34254 research outputs found
Sort by
Inverse Reinforcement Learning Architecture Base on Conditional Variational Autoencoder with Variational Information Bottleneck
No full textWhen agents engage in complex environments or confront challenging problems, designing an effective reward function becomes quite challenging. The difficulty lies in the fact that improperly designed reward functions may lead the agent to focus on short-term gains, potentially trapping it in local optima rather than achieving the desired long-term goals. Inverse Reinforcement Learning (IRL) offers a solution to this problem. By observing demonstrations from experts, IRL helps agents understand the behavior and decision-making processes of experts, enabling the determination of appropriate reward functions to closely mimic the behavior strategies of experts. Nowadays, IRL has found widespread applications in various fields, including robotics, autonomous driving, and healthcare, facilitating the generation of efficient control policies.
This paper presents an Inverse Reinforcement Learning framework based on Conditional Variational Autoencoder (CVAE). This framework allows CVAE to learn synchronously when observing demonstrations from experts. Agents can utilize latent vectors as new reward settings and apply Deep Q-Learning (DQN) for learning to achieve Inverse Reinforcement Learning. Additionally, the model proposed in this paper incorporates the Variational Information Bottleneck (VIB) technique. VIB not only trims parameters but also maintains model performance, reducing the complexity of feature spaces while retaining essential information about the data, thus achieving the effect of Knowledge Distillation. In the experimental phase, we evaluate the performance of the model proposed in this paper by conducting tests in three different game simulation environments and validate its effectiveness
Wireless Power Transfer Circuit with Double-Sided LCC Dual-Frequency Compensation
No full textBatteries are usually charged by constant-current and constant-voltage modes. As contactless charging is gradually adapted, this thesis proposes a double-sided LCC wireless power transfer resonant converter, which toggles the resonant tank behavior to achieve constant current and voltage output by changing the switching frequency. Both modes also can achieve input zero phase angle to promote the circuit efficiency. In this thesis, the converter consists of a full-bridge inverter, a loosely-coupled transformer, transmitter and receiver compensating resonant network, feedback circuit, and an MCU. The MCU collects the output voltage and current to determine the charging mode and applying phase-shift control method in full-bridge to stabilize the output.
This thesis analyzes two output modes of the resonant tank to achieve the constant output voltage or current by changing load, and used ANSYS Maxwell simulation software to simulate the loosely-coupled transformer by according to different distance of wire, width of ferrite, and thicknesses of ferrite, and finally build a 1500 W double-sided LCC resonant converter, which can maintain at 5 A in constant current mode and the output voltage can maintained at 300 V in constant voltage mode. The maximum efficiency in constant-current mode is 92.1% and 90.2% in constant-voltage mode
Uplink Grant-Free Sparse Code Multiple Access Schemes for mMTC Applications
No full textSparse Code Multiple Access (SCMA) is a non-orthogonal multiple access (NOMA) technology that supports large-scale connections, thereby improving the spectrum efficiency. This paper is mainly divided into three parts: The first part focuses on an uplink grant-free scenario. We propose a depth-first tree search algorithm for the joint active user detection and signal demodulation. By incorporating an iterative processing algorithm, an enhanced error rate performance can be achieved. The second part employs deep learning techniques for the optimization of SCMA codebook. In third part, the coded architecture is investigated, and error rate performance can be further enhanced by introducing outer error correcting codes (ECC) of different code lengths. Eventually, uplink grant-free SCMA schemes are constructed for massive machine type communication (mMTC) applications
Resolution Improvement in Single Molecule Localization Microscopy by Carbon Dots
No full text\ue3\ue3Super resolution microscopy is a crucial tool for researching the cell physiology. Among the rest of the techniques, single molecule localization microscopy (SMLM) randomly switched fluorescent molecules between bright and dark states. At last, images were analyzed to detect the sparse on state single molecules and determine the subpixel positions at each frame. All of these localizations were accumulated to construct a high contrast and resolution image. Scientists could observe nanoscale fine structures in the cells. The conventional organic fluorophores such as Alexa Fluor dyes would have the stable photoblinking properties only if they dissolved in the buffer which contained enzymatic oxygen scavengers and thiols. However, these additives were likely to interference the cell behaviors when we visualized the living cell fluorescence images. Therefore, this study showed that rhodamine B and polyethylene glycol 400 synthesized green emissive carbon dots with spontaneous photoblinking through hydrothermal reaction for 8 hours. Carbon dots possessed the highest quantum yield were purified by reversed-phase high-performance liquid chromatography (RP-HPLC). They emitted the large number of photons in comparison with low signal to noise ratio of Alexa Fluor 488. The full width at half maximum of the single particle achieved 36 \uc2\ub1 3.1 nm in the super resolution image. Finally, carbon dots which were modified with triphenylphosphonium (TPP+) derivatives successfully stained the mitochondria in the living cell to confirm that this nanomaterial was worth serving as a fluorescent dye in SMLM. We can also obtain the complete high resolution mitochondria image. What's more, we developed an autofocus system in case focus drift occurred for a long-term photography due to the pull of gravity on the objective. Near-infrared laser was used to generate total internal reflection (TIR) at the boundary between the coverslip and the specimen, then reflected to the quadrant photodiode (QPD) detector. The position of the reflection beam was different when the objective fell, resulting in a difference of the readout voltage. The nanopositioner was controlled by tuning the best proportional-integral-derivative (PID) parameters. The compensatory defocus distance which made the voltage of the reflection beam consistent accomplished long-term instant autofocus
Die Design and Finite Element Analysis in Wheel Frame Forging Process of Aluminum Alloy AL 6061
No full textThe traditional method of producing automobile wheel rims often involves casting. However, literatures suggest that hot forging can lead to improved dimensional accuracy and enhanced mechanical properties. This study aims to investigate how different process parameters of hot forging impact the formability of high-temperature forged aluminum alloy 6061. Additionally, the choice of lubricant for the mold surface significantly influences product formability. Effective lubrication can minimize mold surface wear, enabling a smoother flow of the initial billet material and achieving the desired dimensions and shapes. Moreover, the lubricant also influences the demolding force.
As a result, initial hot ring compression tests will be conducted on aluminum alloy 6061 to determine the friction coefficient with various lubricants and the plastic flow stress of the alloy. The experimental data, including friction coefficient and plastic flow stress from the hot ring compression tests, will be integrated into finite element analysis for the hot forging of the wheel rim. This will help us understand the plastic deformation behavior of the billet within the mold cavity.
Apart from exploring the friction's impact on wheel rim formability, simulations will consider different billet size ratios, forging speeds, and temperatures. Subsequently, by employing the optimal forming conditions, we will proceed with mold design and fabrication, followed by aluminum alloy wheel rim hot forging experiments. The experimental outcomes will be compared against analytical values to validate the applicability of the analytical model. Ultimately, the objective is to achieve high product precision and reduce forging force through this innovative forming method. The comparison between experiments and simulations reveals that simulations exhibit a high level of accuracy in predicting the external dimensions of the product. The simulation results can serve as a valuable reference for process design and the selection of stamping equipment. Through metallographic observations and hardness experiments, information about the grain size and hardness distribution at various locations of the wheel frame has been obtained
The Synthesis and Characterization of Various Amino Acid-Containing Polysulfides Materials
No full textInverse vulcanization has been a method for synthesizing high-sulfur content copolymers in recent years. This reaction involves the use of elemental sulfur and an olefin crosslinker for polymerization. The features of this reaction including solvent-free, high atom economy, and readily available and inexpensive material, which is beneficial for the development of green chemistry. Recent research has shown that binder modified with cysteine can improve the performance of Li-S batteries. Thus this research investigate the use of cysteine containing crosslinkers in inverse vulcanization and applying the materials in Lithium-sulfur batteries. Lithium-sulfur batteries, due to their high energy density and environmental friendliness, are expected to become the new generation of energy storage batteries. However, the shuttle effect during the charging and discharging process leads to a significant decrease in their cycle life. Therefore, this study investigates the effect of polar functional groups of cysteine in reducing the shuttle effect in Li-S batteries as well as how material properties vary, electrochemical performance with respect to the degree of unsaturation of the crosslinker used in inverse vulcanization. From the experimental results show that at 0.4 C, the initial discharge capacity of 841.6 mAh/g. After 100 charge-discharge cycles, the capacity still retain 354 mAh/g
The Impact of Populist Messages by Candidates on Users\ue2 Emotions and Engagement on Facebook: A Case Study of the 2020 Taiwan Presidential Election
No full textPopulism became a hot topic in the 2020 presidential election, mainly due to the campaign strategy and behavior of the Kuomintang (KMT) presidential candidate, Han Kuo-yu. His characteristics of people-centrism and anti-establishment align to a considerable extent with the definition of populism. With the rapid development of online media, candidates have matured in their ways of promoting and marketing themselves on Facebook. However, there is still a lack of mature research examining the effects of populist messages and the characteristics of social media on netizens and the broader impact on the public.
This study attempts to use manual coding to capture elusive populist messages that are challenging to measure. The study investigates the specific impact on netizens' emotional reactions and engagement behaviors. The research found that the higher the degree of using populist messages in candidates' Facebook posts, the significantly more it enhances netizens' feelings of intimacy, anger, anxiety, positive emotional responses, and also significantly increases netizens' engagement behaviors (likes, comments, shares)
Performance enhancement of micromotors by applying sol-gel nano-ceramic coatings
No full textThis research focuses on the additional energy consumption and efficiency reduction caused by heat loss and corrosion problems during the operation of the motor. Use sol-gel method (Sol-gel) and organic/inorganic nano-ceramic hybrid resin prepared by polymer to replace the original organic resin coating on the silicon steel sheet. The new coating has better weather resistance and high temperature resistance. It is expected to increase the service life and performance of existing motors from the material side. The thesis is mainly divided into three stages. The first is the synthesis of nano-hybrid resins. Colloids are synthesized through catalysts with different pH values. After determining the viscosity and adhesion, organic resins are added to make the colloids graft and cross-link. With the goal of improving the thermal conductivity of the motor and increasing its lifespan, silicon oxide, aluminum oxide, and boron nitride are selected as high thermal conductivity and heat dissipation fillers. Then it is coated on silicon steel sheets and copper substrates. The performance of the coating as insulating layer is observed in the adhesion test and dielectric strength test. The second stage is to evaluate the performance of the mixed resin and adjust the process parameters. Through the salt spray-high temperature cycle test of different durations, we can observe the degree of corrosion and compare the performance difference with the commercially available coatings. In addition, after pressing the coated silicon steel sheet, use a universal tensile machine to measure its shear strength. After confirming that the coating meets the rapid self-adhesive film strength standard, glue it into a block to measure the thermal conductivity and compare the performance of heat dissipation with the traditional welded silicon steel sheet blocks and commercially available self-adhesive silicon steel sheet blocks. According to the measurement results of the thermal conductivity meter, the alumina coating can increase the thermal conductivity of the silicon steel sheet by 33.14%. The final stage is to build a dynamometer platform to compare the motor efficiency difference between different coatings. Finally, the convergence temperature at the same rotational speed but different power of each motor will be tested to judge the heat dissipation effect. According to the measurement results at a rotational speed of 1400RPM, the output power of the aluminum oxide coating motor is 63.5W when the temperature converges to 50 degrees, which is about 29.6% higher than the 49W of the traditional welding motor. This study reduces the impact of motor heat accumulation and successfully improves the effect of traditional insulation layers and successively improves the motor efficiency of 6.7% comparing to welding motor as control group
Development and Physical Mechanisms Establishment of Flexible Low Temperature Polycrystalline Silicon and Organic Thin Film Transistors
No full textRecently, the application of flexible thin-film transistors has become more and more striking, which leading many R&D and researchers to put more effort into developing related display products. With the demand in the market of display, smart phones have evolved from the traditional flat display to the fixed curvature boundary and the smart phone that can be repeatedly bent and folded. However, there is still room for improvement in the bending endurance and electrical reliability of the flexible thin film transistor. The topics discussed in this dissertation include the bending endurance of flexible low temperature polysilicon thin film transistors under mechanical bending stress, the physical mechanism of flexible organic thin film transistors after bending, and studies the electrical deterioration of organic thin film transistors with multi-Finger structure.
The first part of this dissertation discusses low-temperature polysilicon thin film transistors on flexible polyimide (Polyimide) substrates to improve the endurance of mechanical stress by improving the quality of the oxide layer. Most of the current flexible thin film transistors use polyimide as the substrate, which not only has flexibility but also has good thermal stability and chemical properties. In addition, the polyimide can be integrated with the solution process, so it can be deposited by spin coating. However, due to the difference Young\ue2s moduli between deposited layers, the result of the low-temperature polysilicon thin film transistor with the flexible substrate shows the strain will be concentrated at the edge of the gate insulator under the mechanical bending stress, that cause the electrical deterioration of the LTPS TFTs. Therefore, the quality of the oxide layer is an important key for LTPS TFTs.
In this chapter, the use of chemical vapor deposition (PECVD) to deposit the gate oxide layer is studied to improve the quality of silicon dioxide by introducing different helium concentrations, thereby enhancing the electrical reliability and mechanical stress endurance of low-temperature polysilicon thin film transistors. Through the extraction of parameters to investigate the effect of different helium concentrations on PECVD deposition the gate insulator, such as subthreshold swing, carrier mobility, electrical measurement, and mechanical stress test. Finally, transmission electron microscopy (TEM) and energy-dispersive X-ray spectroscopy (EDS) to analyze element changes between materials
In the second part, the electrical characteristics of organic thin film transistors after mechanical bending stress in a series of different atmospheres are investigated. In atmospheric environments, severe degradation behavior of the threshold voltage (VT) can be observed after 100,000 channel width axial compressive stress bends and a long-term fixed curved with a bending radius of R=5 mm. Following annealing after bending, VT recovered to a more negative value than the initial VT for both dynamic and fixed curved bending. Performing bending tests in a vacuum, which eliminates the influence of the ambient atmosphere on the OTFTs, shows that the degradation of the device under atmospheric bending is dominated by two mechanisms. Sequentially introduce different gases (humidity, nitrogen, carbon dioxide, oxygen) to clarify their effect on OTFTs. According to the experiments described above, the effects in the vacuum and atmosphere, we are able to separate the physical effects of mechanical bending on the OTFTs from the changes induced by exposure to environmental factors. Due to different mechanical bending conditions, small molecules on the surface of the active layer can induce different electrical properties. Finally, physical models of mechanical bending and atmospheric factors of organic TFTs are proposed.
The third section investigates the effect of Joule heating on organic thin-film transistors (OTFTs) with multi-finger structured under high-current operation. The electrical degradation of OTFTs with different numbers of fingers (N = 4, 6 and 10) is discussed. Significant electrical degradation was observed at high current due to heat accumulation during operation. Under the operation of high current and voltage, with the increase of the number of fingers, the electrical characteristics show different degrees of deterioration, and an abnormal hump is observed in the OTFTs with the N= 10. In order to verify the influence of self-heating effect on OTFTs, OTFTs were fabricated on materials with different thermal conductivity, polyimide and glass substrates, and the causes of degradation were investigated at both room temperature and low temperatures. In addition, AC operation is proposed to change the heating and cooling time. Finally, Silvaco TCAD simulation is used to simulate the self-heating effect of the OTFTs devices, and the effect of the number of fingers on the bias stress is explained, and the reason for the deterioration is proposed.
Index Terms: Flexible electronic devices, Low temperature polycrystalline silicon TFT, Organic TFT, Vacuum bending measurement, Atmosphere measurement, Hopping distance, Multi-finger structure, Self-heating effect
Offshore wind power underwater foundation supplier selection
No full textIn recent years, Taiwan has shown a notable commitment to the advancement of renewable energy, with a particular focus on offshore wind power. The construction of offshore wind power projects, particularly the underwater foundations, has been a central initiative in this pursuit. These foundations play a pivotal role in offshore wind farm development and constitute a significant cost factor. Given the scales and complexities of developing a new wind farm, standardizing the selection criteria by developers would provide clearer guidelines to suppliers, enabling them to better understand the specific requirements and expectations for participating in wind farm projects. This standardization would empower suppliers to make informed decisions regarding their capabilities and invest in areas that require improvement.
This study aims to analyze the factors involved in supplier selection for offshore wind power underwater foundations, by determining the weighting and ranking of each factor in the selection process. It considers the roles played by different stakeholders within the supply chain, including the developers, first- and second-tier manufacturers. The research methodology employs the modified Delphi method to identify the key factors for supplier selection. Based on the identified indicators, a comprehensive FAHP questionnaire is conducted for industry experts to provide their opinions and fill out the questionnaire.
The FAHP questionnaire results have revealed significant differences in the perception and priorities among the various stakeholders involved in supplier selection. Through a thorough analysis of these results, this study highlights the varying levels of importance placed on different factors by different stakeholder groups. This valuable insight not only helps decision-makers gain a clearer understanding of the key considerations in supplier selection but also provides a platform for suppliers to comprehend the gaps in perception between themselves and the evaluators. Armed with this knowledge, suppliers can address areas of improvement and enhance their competitiveness accordingly. By facilitating a better understanding of the supplier selection process, this research serves as a valuable tool for evaluators to make more informed decisions and for suppliers to enhance their competitiveness in the industry. Ultimately, the study contributes to the overall improvement and efficiency of the supplier selection process for offshore wind power underwater foundations in Taiwan