Search

Deep convolutional neural networks (DCNNs) have been proposed as enhanced developments of neural networks (NNs) in the field of artificial intelligence (AI) and successfully applied in deep learning (DL) scenarios. With the advancement of technology, the number of network layers has continuously increased, resulting in a huge number of calculations and memory accesses required in the training and inference process of DCNNs and thereby hindering their further deployment and application. Using a specific dataflow formed by reusable DCNN data in the network-on-chip (NoC), reducing the memory access pressure and improving DCNN processing efficiency has become a promising acceleration scheme for the current DCNN. In this article, a novel convolution layer (CONV) acceleration strategy based on logical PE set segmentation for row stationary (RS) dataflow is proposed to solve the problems of low flexibility and inefficient processing array utilization faced by the conventional folding mapping strategy. The simulation results show that the new mapping strategy based on PE set segmentation can achieve better processing element utilization and CONV acceleration improvement at the expense of little increase in the data movement energy consumption compared with the conventional strategy. [ABSTRACT FROM AUTHOR]

With the growing penetration of power electronic converters in power systems, the issue of reliability becomes more critical than ever before. This paper proposes a hierarchical reliability framework to evaluate the electric power system reliability from the power electronic converter level to the overall system level. On the converter level, the reliability model of a power electronic converter is developed based on the power electronic devices it is composed of, for which various hourly based input profiles and converter topologies are considered. On the system level, reliability metrics such as expected energy not served (EENS) and loss of load expectation (LOLE) are estimated through a non-sequential Monte Carlo simulation. Machine learning regression models, such as support vector regression (SVR), and random forests (RF) are implemented to bridge the nonlinear reliability relationship between two levels. The proposed framework is demonstrated through the modified IEEE Reliability Test System (RTS) 24-bus network. Numerical results show power converter reliability should be considered as an important factor when evaluating overall system reliability performance. [ABSTRACT FROM AUTHOR]

This paper proposes a quadrature-shaped pickup coil, which can realize an angular-misalignment insensitive omnidirectional wireless charging system. Recently, the three-dimensional wireless charging has shown promising prospects for removing the performance deterioration caused by the positional misalignment, which can transfer power to arbitrary spatial objectives. In the conventional design, the strength and the direction of the induced magnetic field can be adjusted by controlling the excitation currents of transmitting coils. However, the existence of angular misalignment caused by the self-rotation of the pickup unit still can result in a sharp decrease of the transmitted power. Accordingly, this paper proposes and implements a novel pickup coil topology to compensate such huge fluctuations in power transfer, which aims to ensure an angular misalignment insensitive omnidirectional wireless charging. Simulated and experimental results are both given to verify the feasibility of the proposed quadrature-shaped topology. [ABSTRACT FROM AUTHOR]

This paper proposes an optimal design of quadrature-shaped pickup to satisfy various demands on performance at special positions for omnidirectional wireless power transfer (WPT). Omnidirectional WPT technology as a brand new way of energy supply has shown great potentials for charging objectives at all directions. In order to ensure the energy can be transmitted to arbitrary spatial directions, the orthogonal structure of three circular transmitting coils has been widely accepted matched with non-identical current control scheme. However, there are still some special positions where the transmitted power drops to the bottom which is not suitable for applications. Accordingly, this paper proposes an optimal design analysis for the quadrature-shaped pickup to satisfy different demands on either power or efficiency at such positions. The simulated analysis and experimental results are both given to verify the enhancement of the proposed quadrature-shaped pickup, which aims to provide a reference to fabricate an appropriate pickup up for utilizations. [ABSTRACT FROM AUTHOR]

Transcranial magnetic stimulation (TMS) is a non-invasive neuromodulation technique investigated for the treatment of various neuropsychiatric disorders such as major depressive disorder and obsessive-compulsive disorder. Coil design and analysis has been a popular topic in TMS studies, primarily as the coil geometry significantly affects the induced electric field ($E$ -field) distribution. In this work, we compare the distribution of the induced $E$ -field of 16 different TMS coils by finite element analysis. The coils are positioned at two head locations: the vertex and the dorsolateral prefrontal cortex (DLPFC). Due to the coil geometry, only nine of the 16 coils are positioned at the DLPFC. Notably, the authors use 50 heterogeneous head models derived from MRI scans of healthy patients in this analysis. As a result, the sensitivity of the $E$ -field intensity and focality to anatomical variations is investigated. The maximum $E$ -field intensity induced in the brain ($E$ -Max brain), and the volume in the brain exposed to the $E$ -field with at least half the $E$ -Max ($V$ -Half) were metrics used to assess the intensity and focality of the induced $E$ -field. It was observed that some of the coils induced high $E$ -field intensity and were highly sensitive to anatomical variations. Other coils exhibited high focality and lower sensitivity to anatomical variations at the same time. For the coils positioned at the DLPFC, less variability was observed when compared to the vertex location. This study provides an understanding of the effect of coil geometry and anatomical variation on the induced $E$ -field during TMS. [ABSTRACT FROM AUTHOR]

Coil design for transcranial magnetic stimulation (TMS) is a critical factor that is constantly being investigated to provide more focused stimulation of the brain. The design of focal coils is important because TMS uses time-varying magnetic fields generated from electromagnetic coils to stimulate a specific region of interest of the brain non-invasively. TMS is a useful tool in both diagnostic and therapeutic purposes for identifying and treating psychiatric and neurological disorders. Currently, TMS is approved by the U.S. Food and Drug Administration for the treatment of chronic depression and is being researched for the treatment of stroke, obsessive compulsive disorder, post-traumatic stress disorder, and Parkinson’s disease. In this paper, the authors are proposing to further improve the focality of the quadruple butterfly coil (QBC) by using a high-permeability ferromagnetic material acting as passive magnetic shield. Analyses of the resulting field profiles with 50 heterogeneous head models using a single and double magnetic shield has been conducted at two different positions in the brain: the vertex and the dorsolateral prefrontal cortex. Comparison of the results has been illustrated with the help of QBC and figure-8 coil by using parameters of interest such as maximum electric field intensities in both the brain and the entire head, the volume of the brain receiving strong stimulation intensities, and the surface area of the brain receiving strong stimulation intensities. There is a significant amount of improvement found in focality when the QBC is used with a magnetic shield. With a shield, the coil is still able to maintain a comparable maximum electric field intensity in the targeted region of the brain to that of the original QBC. Although both shielding solutions show significant improvement, no notable difference has been found in the results between single and double shields with the QBC. [ABSTRACT FROM AUTHOR]

In order to interface one photovoltaic (PV) port, one bidirectional battery port, and one load port of a PV-battery dc power system, a novel nonisolated three-port dc/dc converter named boost bidirectional buck converter (B3C) and its control method based on three-domain control are proposed in this paper. The power flow and operating principles of the proposed B3C are analyzed in detail, and then, the dc voltage relation between three ports is deduced. The proposed converter features high integration and single-stage power conversion from both PV and battery ports to the load port, thus leading to high efficiency. The current of all three ports is continuous; hence, the electromagnetic noise can be reduced. Furthermore, the control and modulation method for B3C has been proposed for realizing maximum power point tracking (MPPT), battery management, and bus voltage regulation simultaneously. The operation can be transited between conductance mode and MPPT mode automatically according to the load power. Finally, experimental verifications are given to illustrate the feasibility and effectiveness of the proposed topology and control method. [ABSTRACT FROM PUBLISHER]