Electromagnetic Simulation of Signal Distribution of Various RF Endoluminal Loop Geometries with Coil Orientation: Towards a Reconfigurable Design

With the objective of improving MR endoluminal imaging of the colonic wall, electromagnetic simulations of different configurations of single-layer and double-layer, and double-turn endoluminal coil geometries were run. Indeed, during colon navigation, variations in coil orientation with respect to B0 are bound to occur, leading to impaired image acquisition due to a loss of signal uniformity. In this work, three typical coil orientations encountered during navigation were chosen and the resulting signal uniformity of the different geometries was investigated through the simulated B 1 x , y / I R t values. Sampling this quantity over a circle of radius r enabled us to calculate the coefficient of variation (= standard deviation/mean) for this given distance. This procedure was repeated for r ∈ 5 ; 15 mm, which represents the region of interest in the colon. Our results show that single-loop and double-layer geometries could provide complementary solutions for improved signal uniformity. Finally, using four microelectromechanical system switches, we proposed the design of a reconfigurable endoluminal coil able to switch between those two geometries while also ensuring the active decoupling of the endoluminal coil during the RF transmission of an MR experiment.

A Runtime Reconfigurable Design of Compute-in-Memory–Based Hardware Accelerator for Deep Learning Inference

Compute-in-memory (CIM) is an attractive solution to address the “memory wall” challenges for the extensive computation in deep learning hardware accelerators. For custom ASIC design, a specific chip instance is restricted to a specific network during runtime. However, the development cycle of the hardware is normally far behind the emergence of new algorithms. Although some of the reported CIM-based architectures can adapt to different deep neural network (DNN) models, few details about the dataflow or control were disclosed to enable such an assumption. Instruction set architecture (ISA) could support high flexibility, but its complexity would be an obstacle to efficiency. In this article, a runtime reconfigurable design methodology of CIM-based accelerators is proposed to support a class of convolutional neural networks running on one prefabricated chip instance with ASIC-like efficiency. First, several design aspects are investigated: (1) the reconfigurable weight mapping method; (2) the input side of data transmission, mainly about the weight reloading; and (3) the output side of data processing, mainly about the reconfigurable accumulation. Then, a system-level performance benchmark is performed for the inference of different DNN models, such as VGG-8 on a CIFAR-10 dataset and AlexNet GoogLeNet, ResNet-18, and DenseNet-121 on an ImageNet dataset to measure the trade-offs between runtime reconfigurability, chip area, memory utilization, throughput, and energy efficiency.

Double-String Battery System with Reconfigurable Cell Topology Operated as a Fast Charging Station for Electric Vehicles

This paper introduces a novel design of an electric vehicle (EV) fast charging station, consisting of a battery energy storage system (BESS) with reconfigurable cell topology. The BESS comprises two battery strings that decouple the power flow between EV and grid, to enable charging powers above the grid capacity. The reconfigurable design is achieved by equipping the battery cells with semiconductor switches and serves two main purposes. First, it aims at solving cell unbalance issues to increase safety, reliability, and lifetime of the battery. Second, it enables the BESS to actively control the EV charging process by changing its cell configuration in a real-time fashion, making a DC-DC converter redundant. The paper presents a modelling approach that captures the reconfigurable design including the controlling algorithm used for cell engagement. The simulation results show that the BESS is able to fulfil the EV request with sufficient accuracy for most of the fast charging process. However, the switching of cells leads to variations in the charging current that can potentially exceed the tolerance band defined in IEC61851-23. Therefore, complementary measures are suggested to achieve a suitable current control during all phases of the charging process. The estimated BESS efficiency during the EV fast charging process is 93.3%. The losses caused by the reconfigurable design amount to 1.2% of the provided energy. It is demonstrated that the proposed design has a competitive efficiency compared to a battery buffered fast charging station with DC-DC converter.

The Assessment of the suitability of reconfigurable design through simulation and decision trees

Positioning of the equipment of a manufacturing company is done foreseeing that it will remain fixed for a long time. This is because changes in the plant layout are thought to be costly, as they involve redesigning equipment support services, developing new foundations, using heavy machinery to move equipment, and causing non-productive time at the plant. However, the increased competition due to globalization, lead us to consider the alternative of making modifications to the layout more frequently to improve productivity. Changing the position of the equipment can be facilitated by specifying smaller mobile process equipment or by anticipating the availability of services at different points in the workshop and the existence of cranes or other means of transportation. These forecasts, logically, result in a greater investment in elements that, at least initially, do not generate any profit. This work shows the evaluation of the feasibility of modifying the positions of the equipment when there are variations in the probability distribution in production speeds. This evaluation is carried out using decision trees and process simulation. The case study addressed consists of a generic production line with serial workstations. Jobs arrive at the first computer and join their product on hold if space is available, otherwise the job is lost. In the same way, the product on hold of the other teams are limited in size, and when a product on hold is full, the team behind stops until there is space available. Due to the use, the speeds of the equipment are changing with respect to the ones they had when new: the process times are increasing and / or becoming more variable. To improve productivity under the new process speeds, the space between the equipment can be modified, however, during this change, production must be suspended. This work shows how, by simulating processes and decision trees, it is possible to determine the ranges of values ​​of the suspension time and the cost of transport for which it is convenient to make changes in the spacing of the equipment.