent Submissions

  • Boundary-sensitive Pre-training for Temporal Localization in Videos

    Xu, Mengmeng; Perez-Rua, Juan-Manuel; Escorcia, Victor; Martinez, Brais; Zhu, Xiatian; Ghanem, Bernard; Xiang, Tao (arXiv, 2020-11-21) [Preprint]
    Many video analysis tasks require temporal localization thus detection of content changes. However, most existing models developed for these tasks are pre-trained on general video action classification tasks. This is because large scale annotation of temporal boundaries in untrimmed videos is expensive. Therefore no suitable datasets exist for temporal boundary-sensitive pre-training. In this paper for the first time, we investigate model pre-training for temporal localization by introducing a novel boundary-sensitive pretext (BSP) task. Instead of relying on costly manual annotations of temporal boundaries, we propose to synthesize temporal boundaries in existing video action classification datasets. With the synthesized boundaries, BSP can be simply conducted via classifying the boundary types. This enables the learning of video representations that are much more transferable to downstream temporal localization tasks. Extensive experiments show that the proposed BSP is superior and complementary to the existing action classification based pre-training counterpart, and achieves new state-of-the-art performance on several temporal localization tasks.
  • VLG-Net: Video-Language Graph Matching Network for Video Grounding

    Qu, Sisi; Soldan, Mattia; Xu, Mengmeng; Tegner, Jesper; Ghanem, Bernard (arXiv, 2020-11-19) [Preprint]
    Grounding language queries in videos aims at identifying the time interval (or moment) semantically relevant to a language query. The solution to this challenging task demands the understanding of videos' and queries' semantic content and the fine-grained reasoning about their multi-modal interactions. Our key idea is to recast this challenge into an algorithmic graph matching problem. Fueled by recent advances in Graph Neural Networks, we propose to leverage Graph Convolutional Networks to model video and textual information as well as their semantic alignment. To enable the mutual exchange of information across the domains, we design a novel Video-Language Graph Matching Network (VLG-Net) to match video and query graphs. Core ingredients include representation graphs, built on top of video snippets and query tokens separately, which are used for modeling the intra-modality relationships. A Graph Matching layer is adopted for cross-modal context modeling and multi-modal fusion. Finally, moment candidates are created using masked moment attention pooling by fusing the moment's enriched snippet features. We demonstrate superior performance over state-of-the-art grounding methods on three widely used datasets for temporal localization of moments in videos with natural language queries: ActivityNet-Captions, TACoS, and DiDeMo.
  • Single-Crystalline All-Oxide α–γ–β Heterostructures for Deep-Ultraviolet Photodetection

    Li, Kuang-Hui; Kang, Chun Hong; Min, Jung-Hong; Alfaraj, Nasir; Liang, Jian-Wei; Braic, Laurentiu; Guo, Zaibing; Hedhili, Mohamed N.; Ng, Tien Khee; Ooi, Boon S. (ACS Applied Materials & Interfaces, American Chemical Society (ACS), 2020-11-17) [Article]
    Recent advancements in gallium oxide (Ga2O3)-based heterostructures have allowed optoelectronic devices to be used extensively in the fields of power electronics and deep-ultraviolet photodetection. While most previous research has involved realizing single-crystalline Ga2O3 layers on native substrates for high conductivity and visible-light transparency, presented and investigated herein is a single-crystalline β-Ga2O3 layer grown on an α-Al2O3 substrate through an interfacial γ-In2O3 layer. The single-crystalline transparent conductive oxide layer made of wafer-scalable γ-In2O3 provides high carrier transport, visible-light transparency, and antioxidation properties that are critical for realizing vertically oriented heterostructures for transparent oxide photonic platforms. Physical characterization based on X-ray diffraction and high-resolution transmission electron microscopy imaging confirms the single-crystalline nature of the grown films and the crystallographic orientation relationships among the monoclinic β-Ga2O3, cubic γ-In2O3, and trigonal α-Al2O3, while the elemental composition and sharp interfaces across the heterostructure are confirmed by Rutherford backscattering spectrometry. Furthermore, the energy-band offsets are determined by X-ray photoelectron spectroscopy at the β-Ga2O3/γ-In2O3 interface, elucidating a type-II heterojunction with conduction- and valence-band offsets of 0.16 and 1.38 eV, respectively. Based on the single-crystalline β-Ga2O3/γ-In2O3/α-Al2O3 all-oxide heterostructure, a vertically oriented DUV photodetector is fabricated that exhibits a high photoresponsivity of 94.3 A/W, an external quantum efficiency of 4.6 × 104%, and a specific detectivity of 3.09 × 1012 Jones at 250 nm. The present demonstration lays a strong foundation for and paves the way to future all-oxide-based transparent photonic platforms.
  • Paper as a Substrate and an Active Material in Paper Electronics

    Khan, Sherjeel M.; Nassar, Joanna M.; Hussain, Muhammad Mustafa (ACS Applied Electronic Materials, American Chemical Society (ACS), 2020-11-16) [Article]
    Paper is an essential part of our daily life in many different ways. It is made by compressing cellulose fibers sourced from wood into thin sheets. Paper is an inherently flexible material which can transport liquids through its medium by capillary action without the need of external force. The mesh network of cellulose in paper gives it a unique set of mechanical properties. Owing to its exclusive and advantageous properties, paper is being used as an active material and a substrate in electronics. Paper as an active material means that paper is utilized in its intrinsic form without modifications. Activated (or functionalized) paper has been widely exploited in many applications, but in order to take true advantage of all the beneficial properties of paper, it needs to be used in its natural produced form. Notably, paper is employed in humidity sensors, pressure sensors, and MEMS devices in its natural form. Additionally, paper is used as a substrate in additively manufactured and origami-inspired electronic devices. Here, we present an overview of how paper is used to make fully flexible and low-cost devices. Furthermore, the emergence of paper-based point-of-care devices is briefly discussed.
  • Performance Characterization of High and Low Power Prism based Tunable Blue Laser Diodes Systems

    Mukhtar, Sani; Holguin Lerma, Jorge Alberto; Ashry, Islam; Ng, Tien Khee; Ooi, Boon S.; Khan, M. Z. M. (IEEE, 2020-11-13) [Conference Paper]
    Comparison of high- and low-power tunable external-cavity blue laser-diode system demonstrates a tunability of 10.6 and 4nm, respectively, with a corresponding SMSR as high as 35 and 32dB and linewidth as low as 97 and 59pm, while showcasing high stability at extreme operating conditions.
  • A highly sensitive, large area, and self-powered UV photodetector based on coalesced gallium nitride nanorods/graphene/silicon (111) heterostructure

    Zulkifli, Nur 'Adnin Akmar; Park, Kwangwook; Min, Jung-Wook; Ooi, Boon S.; Zakaria, Rozalina; Kim, Jongmin; Tan, Chee Leong (Applied Physics Letters, AIP Publishing, 2020-11-13) [Article]
    In this paper, we demonstrate an ultraviolet photodetector (UV-PD) that uses coalesced gallium nitride (GaN) nanorods (NRs) on a graphene/Si (111) substrate grown by plasma-assisted molecular beam epitaxy. We report a highly sensitive, self-powered, and hybrid GaN NR/graphene/Si (111) PD with a relatively large 100 mm2 active area, a high responsivity of 17.4 A/W, a high specific detectivity of 1.23 1013 Jones, and fast response speeds of 13.2/13.7 ls (20 kHz) under a UV light of 355 nm at zero bias voltage. The results show that the thin graphene acts as a perfect interface for GaN NRs, encouraging growth with minimum defects on the Si substrate. Our results suggest that the GaN NR/graphene/Si (111) heterojunction has a range of interesting properties that make it well-suited for a variety of photodetection applications.
  • Using deep neural networks to diagnose engine pre-ignition

    Kuzhagaliyeva, Nursulu; Thabet, Ali; Singh, Eshan; Ghanem, Bernard; Sarathy, Mani (Proceedings of the Combustion Institute, Elsevier BV, 2020-11-12) [Article]
    Engine downsizing and boosting have been recognized as effective strategies for improving engine efficiency. However, operating the engines at high load promotes abnormal combustion events, such as pre-ignition and potential superknock. Currently the most effective method for detecting pre-ignition is by using in-cylinder pressure sensors that have high precision and sensitivity, but also high cost. Due to rapid advances in automotive technology such as autonomous driving, computer-aided designs and future connectivity, we propose to use a complimentary data-driven strategy for diagnosing abnormal combustion events. To this end, a data-driven diagnostics approach for pre-ignition detection with deep neural networks is proposed. The success of convolutional neural networks (CNNs) in object detection and recurrent neural networks (RNNs) in sequence forecasting inspired us to develop these models for pre-ignition detection. For a cost-effective strategy, we use data from less expensive sensors, such as lambda and low-resolution exhaust back pressure (EBP), instead of high resolution in-cylinder pressure measurements. The first deep learning model is combined with a commonly used dimensionality reduction tool–Principal Component Analysis (PCA). The second model eliminates this step and directly processes time-series data. Results indicate that the first model with reduced input dimensions, and correspondingly smaller size of the network, shows better performance in detecting pre-ignition cycles with an F1 score of 79%. Overall, the proposed deep learning approach is a promising alternative for abnormal combustion diagnostics using data from low resolution sensors.
  • NIST-certified secure key generation via deep learning of physical unclonable functions in silica aerogels

    Fratalocchi, Andrea; Fleming, Adam; Conti, Claudio; Di Falco, Andrea (Nanophotonics, Walter de Gruyter GmbH, 2020-11-03) [Article]
    AbstractPhysical unclonable functions (PUFs) are complex physical objects that aim at overcoming the vulnerabilities of traditional cryptographic keys, promising a robust class of security primitives for different applications. Optical PUFs present advantages over traditional electronic realizations, namely, a stronger unclonability, but suffer from problems of reliability and weak unpredictability of the key. We here develop a two-step PUF generation strategy based on deep learning, which associates reliable keys verified against the National Institute of Standards and Technology (NIST) certification standards of true random generators for cryptography. The idea explored in this work is to decouple the design of the PUFs from the key generation and train a neural architecture to learn the mapping algorithm between the key and the PUF. We report experimental results with all-optical PUFs realized in silica aerogels and analyzed a population of 100 generated keys, each of 10,000 bit length. The key generated passed all tests required by the NIST standard, with proportion outcomes well beyond the NIST’s recommended threshold. The two-step key generation strategy studied in this work can be generalized to any PUF based on either optical or electronic implementations. It can help the design of robust PUFs for both secure authentications and encrypted communications.
  • Millimeter-Wave Antenna Array Diagnosis with Partial Channel State Information

    Medina, George; Jida, Akashdeep Singh; Pulipati, Sravan; Talwar, Rohith; J, Nancy Amala; Al-Naffouri, Tareq Y.; Madanayake, Arjuna; Eltayeb, Mohammed (arXiv, 2020-11-02) [Preprint]
    Large antenna arrays enable directional precoding for Millimeter-Wave (mmWave) systems and provide sufficient link budget to combat the high path-loss at these frequencies. Due to atmospheric conditions and hardware malfunction, outdoor mmWave antenna arrays are prone to blockages or complete failures. This results in a modified array geometry, distorted far-field radiation pattern, and system performance degradation. Recent remote array diagnostic techniques have emerged as an effective way to detect defective antenna elements in an array with few diagnostic measurements. These techniques, however, require full and perfect channel state information (CSI), which can be challenging to acquire in the presence of antenna faults. This paper proposes a new remote array diagnosis technique that relaxes the need for full CSI and only requires knowledge of the incident angle-of-arrivals, i.e. partial channel knowledge. Numerical results demonstrate the effectiveness of the proposed technique and show that fault detection can be obtained with comparable number of diagnostic measurements required by diagnostic techniques based on full channel knowledge. In presence of channel estimation errors, the proposed technique is shown to out-perform recently proposed array diagnostic techniques.
  • A Novel HVDC Architecture for Offshore Wind Farm Applications

    Dezem Bertozzi Junior, Otávio José (2020-11) [Thesis]
    Advisor: Ahmed, Shehab
    Committee members: Salama, Khaled N.; Lima, Ricardo
    The increasing global participation of wind power in the overall generation ca- pacity makes it one of the most promising renewable resources. Advances in power electronics have enabled this market growth and penetration. Through a literature review, this work explores the challenges and opportunities presented by offshore wind farms, as well as the different solutions proposed concerning power electron- ics converters, collection and transmission schemes, as well as control and protection techniques. A novel power converter solution for the parallel connection of high power offshore wind turbines, suitable for HVDC collection and transmission, is presented. For the parallel operation of energy sources in an HVDC grid, DC link voltage con- trol is required. The proposed system is based on a full-power rated uncontrolled diode bridge rectifier in series with a partially-rated fully-controlled thyristor bridge rectifier. The thyristor bridge acts as a voltage regulator to ensure the flow of the desired current through each branch, where a reactor is placed in series for filtering of the DC current. AC filters are installed on the machine side to mitigate harmonic content. The mathematical modeling of the system is derived and the control design procedure is discussed. Guidelines for equipment and device specifications are pre- sented. Different setups for an experimental framework are suggested and discussed, including a conceptual application for hardware-in-the-loop real-time simulation and testing.
  • Control Theoretic Approaches to Computational Modeling and Risk Mitigation for Large Crowd Management

    Alrashed, Mohammed (2020-11) [Dissertation]
    Advisor: Shamma, Jeff S.
    Committee members: Shamma, Jeff S.; Feron, Eric; Knio, Omar; Bamieh, Bassam
    We develop a computational framework for risk mitigation in high population density events. With increased global population, the frequency of high population density events is naturally increased. Therefore, risk-free crowd management plans are critical for efficient mobility, convenient daily life, resource management and most importantly mitigation of any inadvertent incidents and accidents such as stampedes. The status-quo for crowd management plans is the use of human experience/expert advice. However, most often such dependency on human experience is insufficient, flawed and results in inconvenience and tragic events. Motivated by these issues, we propose an agent-based mathematical model describing realistic human motion and simulating large dense crowds in a wide variety of events as a potential simulation testbed to trial crowd management plans. The developed model incorporates stylized mindset characteristics as an internal drive for physical behavior such as walking, running, and pushing. Furthermore, the model is combined with a visualisation of crowd movement. We develop analytic tools to quantify crowd dynamic features. The analytic tools will enable verification and validation to empirical evidence and surveillance video feed in both local and holistic representations of the crowd. This work addresses research problems in computational modeling of crowd dynamics, specifically: understanding and modeling the impact of a collective mindset on crowd dynamics versus mixtures of heterogeneous mindsets, the effect of social contagion of behaviors and decisions within the crowd, the competitive and aggressive pushing behaviors, and torso and steering dynamics.
  • Design and topological optimization of nanophotonic devices

    Lin, Ronghui (2020-11) [Dissertation]
    Advisor: Li, Xiaohang
    Committee members: Li, Xiaohang; Fratalocchi, Andrea; Liberale, Carlo; Lu, Tien-Chang
    A central topic in the research of nanophotonics is the geometrical optimization of the nanostructures since the geometries are deeply related to the Mie resonances and the localized surface plasmon resonances in dielectric and metallic nanomaterials. When many nanostructures are assembled to form a metamaterial, the tuning of the geometrical parameters can bring even more profound effects, such as bound states in the continuum (BIC) with infinite quality factors (Q factors). Moreover, with the development of nanofabrication technologies, there is a trend of integrating nanostructures in the vertical direction, which provides more degrees of freedom for controlling the device performance and functionality. The main topic of this dissertation is to explore some of the abovementioned tuning possibilities to enhance the performance of nanophotonic devices. The dissertation contains two major parts: In chapters 2 and 3, the vertical integration of metalenses is studied. We discover a phenomenon similar to the Moiré effect in the bilayer Pancharatnam-Berry phase metalenses and reveal the role of geometrical imperfections on the focusing performance of reflective metalenses. Novel multifocal and reflective metalenses, with smaller footprints and enhanced performance compared to their bulky conventional counterparts, are designed based on the theoretical findings. The study of geometrical imperfections also provides guidelines for analyzing and compensating the fabrication errors, which is vital for large scale production and commercialization of metalenses. In chapters 4 and 5, we use machine learning to harness the full tuning power of the complicated geometries, which is challenging with conventional design methods. Plasmonic metasurfaces with on-demand optical responses are designed by manipulating the coupling of multiple nanodisks using neural networks. An accuracy of ± 8 nm is achieved, which is higher than previous reports and close to the fabrication limits of nanofabrication technologies. We also demonstrate, for the first time, the control of multiple BIC states using freeform geometries with predefined symmetry. It is a new method to exploit the untapped potential of freeform photonics structures. The discoveries we have made in both dielectric and plasmonic nanophotonic devices could benefit applications such as imaging, sensing, and light-emitting devices.
  • Micro-electromechanical Resonator-based Logic and Interface Circuits for Low Power Applications

    Ahmed, Sally (2020-11) [Dissertation]
    Advisor: Fariborzi, Hossein
    Committee members: Fariborzi, Hossein; Shamim, Atif; Younis, Mohammad I.; Weinstein, Dana
    The notion of mechanical computation has been revived in the past few years, with the advances of nanofabrication techniques. Although electromechanical devices are inherently slow, they offer zero or very low off-state current, which reduces the overall power consumption compared to the fast complementary-metal-oxide-semiconductor (CMOS) counterparts. This energy efficiency feature is the most crucial requirement for most of the stand-alone battery-operated gadgets, biomedical devices, and the internet of things (IoT) applications, which do not require the fast processing speeds offered by the mainstream CMOS technology. In particular, using Micro-Electro-Mechanical (MEM) resonators in mechanical computing has drawn the attention of the research community and the industry in the last decade as this technology offers low power consumption, reduced circuit complexity compared to conventional CMOS designs, run-time re- programmability and high reliability due to the contactless mode of operation compared to other MEM switches such as micro-relays. In this thesis, we introduce digital circuit design techniques tailored for clamped-clamped beam MEM resonators. The main operation mechanism of these circuit blocks is based on fine-tuning of the resonance frequency of the micro-resonator beam, and the logic function performed by the devices is mainly determined by factors such as input/output terminal arrangement, signal type, resonator operation regime (linear/non-linear), and the operation frequency. These proposed circuits include the major building blocks of any microprocessor such as logic gates, a full adder which is a key block in any arithmetic and logic operation units (ALU), and I/O interface units, including digital to analog (DAC) and analog to digital (ADC) data converters. All proposed designs were first simulated using a finite element software and then the results were experimentally verified. Important aspects such as energy per operation, speed, and circuit complexity are evaluated and compared to CMOS counterparts. In all applications, we show that by proper scaling of the resonator’s dimensions, MHz operation speeds and energy consumption in the range of femto-joules per logic operation are attainable. Finally, we discuss some of the challenges in using MEM resonators in digital circuit design at the device level and circuit level and propose solutions to tackle some of them.
  • UAV Enabled IoT Network Designs for Enhanced Estimation, Detection, and Connectivity

    Bushnaq, Osama (2020-11) [Dissertation]
    Advisor: Al-Naffouri, Tareq Y.
    Committee members: Al-Naffouri, Tareq Y.; Alouini, Mohamed-Slim; Shamma, Jeff S.; Shihada, Basem; Gesbert, David
    The Internet of Things (IoT) is a foundational building block for the upcoming information revolution. Particularly, the IoT bridges the cyber domain to anything within our physical world which enables unprecedented monitoring, connectivity, and smart control. The utilization of Unmanned Aerial Vehicles (UAVs) can offer an extra level of flexibility which results in more advanced and efficient connectivity and data aggregation. In the first part of the thesis, we focus on the optimal IoT devices placement and, the spectral and energy budgets management for accurate source estimation. Practical aspects such as measurement accuracy, communication quality, and energy harvesting are considered. The problem is formed such that a set of cheap and expensive sensors are placed to minimize the estimation error under limited system cost. The IoT revolution relies on aggregating big data from massive numbers of devices that are widely scattered in our environment. These devices are expected to be of low- complexity, low-cost, and limited power supply, which impose stringent constraints on the network operation. Aerial data transmission offers strong line-of-sight links and flexible/instant deployment. The UAV-enabled IoT networks can, for instance, offer solutions to avoid and manage natural disasters such as forest fire. We investigate in this thesis the aerial data aggregation for field estimation, wildfire detection, and connection coverage enhancement via UAVs. To accomplish the network task, the field of interest is divided into several subregions over which the UAVs hover to collect samples from the underlying nodes. To this end, we formulate and solve optimization problems to minimize total hovering and traveling times. This goal is fulfilled by optimizing the UAV hovering locations, the hovering time at each location, and the trajectory traversed between hovering locations. Finally, we propose the utilization of the tethered UAV (T-UAV) to assist the terrestrial network, where the tether provides power supply and connects the T-UAV to the core network through a high capacity link. The T-UAV however has limited mobility due to the limited tether length. A stochastic geometry-based analysis is provided for the optimal coverage probability of T-UAV-assisted cellular networks.
  • 3D Near Isotropic Antenna in Package for IoT Applications

    Su, Zhen (2020-11) [Dissertation]
    Advisor: Shamim, Atif
    Committee members: Shamim, Atif; Bagci, Hakan; Wu, Ying; Wu, Ke
    Internet of Things (IoT) is an emerging paradigm about building a massive internet to link billions of non-living things to make smart decisions for humans and improve their quality of life. For many of IoT devices, such as wireless sensor nodes dispersed in the environment, there is not much control over their placements or orientations. Thus, there is a need to develop orientation insensitive antennas that ensure reliable data transmission irrespective of devices’ positions or orientations. As billions of such IoT devices required in the future, a low-cost fabrication process suitable for mass manufacturing must be adapted. Antenna in package (AiP) concept is beneficial that the package is utilized to realize the antennas, not only saving space but also reducing the overall cost. For orientation insensitivity, antennas must be near isotropic and even have to maintain their radiation pattern for multi-bands or wide bandwidths in most applications. However, there is a dearth in the literature about design methodologies for near isotropic antennas, particularly for multi-bands near isotropic AiP designs. In addition, a near isotropic behavior is also important for polarization, particularly for CP antennas. To have simultaneous isotropy in radiation pattern and circular polarization is challenging. In the nut shell, this thesis presents theoretical models and derives conditions for wire AiP design for different specifications, single-band and dual-band near isotropic antennas, null free near isotropic antenna with wide CP coverage, and a full CP antenna with decent near isotropy (with very narrow null beam). The single-band AiP has only 5.05 dB gain variation at WiFi/BLE band and the dual-band AiP has a decent near isotropic radiation property and covers both GSM900 and GSM1800 bands. The theoretical model for null-free near isotropic antenna with wide CP coverage is presented with particle swarm optimization (PSO). The full CP antenna has a measured CP coverage of 70% with a small null in the radiation pattern. The results are promising and indicate that the conditions and methods proposed are useful for the future near isotropic AiP design. Also, this work provides designers flexibility to adjust the AiP design according to their own applications.
  • A Closer Look at Neighborhoods in Graph Based Point Cloud Scene Semantic Segmentation Networks

    Itani, Hani (2020-11) [Thesis]
    Advisor: Ghanem, Bernard
    Committee members: Ghanem, Bernard; Al-Naffouri, Tareq Y.; Wonka, Peter; Thabet, Ali K.
    Large scale semantic segmentation is considered as one of the fundamental tasks in 3D scene understanding. Point clouds provide a basic and rich geometric rep- resentation of scenes and tangible objects. Convolutional Neural Networks (CNNs) have demonstrated an impressive success in processing regular discrete data such as 2D images and 1D audio. However, CNNs do not directly generalize to point cloud processing due to their irregular and un-ordered nature. One way to extend CNNs to point cloud understanding is to derive an intermediate euclidean representation of a point cloud by projecting onto image domain, voxelizing, or treating points as vertices of an un-directed graph. Graph-CNNs (GCNs) have demonstrated to be a very promising solution for deep learning on irregular data such as social networks, bi- ological systems, and recently point clouds. Early works in literature for graph based point networks relied on constructing dynamic graphs in the node feature space to define a convolution kernel. Later works constructed hierarchical static graphs in 3D space for an encoder-decoder framework inspired from image segmentation. This thesis takes a closer look at both dynamic and static graph neighborhoods of graph- based point networks for the task of semantic segmentation in order to: 1) discuss a potential cause for why going deep in dynamic GCNs does not necessarily lead to an improved performance, and 2) propose a new approach in treating points in a static graph neighborhood for an improved information aggregation. The proposed method leads to an efficient graph based 3D semantic segmentation network that is on par with current state-of-the-art methods on both indoor and outdoor scene semantic segmentation benchmarks such as S3DIS and Semantic3D.
  • Investigation of a New Voltage Balancing Circuit for Parallel-connected Offshore PMSG-based Wind Turbines

    Elserougi, Ahmed A.; Bertozzi, Otavio; Massoud, Ahmed M.; Ahmed, Shehab (IEEE, 2020-10-30) [Conference Paper]
    Parallel connection of wind turbines (WTs) is beneficial in high-power applications. For successful operation of parallel WT-based energy conversion systems, a well-regulated voltage is needed at the collection point. Due to wind speed variation, the generated voltage from each WT may differ from one to another. Conventional solutions use regulating converters with full power rating. In this paper, a new concept is presented which depends on using fully-rated uncontrolled rectifier bridges for AC-DC conversion, and partially-rated fully-controlled bridge rectifiers which are used as voltage tuners to guarantee the flow of desired maximum power point DC currents through the parallel connected branches. The proposed system is simple, cost effective, reliable and efficient. The main drawback of the proposed system is the critical need for filters and VAR compensators on the AC side to ensure acceptable performance of the WT generator. Also, smoothing reactors are needed on the DC side for filtering of the transmitted DC current. A simulation model has been built to validate the proposed concept, and the simulation results show the effectiveness of the approach.
  • Symmetrical orientation of spiral-interconnects for high mechanical stability of stretchable electronics

    Qaiser, Nadeem; Damdam, Asrar Nabil; Khan, Sherjeel Munsif; Hussain, Muhammad Mustafa (IEEE, 2020-10-30) [Conference Paper]
    Recently, interconnect based stretchable electronic devices have attained growing interest due to its application for various state-of-the-art technologies. Here, we report an engineered design of spiral interconnects for a series of stretchable networks referred to as the symmetrical series; wherein spirals connect to the island in the symmetry manner. A systematic analysis of Si-based spiral interconnects by numerical modeling, and experiments show that our design provides higher stretchability of 165% in comparison to the conventionally used nonsymmetrical design. The reason for high mechanical reliability is attributed to the favorable unwrapping profile of spiral interconnect due to the nature of forces acting on it during the stretching process. In contrast, for the nonsymmetrical series, the nature of tensile forces produces the rotation, and resultant tilting of spiral arm results in low stretchability of 150%. As a result, nonsymmetrical interconnect fails at earlier stages of stretching. Our study demonstrates the significance of the orientation of spiral interconnects linked to the island to attain the high performance of stretchable electronic devices.
  • Modeling of Beam Electrothermal Actuators

    Hussein, Hussein; Fariborzi, Hossein; Younis, Mohammad I. (Journal of Microelectromechanical Systems, Institute of Electrical and Electronics Engineers (IEEE), 2020-10-30) [Article]
    Beam electrothermal actuators amplify the thermal expansion of pre-shaped beams and use the symmetrical structure to create a linear motion. These actuators, including V and Z shapes, are widely used in microsystems. Explicit analytical expressions are derived in this paper governing the structural deformation of these actuators due to electrothermal expansion and interaction with external lateral load. The analytical expressions are developed for an arbitrary initial shape of the actuator beams with symmetry and uniform cross-section. The modeling is based on the elastic beam theory, and all the modes of buckling are considered in the analytical solution. The solution and analytical expressions account for axial forces in tension and compression. The modeling considers the buckling with the third mode, which occurs at a certain limit of the axial compression and leads to significant variation in the stiffness of the actuator. This phenomenon is well studied for bistable pre-shaped beams, but related studies are limited for beam electrothermal actuators. As a case study, the V shape actuator is specifically investigated. The modeling shows very good agreement with finite element simulations and experimental data based on micro-machined in-plane silicon actuators. [2020-0262]
  • Class-Agnostic Segmentation Loss and Its Application to Salient Object Detection and Segmentation

    Sharma, Angira; Khan, Naeemullah; Sundaramoorthi, Ganesh; Torr, Philip (arXiv, 2020-10-28) [Preprint]
    In this paper we present a novel loss function, called class-agnostic segmentation (CAS) loss. With CAS loss the class descriptors are learned during training of the network. We don't require to define the label of a class a-priori, rather the CAS loss clusters regions with similar appearance together in a weakly-supervised manner. Furthermore, we show that the CAS loss function is sparse, bounded, and robust to class-imbalance. We apply our CAS loss function with fully-convolutional ResNet101 and DeepLab-v3 architectures to the binary segmentation problem of salient object detection. We investigate the performance against the state-of-the-art methods in two settings of low and high-fidelity training data on seven salient object detection datasets. For low-fidelity training data (incorrect class label) class-agnostic segmentation loss outperforms the state-of-the-art methods on salient object detection datasets by staggering margins of around 50%. For high-fidelity training data (correct class labels) class-agnostic segmentation models perform as good as the state-of-the-art approaches while beating the state-of-the-art methods on most datasets. In order to show the utility of the loss function across different domains we also test on general segmentation dataset, where class-agnostic segmentation loss outperforms cross-entropy based loss by huge margins on both region and edge metrics.

View more