Researchers are motivated recently to generate power using renewable resources especially wind energy. Induction generators are one of the most preferable machines that can be used to extract electrical power from wind. In our research transient performance of double-fed induction generators during abnormal conditions including internal faults has been studied using the EMTP-ATP simulation tool. To validate our simulation results a practical model of a self-excited induction generator which acts as an activated crowbar double-fed induction generator connected mechanically with a variable speed dc motor as a prim-mover has been built. The National Instruments interface card USB-6009 with LabVIEW software is used to extract voltage and current signals from voltage and current transducers. The proposed research work presents a good overview of the internal fault behavior of induction generators with different types of faults.
We describe a silicon photonics-based fabrication tolerant Wavelength Division Multiplexing (WDM) filter with cascaded Mach-Zehnder Interferometers (MZIs), using wavelength independent couplers (WICs). 3D Finite Difference Time Domain (FDTD) simulations show broadband device operation with a spectral shift per waveguide width offset of 6.3 pm/nm which. This offers a fabrication tolerance improvement of over 100 times compared to a standard MZI based WDM filter.
We propose apodized Distributed Bragg Reflector (DBR) bends that offers state of the art performance and compact (0.0024mm2/ch) 4 channel cWDM filter. It is compatible with the most advanced CMOS photonics technology. Simulations show -39 dB sidelobe suppression, -32 dB cross-talk, and -0.05 dB insertion loss for an optimized device configuration.
A multi-objective criterion for tuning a PD controller for satellite attitude control is proposed. The proposed criterion optimizes both the transient response and energy consumption of the system. It is shown via simulations that the multi-objective criterion surpasses the single objective ITAE criterion. The consideration of the energy consumption sets an upper bound to the optimization problem resulting in more practical controller gains.
In recent years, electric vehicles are gaining popularity over conventional internal combustion engine driven vehicles as electric vehicles have better control, less noise, low air pollution, compact size and better efficiency due to regenerative breaking. In this paper, two Open-End Winding Induction Motors are used in Differential 4 Wheel Drive topology, with four 2-level Voltage Source Inverters powering the two motors. Each side of the motor is powered through an inverter, resulting a 3-level inverter output. An improved 7-level Direct Torque Controller is used to control the switching states of the four inverters with the aim of reducing the torque ripples while achieving a stable normal operation and fault tolerant operation. The necessary models are developed, and the proposed scheme is verified through simulations.
Brain-inspired architectures gain increased attention, especially for limited resources edge devices to perform cognitive tasks. The hyperdimensional computing (HDC) paradigm is an emerging framework inspired by an abstract representation of neuronal circuits' attributes in the human brain. HDC has shown promising results for 1D applications, such as text classification, utilizing less power and lower latency than convolutional neural networks (CNN). Hence, in this paper, HDC is analyzed in terms of accuracy for image classification tasks. The paper reveals for 2D applications, HDC can achieve an adequate performance utilizing only 16 % of the training dataset.
In this research, we study the optimal batching and scheduling in a single machine system where production costs have non-linear dependence on the batch composition. The main objective is to minimize the production costs while taking into account the due production dates. The problem originates in using metal Additive Manufacturing (aka 3D printing) for spare parts production where multiple items of different types can be batched together in the printing chamber, and the printing costs heavily depend on the batch composition. This paper presents a mixed-integer linear scheduling problem and demonstrates how it can be efficiently solved using the set covering reformulation.
Cumulative Sum (CUSUM) quality control chart is commonly used in many industrial applications. In this research, an improved scheme of the CUSUM chart is introduced to monitor attribute data. In this scheme, the difference between actual and in-control numerals of nonconforming items is raised to an exponent w to enhance the detection effectiveness. The exponent w is optimized, along with other charting parameters to minimize the Average Number of Defectives (AND) which is used as an objective function. The proposed scheme is found to outperform the conventional CUSUM chart for detecting a wide range of shifts in fraction non-conforming.
Supply chains are a vital component in many industries that provide vital services on a daily basis, and healthcare supply chains are not an exception. However, these supply chains are very complicated in the way they are established which introduces critical issues such as imprecise or incorrect information, lack of appropriate historical record, and lack of transparency. These issues all together hinder the process of tracking and tracing throughout the pharmaceutical supply chain. Therefore, to guarantee that only authentic drugs are being transported within the supply chain, a comprehensive, end-to-end solution is needed. There has already been a number of attempts to resolve these issues, however, they are conceptualized and built in a centralized system resulting in transparency, data privacy, and authenticity issues. This paper leverages the smart contract feature in the Ethereum blockchain to provide traceability, authenticity, and data provenance. In addition, decentralized off-chain storage is utilized when needed.
This study proposes a framework to predict ma- chine failures using sensor data and optimize the predictive/corrective maintenance schedule. Machine learning (ML) models are trained to predict the failure probabilities. The ML model's output is fed to an optimization model to propose a maintenance policy. Hence, demonstrating how prediction models can help increase system reliability at lower costs.