Search results
(421 - 440 of 463)
Pages
- Title
- STRATEGIES TO MAXIMIZE DOSE REDUCTION IN SPECT MYOCARDIAL PERFUSION IMAGING
- Creator
- Juan Ramon, Albert
- Date
- 2019
- Description
-
Radiation exposure in medical imaging has become a topic of major concern, gaining intense attention within the clinical and research...
Show moreRadiation exposure in medical imaging has become a topic of major concern, gaining intense attention within the clinical and research communities. In 2009, the National Council on Radiation Protection and Measurements (NCRP) announced radiation exposure of patients via medical imaging increased more than sixfold between the 1980s and 2006, with cardiac nuclear medicine, specifically myocardial perfusion imaging (MPI) with single-photon emission computed tomography (SPECT) being the second biggest culprit. The goal of this work is to evaluate several strategies to enable radiation dose to be minimized while maintaining current levels of diagnostic accuracy in the clinic. We achieve dose reduction through optimization of advanced image reconstruction strategies, to obtain higher-quality images at a given dose (noise) level, through a machine learning approach to predict the optimal dose for each patient, and through advanced deep learning (DL) algorithms to improve the quality of reconstructed images. Our ultimate objective is to provide the nuclear cardiology field with a new set of algorithms and guidelines for selecting administered activity levels and image reconstruction procedures in the clinic. The project is based on a clinical study in which imaging and various other data are being collected for a set of patients. The project has the following components. First, we investigate a global dose-reduction approach (i.e., reducing dose by a uniform proportion across all patients) via optimization of image reconstruction strategies. Specifically, we maximize perfusion-defect detection (diagnostic accuracy) over a range of simulated dose levels using clinical data into which we have introduced simulated defects. We measure diagnostic performance using clinically validated model observers from the Quantitative Perfusion SPECT (QPS) software package. We investigate the diagnostic accuracy over a range of dose levels ranging from those currently used in the clinic down to one-eighth of this level. We consider the following image-reconstruction: filtered-backprojection (FBP) with no correction for physics effects, and ordered-subsets expectation-maximization (OS-EM) with several combinations of attenuation correction (AC), scatter correction (SC), and resolution correction (RC).Second, we propose a patient-specific ("personalized") dose reduction approach based on machine learning that aims to predict the minimum radiation dose needed to obtain consistent perfusion-defect detection accuracy for each individual patient. This prediction is based on patient attributes, especially body measurements, and various clinical variables. We compare the diagnostic accuracy produced by predicted personalized doses to that produced by standard clinical dose levels to validate the predictive models.Third, we verify that the dose minimization results obtained in the context of perfusion-defect detection also maintain diagnostic accuracy in evaluating cardiac function, as characterized by myocardial motion.Finally, we propose a deep learning (DL) method to denoise SPECT-MPI reconstructed images. The method is a 3D convolutional neural network trained to predict standard-dose images from low-dose images. We quantify the extent to which dose reduction can be achieved using the proposed DL structure when dose is reduced uniformly across patients or by means of our patient-specific approach.
Show less
- Title
- LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
- Creator
- Song, Chao
- Date
- 2019
- Description
-
Single photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery...
Show moreSingle photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery diseases. The image quality in cardiac SPECT can be adversely affected by cardiac motion and respiratory motion, both of which can lead to motion blur and non-uniform heart wall. In this thesis, we mainly investigate imaging de-noising algorithms and motion correction methods for improving the image quality in cardiac SPECT on both standard dose and reduced dose.First, we investigate a spatiotemporal post-processing approach based on a non-local means (NLM) filter for suppressing the noise in cardiac-gated SPECT images. Since in recent years low-dose studies have gained increased attention in cardiac SPECT owing to its potential radiation risk, to further improve the image quality on reduced dose, we investigate a novel de-noising method for low-dose cardiac-gated SPECT by using a three dimensional residual convolutional neural network (CNN). Furthermore, to reduce the negative effect of respiratory-binned acquisitions and assess the benefit of this approach in both standard dose and reduced dose using simulated acquisitions. Inspired by the success in respiratory correction, we investigate the potential benefit of cardiac motion correction for improving the detectability of perfusion defects. Finally, to combine the benefit of above two types of motion correction, dual-gated data acquisitions are implemented, wherein the acquired list-mode data are further binned into a number of intervals within cardiac and respiratory cycle according to the electrocardiography (ECG) signal and amplitude of the respiratory motion.
Show less
- Title
- Rapidly Deployable PV-Based Smart Irrigation System
- Creator
- Usta, Salih
- Date
- 2019
- Description
-
There are many agricultural fields in developing countries such as Turkey which do not have electricity on site. In order to water these...
Show moreThere are many agricultural fields in developing countries such as Turkey which do not have electricity on site. In order to water these fields, there is usually a need to store water in a water reservoir nearby. This purpose is achieved by manpower or by using diesel-operated water pumps which are often inefficient and require a high degree of maintenance over time. Furthermore, extending the power supply grid to the field is not considered an option by governors, due to the high cost for a relatively small-scale application. Along with this, watering the field is done by farmers, which frequently leads to waste of water, or leads to watering one particular area of the field less than the others, which causes a drop in crop efficiency. Preventing water waste is considered an important issue in the 21st century. Also, increasing crop efficiency in a developing country is an important consideration. To prevent water waste and to enhance crop efficiency, an automated irrigation system is needed. The objective of this study is to develop a photovoltaic-based irrigation system for an agricultural field that is not tied to an existing conventional electric grid. Firstly, a stand-alone PV system is designed according to the field requirements. Secondly, a soil moisture sensor-based smart irrigation system is developed for an automated irrigation process compatible with drip irrigation systems. This system also enables users to monitor and analyze soil moisture data. By developing this type of complete irrigation mechanism, a long-term lower cost, efficient, and environmental-friendly system is designed.
Show less
- Title
- Multi-function multi-modality sensing and communication system: a designer's perspective
- Creator
- Fepeussi, Tonmo Vanessa Carine
- Date
- 2019
- Description
-
The combination of sensing and communication functionalities on the same electronic device is the key to autonomous sensing applications in...
Show moreThe combination of sensing and communication functionalities on the same electronic device is the key to autonomous sensing applications in the transportation industry, including driverless vehicles and structural health monitoring (SHM) of aero-vehicles. Due to the limited availability of spectral and hardware resources, there is a need for resource sharing between sensing and communication systems. This is achieved by the efficient integration of sensing and communication functions through a unified design of both systems into smart sensors. To that end, a multi-modality approach is employed in this research to design multi-functional systems at two different bands of the frequency spectrum, namely radio and acoustic frequencies.First, a radio-frequency (RF) software-defined system capable to support radar sensing and RF communication is proposed for use in modern interconnected automotive applications such as driverless vehicles. The proposed RF radar is designed on a software-defined homodyne transceiver prototype capable of radio communication. The system is implemented in the S band over a narrow frequency bandwidth of 34 MHz between 3.550 GHz and 3.584 GHz. Experimental measurements show that the designed radar sensor can measure short-range targets with a range accuracy of less than 21 cm.An acoustic sensing and communication system is developed in parallel for use in autonomous SHM monitoring of aero-vehicles. The proposed communication system uses M-ary time-reversal pulse position modulation (M-TRPPM) as the modulation scheme for dispersion compensated wireless communication across the elastic channel. The time reversal based time division multiple access (TR-TDMA) protocol is introduced to regulate channel access by multiple sensors. Simulation and experimental validation demonstrate that the designed system, using an excitation signal generated by a PZT sensor disc at 300 kHz resonant frequency, is capable of reliable data transmission with a bit error rate (BER) approximating zero at low data rates of a few kilobits per seconds.
Show less
- Title
- Wireless Body Sensor Network for Tracking Human Mobility using Long Short-Term Memory Neural Network for Classification
- Creator
- Gupta, Saumya
- Date
- 2019
- Description
-
A large number of sensors are used without justification of the number chosen or placement choice. Many papers about body sensor networks...
Show moreA large number of sensors are used without justification of the number chosen or placement choice. Many papers about body sensor networks explore how to capture a type or types of motion, but all their sensors are placed in different locations; making their algorithms very specific to that movement. In this research, we explore the enhancement of human activity classification algorithm using long short-term memory (LSTM) neural network and wearable sensor network. There are five identical nodes used in the body sensor network to collect data. Each node incorporates an ESP8266 Microcontroller with Wi-Fi which is connected to an inertial measurement unit consisting of triple axis accelerometer and gyroscope sensor board. An analysis on the accuracy that each sensor node provides separately and in different combinations has been conducted to allow future research to focus their positioning in optimal positions. A Robot Operating System (ROS) central node is used to illustrate the in-built multi-threading capability. For demonstration, the positions chosen are waist, ankles and wrists. The raw sensor data can be observed on screen while it is being labelled live to create fitting dataset for developing an artificial neural network. Expectation is that increasing the number of sensors should raise the overall accuracy of the output but that isn’t the case observed, positioning of the sensor is pertinent to improvement. These platforms can be further extended to understand different motions and different sensor positions, also expanded to include other sensors.
Show less
- Title
- Integrated Design Framework For Electric Motor Drive Systems
- Creator
- Salameh, Mohamad
- Date
- 2020
- Description
-
This thesis aims to develop a flexible and time-efficient framework for machine design optimization that considers driving cycles,...
Show moreThis thesis aims to develop a flexible and time-efficient framework for machine design optimization that considers driving cycles, multiphysics domains and current design. The proposed development of the framework is based on the enhancement of three key aspects in the machine design process. A data mining algorithm – the X-means – is employed in the driving cycle analysis, to establish a trade-off between the optimization objectives and the computational intensity. A novel vibration surrogate model is proposed to evaluate the vibroacoustic behavior of the machine in an accurate and time-efficient way. In the identification process, the time effectiveness of the model is attained with a minimized number of finite element simulations. Furthermore, the principle of simultaneous coupled optimization is considered in the framework, where current design variables are included in the optimization environment to allow identifying design candidates with improved performance.
Show less
- Title
- ANALYSIS AND OPTIMIZATION OF VIBRATION AND ACOUSTIC NOISE IN SWITCHED RELUCTANCE MACHINES
- Creator
- Yaman, Selin
- Date
- 2019
- Description
-
One of the main drawbacks of switched reluctance machines (SRM) is the vibration and high acoustic noise compared to other electrical motors....
Show moreOne of the main drawbacks of switched reluctance machines (SRM) is the vibration and high acoustic noise compared to other electrical motors. The root cause of the high level of acoustic noise is radial forces with high harmonic content. These harmonics may trigger resonant modes in the stator and cause the machine to create high vibration and acoustic noise. To better understand the factors influencing vibration and acoustic noise in an SRM, this dissertation first develops a multi-physics model in ANSYS Workbench environment and carries out a comprehensive analysis of multiple variations in stator and rotor geometries. Based on this understanding, this dissertation identifies distinct factors affecting noise in the machine, which are affected by electromagnetic design and power electronic control. From the electromagnetic perspective, geometrical optimizations in the stator and rotor structures are evaluated to understand the impact on NVH (noise, vibration and harshness) performance. This background is used to develop a fast geometry-sensitive analytical approach to reduce acoustic noise in the machine. While optimizing the geometry for a silent machine design, different design of experiments (DoE) methods and response surface (RS) optimization methods are also compared and presented. Furthermore, material analysis is included in structural design, where high flux material effect on vibration and acoustic noise is observed. The second topic of the NVH analysis is power electronic and switching solutions. In this study, multiple basic and advanced switching techniques have been considered and optimized to reduce acoustic noise under a preset efficiency constraint. Further, a time efficient model of SRM is introduced with vibro-acoustic noise perspective by developing a computationally cost effective SRM modeling. By using this analytical time-efficient NVH model, a current shape optimization is implemented, and results are discussed. Finally, experimental validations are provided for NVH and psychoacoustics analysis for different operating conditions and current control methods.
Show less
- Title
- COMMUNICATION-BASED DISTRIBUTED CONTROL IN MICROGRIDS AND NETWORKED MICROGRIDS
- Creator
- Zhou, Quan
- Date
- 2019
- Description
-
Microgrids representing localized small-scale power systems are capable of operating as self-controlled entities, which cluster and manage...
Show moreMicrogrids representing localized small-scale power systems are capable of operating as self-controlled entities, which cluster and manage distributed energy resources (DERs) and other smart devices within a defined electrical boundary. By utilizing locally available resources, microgrids reduce their dependencies on the utility grid, which provide more reliable, resilient, and economic power services to local customers. Geographically close microgrids can be connected for forming a networked microgrid system, which provides additional operational flexibility and further enhances the system reliability and resilience by sharing available DERs.Considering variable and controllable characteristics of DERs, locally available DERs need to be appropriately coordinated and controlled to respond to changing loads. The proliferation of microgrids will make it inevitable to rely on communication systems among microgrids for realizing the coordinated control of participating DERs in networked microgrid systems. The networked microgrid system is considered as a cyber-physical system (CPS) which requires sophisticated network technologies to cope with the massive adaption of communication, computation and control devices. Conventionally, a networked system has been managed by a centralized master controller, which processes the data collected from participating DERs and sends operational set points to each participant.Compared with the centralized control strategies, distributed control is more advantageous for connecting participating DERs. The connectivity of distributed control system (i.e., meshed network) is higher than that of a centralized structure (i.e., star network), in particular when critical circumstances are encountered in which some of the network connections are lost. Also, the distributed control system enables parallel data processing and control, which speeds up the networked system response to variable DERs and loads and promote economic merits. The communication-based distributed control strategies have proven to demonstrate higher reliability, resilience, and scalability while possessing lower implementation costs as compared with centralized control strategies.We have proposed several communication-based distributed control strategies for realizing the coordinated operation of participating microgrids and DERs, which can be applied to achieve various operational objectives, including proportional active power sharing, DER plug-and-play capability, seamless microgrid islanding, and resynchronization operations, and optimal economic operations. The benefits and challenges of communication-based distributed control strategies in networked microgrid systems are discussed and addressed in our work. Extensive case studies have been conducted in this thesis to validate the effectiveness of the proposed communication-based distributed controller design.
Show less
- Title
- BLOCKCHAIN FOR TRANSACTIVE ENERGY MARKET WITH NETWORKED MICROGRIDS
- Creator
- Yan, Mingyu
- Date
- 2021
- Description
-
Transactive energy, which allows and incentivizes microgrids (MGs) to trade energy with each other, is regarded as the next-generation energy...
Show moreTransactive energy, which allows and incentivizes microgrids (MGs) to trade energy with each other, is regarded as the next-generation energy management scheme to accommodate the penetration of distributed energy resources (DERs). Blockchain provides an effective and decentralized strategy, which can address the operational challenges introduced by the transactive energy market. This thesis is aimed at providing effective transactive energy markets for incentivizing MGs to trade energy and utilizing blockchain technologies to provide a secure and efficient energy trading environment for all participants.First, this thesis offers a centralized transactive market for networked MGs to transact energy through the centralized distribution system operator (DSO) while ensuring the power network limits. All MGs cooperate in this market and the cooperative behaviors are captured using the cooperative game with externalities. A two-level problem is studied to allocate the total payoff to all participating MGs. Numerical results for a 4-MG system and the IEEE 33-bus show the validity of the centralized transactive energy model. Second, this thesis proposes a two-level network-constrained peer-to-peer (P2P) transactive energy for multi-MGs, which guarantees the distribution power network security and allows MGs to trade energy with each other flexibly. At the lower level, a P2P transactive energy is employed for multi-MGs to trade energy with each other. A multi-leader multi-follower (MLMF) Stackelberg game approach is utilized to model the energy trading process among MGs. At the upper level, the DSO reconfigures the distribution network based on the P2P transactive energy trading results by applying the AC optimal power flow considering the distribution network reconfiguration. If there are any network violations, the DSO requests energy trading adjustments at the lower level for network security. Numerical results for a 4-MG system, the modified IEEE 33-bus, and the 123-bus distribution power systems show the effectiveness of the proposed transactive energy model and its solution technique. Third, this thesis adopts the blockchain for the peer-to-peer transactive energy market among MGs. A two-level integrated blockchain-power system is provided, in which all MGs and the DSO are equipped with blockchain. At the lower level, MGs trade energy with each other through the lower-level MG blockchain, while the DSO manages the network security through the upper level DSO blockchain. We illustrate how to utilize blockchain technologies, i.e., public and private keys and smart contracts, to provide an efficient and secure energy trading environment for all MGs. Last, this thesis applies the blockchain for transacting energy and carbon right for networked MGs. MGs transact energy and carbon right through the centralized DSO while ensuring the power network limits. The introduction of blockchain achieves secure and decentralized market settlements in this centralized market. Numerical results for a 4-MG system and modified IEEE 33-bus systems show the effectiveness of the proposed transactive energy and carbon market.
Show less
- Title
- IMPROVING DEEP LEARNING BASED SEMANTIC SEGMENTATION USING CONTEXT INFORMATION
- Creator
- Xia, Zhengyu
- Date
- 2021
- Description
-
Semantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately...
Show moreSemantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately. Nowadays, applications such as autonomous driving, path navigation, image search engine, or augmented reality require accurate semantic analysis and efficient segmentation mechanisms. In this thesis, we propose multiple models to improve the performance of semantic segmentation. In the first part, we focus on the single-task network, which aims to improve the performance of semantic segmentation. Our research includes exploiting context information using mixed spatial pyramid pooling to extract dense context-embedded features in FCN-based semantic segmentation. We also propose a GAF module to generate a global context-based attention map to guide the shallow-layer feature maps for better pixel localization. In the second part, we focus on a multi-task network that incorporates semantic segmentation to improve other computer vision tasks such as object detection. Specifically, a multi-task network, along with a learning strategy is designed to let semantic segmentation and object detection assist each other since they are highly correlated. Also, we include weakly-supervised multi-label semantic segmentation learning to deal with the shortage of high-quality training examples and to improve the performance of cross-domain object detection. In the third part, we focus on improving the performance of video panoptic segmentation, which is a unified network that incorporates semantic segmentation and instance segmentation using video streams. We design a new ConvLSTM pyramid to transmit spatio-temporal contextual information in our video panoptic segmentation network. Specifically, we propose a modified ConvLSTM to generate temporal contextual information. Also, we design an MSTPP module to obtain mixed spatio-temporal context-embedded feature maps. Experimental results on different datasets show that our proposed method achieves better performance compared with the state-of-the-art methods.
Show less
- Title
- MACHINE VISION NAVIGATION SYSTEM FOR VISUALLY IMPAIRED PEOPLE
- Creator
- Yang, Guojun
- Date
- 2021
- Description
-
Visually impaired people are often challenged in the efficient navigation of complex environments. Moreover, helping them navigate intuitively...
Show moreVisually impaired people are often challenged in the efficient navigation of complex environments. Moreover, helping them navigate intuitively is not a trivial task. Cognitive maps derived from visual cues play a pivotal role in navigation. In this dissertation, we present a sight-to-sound human–machine interface (STS-HMI), a novel machine vision guidance system that enables visually impaired people to navigate with instantaneous and intuitive responses. This proposed system extracts visual context from scenes and converts them into binaural acoustic cues for users to establish cognitive maps. The development of the proposed STS-HMI system encompasses four major components: (i) a machine vision–based indoor localization system that uses augmented reality (AR) markers to locate the user in GPS-denied environments (e.g., indoor); (ii) a feature-based object detection and localization system called the simultaneous localization and mapping (SLAM) algorithm, which tracks the mobility of users when AR markers are not visible; (iii) a path-planning system that creates a course towards a destination while avoiding obstacles; and (iv) an acoustic human–machine interface to navigate users in complex navigation courses. Throughout the research and development of this dissertation, each component is analyzed for optimal performance. The navigation algorithms are used to evaluate the performance of the STS-HMI system in a complicated environment with difficult navigation paths. The experimental results confirm that the STS-HMI system advances the mobility of visually impaired people with minimal effort and high accuracy.
Show less
- Title
- EVENT-BASED NONINTRUSIVE LOAD MONITORING
- Creator
- Yan, Lei
- Date
- 2021
- Description
-
Non-Intrusive Load Monitoring (NILM) is an important application to monitor household appliance activities and provide related information to...
Show moreNon-Intrusive Load Monitoring (NILM) is an important application to monitor household appliance activities and provide related information to house owner or/and utility company via a single sensor installed at the electrical entry of the house. With this information, utilities can perform many tasks such as energy conservation, planning gen-eration more wisely, and demand response (DR) study. For house owners, they can un-derstand their bill more clearly and make better budget plan. For researchers, NILM sys-tem is a good foundation for energy management in buildings and can provide valuable power information for smart homes design. This dissertation aims to develop and demon-strate a complete and accurate event-based NILM system, which includes (1) an edge-cloud framework for event-based NILM, (2) an adaptive event detection method, (3) a two-stage event-based load disaggregation method; and (4) a high-resolution (50Hz) NILM dataset. Event detection is the first step in event-based NILM and it can provide deter-ministic transient information to identify appliances. However, existing methods with fixed parameters suffer from unpredictable and complicated changes in smart meter data such as long transition, high fluctuation and near-simultaneous events in both power and time domains. This dissertation presents an adaptive method to detect events based on home appliance load data with high sampling rate (>1Hz) by flexibly tuning the parame-ters according to the data being processed. The proposed method runs fast over the data stream and captures the transient process by multi-timescales searching as well. The mi-cro-timescale and macro-timescale window could deal with near-simultaneous events and long-transition events, respectively. Transient load signatures are extracted from detected events and stored in a sequential tree struct that can be used for NILM and load recon-struction, etc. Case studies on a 20Hz dataset, the LIFTED dataset of 50Hz, and the BLUED dataset of 60Hz demonstrate that the proposed method is able to work on data of different sampling rates and outperforms other methods in event detection. The ex-tracted load signatures could also improve the efficiency of NILM and help develop oth-er applications. This dissertation presents an online transient-based electrical appliance state track-ing method for NILM. The proposed Factorial Particle based Hidden Markov Model (FPHMM) method takes advantage of transient features in high-resolution data to infer states in the transient process and conducts steady state verification to rectify falsely identified appliances. The FPHMM method can overcome the common feature similarity problem in NILM by combining particle filter method and Markov Chain Monte Carlo sampling method, and by mining the intra-relationship of states inside a single appliance and the inter-relationship of states among multiple appliances. The FPHMM method is tested on the LIFTED dataset with appliance-level details and high sampling rates. Test-ing results demonstrate that the FPHMM method is effective in resolving the feature similarity issue. A modified mean shift algorithm with different levels of bandwidth is proposed as well to cluster the extracted features from event detection. Based on the clustered fea-tures, another solution is proposed to decode the states of appliance in two stages. The first stage uses Bayesian Inference Factorial HMM (BI-FHMM) solver to accelerate com-putational speed and improve accuracy by integrating the load signatures and statistical inference. The second stage then verifies and rectifies the results obtained from the first stage. Test results demonstrate that the proposed approach achieves good performance and can be applied to existing smart meters.
Show less
- Title
- RESIDENTIAL LOAD DATA COMPRESSION AND LOAD DISAGGREGATION
- Creator
- Xu, Runnan
- Date
- 2021
- Description
-
Non-Intrusive Load Monitoring (NILM) for residential applications aims to dis-aggregate the total electricity consumption of a household into...
Show moreNon-Intrusive Load Monitoring (NILM) for residential applications aims to dis-aggregate the total electricity consumption of a household into the single appliance information. For the customer side, users can change their consumption habit and save more electricity. For the utility, generation scheduling will be more accurate, efficient, and secure. Furthermore, energy management system, demand response and fault diagnosis will benefit from the real time information provided by the NILM. This dissertation first proposes a data compressed method suitable for the NILM data. Then a real time disaggregation based on the Kalman filter is proposed to obtain the appliance state information. A model-free lossless data compression method for time series in smart grids (SGs), namely, Lossless Coding considering Precision (LCP) method is proposed. The LCP method encodes the current datapoint only using the immediate previous datapoint by differential coding, XOR coding, and variable length coding and transmits the encoded data once generated. It does not use the dynamics (e.g., many previous datapoints) or prior knowledge (e.g., mathematical models) of the time series. It considers the patterns, potential applications, and associated precision to preprocess the time series and especially suits high-resolution time series with long steady periods. The LCP method features low-latency and generalizability which enables real-time data communication for different time-critical tasks. Sub-metered load profiles in REDD dataset, high-resolution LIFTED dataset, AMPds dataset and PMU dataset are used to evaluate the performance of the LCP method. The results show that the LCP method demonstrates high compression ratio, low latency, and low complexity compared to state-of-the-art Resumable Data Com-pression (RDC) method, DEFLATE based on LZ77 & Huffman coding, and Lempel-Ziv-Markov Chain Algorithm (LZMA). An online method based on the transient features of individual appliances and system steady-state characteristics is proposed to estimate the appliances’ working states. It determines the number of states for each appliance via Density-based Spatial Clustering of Applications with Noise (DBSCAN) method and models the transition relationship among different states. The states of working appliances are identified from aggregated power signals by implementing the Kalman filtering method into the Factorial Hidden Markov Model (FHMM) and by the verification of system states which are the combination of working states of individual appliances. The proposed method is event based and the use of transient features extracted from event detection could achieve fast state inference and is suitable for online load disaggregation. The proposed method is tested on high-resolution dataset such as LIFTED and outperforms other related methods, including Segment-wise Integer Quadratic Constraint Programming (SIQCP), Combinatorial Optimization (CO), and the exact FHMM (FHMM_EXACT), in terms of accuracy, f1 score, and computational time.
Show less
- Title
- TASK-BASED LOAD FORECASTING AND ROBUST RESOURCE SCHEDULING IN SMART GRID
- Creator
- Han, Jiayu
- Date
- 2021
- Description
-
In microgrids, the uncertainty of load and renewables and lack of generation capacity will lead to a wide variety of operation problems in...
Show moreIn microgrids, the uncertainty of load and renewables and lack of generation capacity will lead to a wide variety of operation problems in both grid-connected mode and islanded mode. This motivates the design of the state-of-art microgrid master controller for microgrid energy management, load forecasting, and demand response. Uncertainty in renewables and load is a great challenge for microgrid operation, especially in islanded mode as the microgrid may be small in size and has limited flexible resources. A multi-timescale, two-stage robust dispatch model is proposed to optimize the microgrid operation. The proposed one uses only one model to combine the hourly and sub-hourly dispatch together, which means the day-ahead hourly dispatch results must also satisfy the sub-hourly conditions. At the same time, the feasibility of the day-ahead dispatch result is verified in the worst-case condition considering the high-level uncertainty in renewable energy output and load consumptions. In addition, battery energy storage system (BESS) and solar PV units are integrated as a combined solar-storage system in the proposed model and the output power of the combined solar-storage system remains unchanged on an hourly basis. Furthermore, both BESS and thermal units provide regulating reserve to manage solar and load uncertainty. The model has been tested in a controller hardware in loop (CHIL) environment for the Bronzeville Community Microgrid system in Chicago. The simulation results show that the proposed model works effectively in managing the uncertainty in solar PV and load and can provide a flexible dispatch in both grid-connected and islanded modes.When the generation capacity of an islanded microgrid is less than the load demand, load curtailment is inevitable. This dissertation proposes a multi-objective optimization model to minimize the load curtailments. Specifically, the proposed model minimizes the generation cost and total load curtailments and also minimizes the maximum load curtailment. Furthermore, the impact of the penalty coefficients of total load curtailment and maximum load curtailment is analyzed, which provides a strategy to choose the value of the two penalty coefficients according to different practical purposes. The proposed model can be used in both microgrid generation scheduling and microgrid planning problems. It was tested in the Bronzeville Community Microgrid system and the results showed that the proposed model can reduce the total load curtailment and maximum load curtailment.Load forecasting is one of the most important and studied topics in modern power systems. However, traditional load forecasting is an open-loop process as it does not consider the end use of the forecasted load. This dissertation proposes a closed-loop task-based day-ahead load forecasting model labeled as LfEdNet that combines two individual layers in one model, including a load forecasting layer based on deep neural network (Lf layer) and a day-ahead stochastic economic dispatch (SED) layer (Ed layer). The training of LfEdNet aims to minimize the cost of the day-ahead SED in the Ed layer by updating the parameters of the Lf layer. Sequential quadratic programming (SQP) is used to solve the day-ahead SED in the Ed layer. The test results demonstrate that the forecasted results produced by LfEdNet can lead to lower cost of day-ahead SED at the expense of slight reduction in forecasting accuracy.
Show less
- Title
- TOPOLOGY OPTIMIZATION OF SYNCHRONOUS ELECTRIC MACHINES
- Creator
- Guo, Feng
- Date
- 2021
- Description
-
Topology optimization of electric machine is attractive because of the increased design degree of freedom compared to conventional electric...
Show moreTopology optimization of electric machine is attractive because of the increased design degree of freedom compared to conventional electric machine design techniques. Also, a topology optimization approach does not necessarily require the use of a geometric template where dimensions are controlled by parameters. In this dissertation, a density-based magneto-structural topology optimization approach for the design of synchronous reluctance machine (SynRel), interior permanent magnet synchronous machine (IPMSM), and wound field synchronous machine (WFSM) rotors is developed. Depending on the electric machine type, the optimization problems are divided into single material and multi-material topology optimizations. A mass thresholding function is introduced to overcome the intermediate density issue which is caused by combining the magnetic and structural topology optimization problems. SynRel and IPMSM optimization examples are presented in the single material topology optimization section. For the multi-material topology optimization, in order to properly define the boundary conditions between multiple materials, a virtual region calculation approach is proposed. In the WFSM topology optimization, the copper field winding is represented by a virtual region. The contact and frictionless boundary conditions between the copper field winding and the electrical steel is defined and the centripetal load of the copper winding are equivalently calculated and applied on the elements on the electrical steel next to the boundary between the copper field winding and the steel of the WFSM pole tip. In additional to the total free-form magneto-structural topology optimization, a density-based combined dimensional and topology optimization is developed for the design of IPMSM and WFSM rotors. Both the dimensional and topological control variables are integrated to simplify the optimization problem. For IPMSM rotor design, the permanent magnet (PM) block shape is preferred to be retained where dimensional optimization could be used. The proposed dimensional topology optimization approach can fit in this design situation, where the PM is designed using dimensional control variables where the rest of the design domain is optimized using topology optimization. To allow the block or rectangular magnet to move and change size, the surrounding design domain mesh must deform or distort. The Laplace's smoothing mesh deformation technique is used in this approach and helper lines are connected to allow greater mesh deformation range and to avoid over mesh distortion. In addition to IPMSMs, a WFSM example is presented optimizing the winding region using dimensional optimization and the rotor core using topology optimization. An alternative combined dimensional and topology optimization approach has also been developed primarily for the design of the IPMSM rotors. In this approach, the mesh deformation is not required but there is no explicit geometric boundary between the rectangular permanent magnet and the surrounding electrical steel and air. In this approach, the PM density is expressed as a Heaviside rectangular function of dimensional variables. The function is projected onto the rotor mesh. Modified material penalizations are used. Topology optimization then controls the deposition of electrical steel and air. Three different IPMSM examples are presented with different dimensional control variables, including the PM position, size and angle.
Show less
- Title
- A Novel CNFET SRAM-Based Computing-In-Memory Design and Low Power Techniques for AI Accelerator
- Creator
- Kim, Young Bae
- Date
- 2023
- Description
-
Power consumption and data processing speed of integrated circuits (ICs) is an increasing concern in many emerging Artificial Intelligence (AI...
Show morePower consumption and data processing speed of integrated circuits (ICs) is an increasing concern in many emerging Artificial Intelligence (AI) applications, such as autonomous vehicles and Internet of Things (IoT). In addition, according to the 2020 International Technology Road map for Semiconductors (ITRS), the high power consumption trend of AI chips far exceeds the power requirements. As a result, power optimization techniques are highly regarded in nowadays AI chip designs. There are various low-power methodologies from the system level to the layout level, and we are focusing on transistor level and register transfer level (RTL) through this thesis. In this thesis, we propose a novel ultra-low power voltage-based computing-in- memory (CIM) design with a new SRAM bit cell structure for AI Accelerator. The basic working principle of CIM (Computing-in-memory) is to use the existing internal embedded memory array (e.g. SRAM) instead of external memory, and it reduces unnecessary access to external memory by calculating with internal embedded mem- ory. Since the proposed our SRAM bit cell uses a single bitline for CIM calculation with decoupled read and write operations, it supports much higher energy eciency. In addition, to separate read and write operations, the stack structure of the read unit minimizes leakage power consumption. Moreover, the proposed bit cell structure provides better read and write stability due to the isolated read path, write path and greater pull-up ratio. Compared to the state-of-the-art SRAM-CIM, our proposed SRAM-CIM does not require extra transistors for CIM vector-matrix multiplication. We implemented a 16k (128⇥128) bit cell array for the computation of 128x neurons, and used 64x binary inputs (0 or 1) and 64⇥128 binary weights (-1 or +1) values for the binary neural networks (BNNs). Each row of the bit cell array corresponding to a single neuron consists of a total of 128 cells, 64x cells for dot-product and 64x replicas cells for ADC reference. And 64x replicas cells consist of 32x cells for ADC reference and 32x cells for o↵set calibration. We used a row-by-row ADC for the quantized outputs of each neuron, which supports 1-7 bits of output for each neuron. The ADC uses the sweeping method using 32x duplicate bit cells, and the sweep cycle is set to 2N1 +1, where N is the number of output bits. The simulation is performed at room temperature (27C) using 32nm CNFET and 20nm FinFET technology via Synopsys Hspice, and all transistors in bitcells use the minimum size considering the area, power, and speed. The proposed SRAM-CIM has reduced power consumption for vector-matrix multiplication by 99.96% compared to the existing state-of-the-art SRAM-CIM. Moreover, because of the separated reading unit from an internal node of latch, there is no feedback from the read access circuit, which makes it read static noise margin (SNM) free. Furthermore, for the low power AI accelerator design, we propose a new AI accelerator design method that applies low power techniques such as bus specific clock gating (BSCG) and local explicit clock gating (LECG) at the register-transfer- level (RT-level). And evaluates them on the Xilinx ZCU-102 FPGA SoC hardware platform and 45nm technology for ASIC, respectively. It measures dynamic power using a commercial EDA tool, and chooses only a subset of FFs to be gated selectively based on their switching activities. We achieve up to a 53.21% power reduction in the ASIC implementation and saved 32.72% of the dynamic power dissipation in the FPGA implementation. This shows that our RTL low power schemes have a powerful possibility of dynamic power reduction when applied to the FPGA design flow and ASIC design flow for the implementation of the AI system.
Show less
- Title
- Electric Machine Windings with Reduced Space Harmonic Content
- Creator
- Tang, Nanjun
- Date
- 2023
- Description
-
The reduction of magnetomotive force (MMF) space harmonic content in electric machine windings can significantly improve the machine's...
Show moreThe reduction of magnetomotive force (MMF) space harmonic content in electric machine windings can significantly improve the machine's electromagnetic performance. Potential benefits include a reduction of torque ripple, a more sinusoidal back EMF, and reduced power losses. With the proposal of a uniform mathematical representation that applies to both distributed windings and fractional-slot concentrated windings (FSCWs), closed-form expressions can be derived for harmonic magnitudes, winding factors, etc. These expressions can then be used to formulate the MMF space harmonic suppression problem for windings, which looks for improved windings with certain harmonic orders reduced or even eliminated, by varying the slot distribution and coil turns. Different solution techniques are explored to gain additional insights about the solution space. The underlying mathematical relations between different harmonic orders are mathematically proved to establish the family phenomenon, which presents clear pictures of the higher order part of the harmonic spectrum and is the foundation for exact calculation of the total harmonic distortion (THD) of windings. The exact THD calculation further indicates how the minimal THD can be achieved for a winding. Windings can also be analyzed and designed from the view of subsets to incorporate distribution and excitation phase shift effects. With reduced or the minimal space harmonic content, new winding designs can help significantly improve the Pareto front when combined with motor geometry optimization. Design examples including a 12-slot 2-pole mixed-layer distributed winding, a 18-slot 2-pole mixed-layer distributed winding, and a four-layer 24-slot 22-pole FSCW with excitation phase shift are presented with finite element analysis (FEA) results to verify the performance improvements.
Show less
- Title
- Polarization Induced by A Terahertz Electric Field in A Semiconductor Nanodimer in the Overlapping Regime
- Creator
- Wang, Zi
- Date
- 2023
- Description
-
Boltzmann transport equation is a theoretical framework for the description of thermodynamics or charge reactions in a system not in...
Show moreBoltzmann transport equation is a theoretical framework for the description of thermodynamics or charge reactions in a system not in equilibrium, which can be applied to the analysis of the interactions of mobile charges with an electromagnetic wave. When the dimensions of the object are small compared to the wavelength, the induced dipole moment provides a means to characterize the collective response while providing insight to the nature of the charge-field interactions. Semiconductor nanoparticles exhibit surface plasmon resonance in the terahertz frequency range and are of current interest for the development of components and circuits in that part of the electromagnetic spectrum. By changing the plasmon frequencies of doped semiconductors through the change of carrier concentration, new opportunities arise for plasmonic manipulation in terahertz region leading to various promising applications. Despite the Drude model's long-term success and convenience in describing the electrical conductivity of metals in terms of dielectric functions, some aspects of polarization are not accounted for by bulk properties. By incorporating the transport equations of the charge carriers with Maxwell's equations, screening effects of charge carriers can be accounted for, enabling the internal field, space charge and induced dipole moment of a semiconductor nanoparticle to be studied.The computations performed for elementary dimer structures in overlapping cases revealed the internal field screening, while the complex dipole moments show dispersion and absorption effects. The numerical algorithms are implemented using the finite element method to investigate the surface plasmon resonance (SPR) induced on the semiconductor particles. Unique SPR modes evolution is observed as the thickness of the overlap region is varied. The characteristics can be interpreted by the migration of local space charge as the level of overlap is varied. This degree of freedom provided by a semiconductor nanodimer could be employed to control the local field near a simple cluster of nanoparticles, with potential for application in sensing and circuit components in the terahertz frequency range.
Show less
- Title
- Machine Learning for NDE Imaging Applications
- Creator
- Zhang, Xin
- Date
- 2023
- Description
-
Infrared Thermography and Ultrasonic Imaging of materials are promising non-destructive evaluation (NDE) methods but signals face challenges...
Show moreInfrared Thermography and Ultrasonic Imaging of materials are promising non-destructive evaluation (NDE) methods but signals face challenges to be analyzed and characterized due to the nature of complex signal patterns and poor signal-to-noise ratios (SNR). Industries such as nuclear energy, are constructed with components produced using high-strength superalloys. These metallic components face challenges for wide deployment because material defects and mechanical conditions need to be non-destructively evaluated to identify potential danger before they enter service. Low NDE performance and lack of automation, particularly considering the complex environment in the in-situation NDE and nuclear power plant, present a major challenge to implement conventional NDE. This study solves the problems of using the advantages of machine learning as signal processing methods for Infrared Thermography and Ultrasonic NDE imaging applications. In Pulsed Infrared Thermography (PIT), for quality control of metal additive manufacturing, we proposed an intelligent PIT NDE system and developed innovative unsupervised learning models and thermal tomography 3D imaging algorithms to detect calibrated internal defects (pores) of various sizes and depths for different nuclear-grade metallic structures. Unsupervised learning aims to learn the latent principal patterns (dictionaries) in PIT data to detect defects with minimal human supervision. Difficulties to detect defects by using PIT are thermal imaging noise patterns; uneven heating of the specimen; defects of micron-level size with overly weak temperature signals and so on. The unsupervised learning methods overcome these barriers and achieve the high defect detection accuracies (F-score) of 0.96 to detect large defects and 0.89 to detect microscopic defects, and can successfully detect defects with diameter of only 0.101-mm. In addition, we researched and developed innovative unsupervised learning models to compress high-resolution PIT imaging data and achieve the average high compression ratio >30 and a highest compression of 46 with reconstruction accuracy peak signal-to-noise ratio (PSNR) >73dB while preserving weak thermal features corresponding to microscopic defects. In ultrasonic NDE imaging, for structural health monitoring of materials, we built a high-performance ultrasonic computational system to inspect the integrity of high-strength metallic materials which are used in high-temperature corrosive environments of nuclear reactors. For system automation, we have been developing neural networks with various architectures for grain size estimation by characterizing the ultrasonic backscattered signals with high accuracy and data-efficiency. In addition, we introduce a response-based teacher-student knowledge distillation training framework to train neural networks and achieve 99.27% characterization accuracy with a high image processed throughput of 192 images/second on testing. Furthermore, we introduce a reinforcement learning based neural architecture search framework to automatically model the optimal neural networks design for ultrasonic flaws detection. At last, we comprehensively researched the performance of using unsupervised learning methods to compress 3D ultrasonic data and achieve high compression performance using only 4.25% of the acquired experimental data.
Show less
- Title
- A New Control and Decision Support Framework To Avoid Fast-Evolving System Collapse and Cascading Failure
- Creator
- Guha, Bikiran
- Date
- 2022
- Description
-
The modern power system is a vast and incredibly complex network with a very large number of equipment operating round the clock to reliably...
Show moreThe modern power system is a vast and incredibly complex network with a very large number of equipment operating round the clock to reliably transport electricity from generators to consumers. However, factors such as aging and faulty equipment, extreme and unpredictable weather, cyber attacks and increasing amounts of unpredictable renewable generation have made it increasingly vulnerable to cascading failure and wide-area collapse. Therefore, a lot of work has been done over the years on cascading failure vulnerability analysis and mitigation. However, to the best of our knowledge, the existing literature on this topic focus on preventive analysis and mitigation, mostly from a planning perspective. There is a lack of decision support schemes which can take real-time preventive action when the system becomes vulnerable to cascading failure, while taking into account the various dynamics and uncertainties involved in these types of failures. The only defense under these situations are pre-designed emergency control schemes. However, they are only effective against known vulnerabilities and can make matters worse if not accurately designed and calibrated.This research work has proposed a novel wide-area monitoring protection and control (N-WAMPAC-20) framework designed to make decisions in real-time to assess the vulnerabilities of the system (when a disturbance happens) and to implement mitigation actions, if necessary. The main contributions of this dissertation focus on the disturbance monitoring, real-time control and decision making aspects of this framework. The proposed framework has been divided into two major parts: an offline part and an online part. The offline part continuously runs extreme contingency analysis in the background (using combined dynamics and protection simulators) to generate elements which can assess system vulnerabilities and suggest suitable mitigation actions, if necessary. In this regard, a novel load shedding adjustment scheme is also proposed, which has been shown to be effective against a variety of fast-evolving cascading failure scenarios. The online part consists of real-time disturbance monitoring and decision-making components. The disturbance monitoring component focuses on real-time fault detection and location. If a fault has been identified and located, the real-time decision making component determines the vulnerability of the system, by consulting with the elements designed offline. If vulnerabilities are identified, targeted mitigation actions are implemented. The design and applicability of a prototype of N-WAMPAC-20 has been presented using a case of voltage collapse and a case of wide-area loss of synchronization on a synthetic model of the Texas grid.
Show less