56016 research outputs found
Sort by
Improving Binarized MobileNet by Searching Group Parameters and Inter-Block Skip Connections
No full textGraduate School of Artificial Intelligence ArtificiIn recent years, thanks to the efficiency of the structure, the several MobileNetV1-based binary neural networks(BNNs) are proposed. However, most of them replaced the depth-wise convolution(DWC) with the vanilla convolution due to the significant performance degradation though it hurts the efficiency of the MobileNetV1 structure. In this paper, we replace DWC with grouped convolution(GC) and propose a new hyperparameter, group divisor, to deal with the number of groups in GC and search optimal trade-off between computational complexity and performance. Furthermore, we mitigate the performance degradation by adding inter-block skip-connections to the network without any additional parameters. We propose new hyperparameter, block interval, which is the interval of the skip-connection between blocks, to search optimal inter-block skip-connections. As a result, based on ReActNet, our method achieves more compact network with comparable accuracy on CIFAR and ImageNet dataset.clos
Multipurpose Solid Additives for Efficient and Stable Organic Solar Cells
Get PDFSchool of Energy and Chemical Engineering (Energy Engineering)Organic solar cells (OSCs) are one of the efficient and practical ways to use solar energy as a green energy source. Based on several advantages of organic materials, organic solar cells have attracted great attention as next-generation solar cells. However, many researchers are still needed for commercialization of organic solar cells. There are three key points to fabricate efficient solar cell systems: efficiency, stability, and reproducibility. With these three keywords, researchers are conducting various studies such as the introduction of a new donor or acceptor, and interlayer engineering. Among several approaches for efficient solar cells, the introduction of additive (only small amounts) into the active layer is one of the facile methods owing to low-cost, easy processing, and unique property. In addition, this promising approach is expected to improve the power conversion efficiency together with several kinds of stability. Thereby, in this dissertation, I look at the role of additives in OSCs by applying newly synthesized donor materials using the commonly used solvent additive. Based on this, I design the multipurpose solid additives for organic solar cells based on each characteristic by categorizing additives into three types (nonvolatile polymer additives, nonvolatile small molecule additives, and volatile small molecule additives). Furthermore, I investigate the influence of solid additives on performance, stability, and reproducibility. In the first study, with the increasing concern for discovering a new processing additive, the effects of polymer additives (polystyrene, poly(styrene-b-pentafluorostyrene) and poly(pentafluorostyrene)) on the bulk heterojunction blend system were thoroughly investigated to obtain a direct comparison with the widely used volatile solvent 1,8-diiodooctane solvent additive. In the second study, dibutylhydroxytoluene (BHT)-based nonvolatile antioxidant additives with polar cyanide (CN) and perfluorinated alkyl chains (designated as BHT???CN and BHT???PF) are developed, demonstrating that the OSCs will have significantly improved long-term stability by using them when exposed to the H2O, O2. In particular, the use of BHT???PF in the various given test-bed OSC systems can remarkably enhance the long-term stability, as well as the high initial PCEs similar to the maximized values obtained from the highly optimized OSCs with each well-known suitable solvent additive. Last study, I carry out a comprehensive investigation into the effect of benzothiadiazole (BT) and its fluorinated analogs (FBT and 2FBT) as solid additives on the device performance of layer-by-layer (LBL) platform. The use of FBT in the donor layer results in a suitable morphology that ensures efficient charge transport/generation properties and suppresses recombination loss, boosting the photovoltaic performance of the LBL device. These findings are not only invaluable in shaping our understanding of OSCs, but also provides the possibility of development of next-generation solar cell technologies.ope
Survey on the converse theorem
Get PDFDepartment of Mathematical SciencesThe task of analyzing the properties of a given function is attempted in many mathematical studies. Conversely, one may ask the question of which of those properties is based on to specify the original function.
In the number theory, Hecke and Weil completed the characterization of the classical modular form, based on Hamburg's monumental theorem on the Riemann zeta function. This is called a converse theorem, and studies on variable automatic form have been actively conducted until recently.
In this thesis, we fi rst examine the proof method of the converse theorems in modular forms and Maass forms. Then by applying these methods, we will survey the converse theorem for harmonic Maass forms. We will also briefly review follow-up studies on these theorems.ope
Verification and Validation of Whole Core Three Dimensional MOC Code STREAM3D with Watts Bar Unit 1 Multi-cycle Operation
No full textDepartment of Nuclear EngineeringThis research verifies and verifies the STREAM3D (Steady State and Transient Reactor Analysis code with Method of Characteristics) with Watts Bar Unit 1 multi-cycle operation. A direct neutron transport analysis code, STREAM3D, has been developed to perform pressurized water reactor (PWR) core calculation. The 3D MOC/DD (Three-dimensional Method of Characteristics/Diamond-Difference) method is implemented in STREAM3D to solve the neutron transport equation for 3D reactor core problems. In the scope of verification and validation (V&V), Organization for Economic Co-operation and Development (OECD) Nuclear Energy Agency (NEA) published Watts Bar Unit 1 Cycle 1-2-3 3D whole core problems. The Watts Bar Unit 1 is Westing-house type PWR with 3411 MW thermal power. These problems comprise seven whole-core exercises: multi-cycle, multi-physics, depletion, and stand alone problems. Compared parameters during the zero-power physics testing in all given cycles: ITC, DBW, CRW, critical boron concentration (CBC). STREAM3D results show good agreement with measured data and code systems. The verification and validation of STREAM3D are conducted at hot full power conditions along with various feedback required for 3D whole core simulation, such as depletion, equilibrium xenon feedback, and CBC search. The influence on parameters, including CBC values, normalized radial assembly power distribution, and 1D core axial power distribution.
STREAM3D is compared against the measured data and KENO-VI code in the benchmark for Cycle 1. Additional comparisons are made with the two-step nodal diffusion code STREAM/RASTK v2 and Monte Carlo Code MCS for all given cycles. STREAM3D shows the RMS error of assembly power within 1.3% and RMS error of CBC within 22 ppm for Cycle 1. RMS error of assembly power within 1.8%, and RMS error of CBC within 31 ppm for Cycle 2. STREAM3D shows the RMS error of assembly power within 2.8% and RMS error of CBC within nearly 50 ppm for Cycle 3. A good agreement is observed at 1D core axial power during Cycle 1. A little higher tendency at 1D core axial power is observed in Cycle 2 and Cycle 3. Multi-cycle, multi-physics, and stand-alone capability of STREAM3D has been verified and validated against Watts Bar Unit 1 multi-cycle operation.clos
??????????????? ??? ????????? ???????????? ???????????? ??? ???????????? ?????? ?????????
Get PDFDepartment of Mechanical EngineeringThis work presents fault detection and recovery system when the bias fault occurs in the yaw rate part of the gyro (z-gyro) during the quadrotor flight. Kalman filter based on the dynamic model of the quadrotor and chi-square test is used to detect the fault. The auxiliary yaw rate was calculated through the homography matrix between the image frames obtained by the camera which is filming the ground plane. This was used as a redundant z-gyro sensor and switched with the faulty sensor to perform the recovery process. Additionally, in constant bias fault scenario, camera is used to correct biased sensor data. Through this, it was possible to follow the path stably even if there was a bias during flights. The system was implemented with the ROS environment and simulated in Gazebo.ope
Development of a virtual metrology system for smart manufacturing: A case study of spandex fiber production
No full textVirtual metrology (VM) is one of the most important enabling technologies in smart manufacturing. Although there is an abundance of literature on VM applications, the context of continuous production has received less attention. Fundamental challenges involved in the application of VM to a continuous process have been overlooked in comparison with intermittent (or batch) processes. Here, we described a real-world VM system for the manufacturing of spandex fiber, focusing on how practical challenges associated with a continuous process, including time synchronization, recirculation process, and autocorrelated features, can be addressed using data analytics. A model refresh strategy is discussed for the deployed VM system to ensure continuous usability and high-quality predictions. The virtual-physical connection established by the VM system creates a virtuous cycle in which constantly updated data render the system realistic, and valuable insights generated by the system can be applied to the physical production environment
Price impact in Nash equilibria
No full textWe prove global existence of a continuous-time Nash equilibrium with endogenous persistent and exogenous temporary price impact. Relative to the analogous Radner and Pareto-efficient equilibria, the Nash equilibrium has a lower interest rate but has similar Sharpe ratios and stock-return volatility
Modeling and Simulation of a Compression Molding Process for Aircraft Structures with Recycled Thermoplastic Composites
No full textAdvanced thermal fluid leakage detection system with machine learning algorithm for pipe-in-pipe structure
No full textPipe-in-pipe (PIP) system is essential for high thermal and high pressure fluid transportation. However, in the existing PIP systems, fluid leakage between inner and outer pipe has been difficult to discover or detect, which has worked as bottle neck to utilize PIP system in high risk industries as nuclear reactor, chemical plant or oil drilling systems. Here, we propose a noble PIP leakage detection system utilizing distributed temperature sensing (DTS) with Machine Learning (ML). With the Fourier transformed spectrogram data from DTS, the ML assisted system was able to detect 0.2 similar to 7 ml/min liquid leakage between inner and outer pipe with the accuracy of 91.67% with a single embedded optical fiber. Under varying operating temperature, the system successfully distinguished leakage and non-leakage states using the optimized convolutional neural network. Our developed PIP leakage detection system can be deployed in safety-critical industrial systems for autonomous leakage detection
