Search

Event detection (ED) aims to identify events and their types from the given text, which is the vital subtask of event extraction. However, poor quality training corpus can seriously affect the validity of current dominating deep learning-based techniques. In order to improve the robustness of the model, this paper converts the ED task into the named entity recognition (NER) task, and uses the span-based NER framework to identify and classify event triggers. In addition, we ameliorate the framework through data cleaning strategy, hierarchical learning rate strategy, sentence-level loss function and model ensemble strategy. The experimental results suggest that our approach achieves state-of-the-art performance on the widely used Chinese dataset. Moreover, the deep analysis has confirmed the effectiveness of the improvement strategy in resisting data pollution.

Based on the aerodynamic layout of special projectiles, a targeted aerodynamic parameter solution algorithm is built by a mechanical analysis model. The simulation software is used to analyze the ballistic stability. The results show that the aerodynamic layout with the center of gravity moving down is helpful to overcome the rolling interference in uncontrolled flight.

Power big data is the practice of big data concepts, methods, and technologies in the power industry. It is composed of structured data and unstructured data. It mainly involves power generation, transmission, transformation, and distribution. Power big data can reflect the macro-economy in real time, promote economic and social development, and improve the internal management level and economic benefits of the power industry enterprises. The integration of power big data technology and power system simulation calculation can update system analysis in real-time and provide new perspectives and methods for system analysis.

Real-time big data analysis provides a better value than standard offline big data analysis. Academic groups are still researching and exploring the technologies of real-time big data analysis. Due to the vast scale and poor transmission rate of big data, there are several impediments in the research process, making it hard to fully access the potential value of data.A real-time large data visualization analysis platform based on Flink is presented to overcome this challenge.The platform first collects data in real time and sends it to the message middleware Kafka, and then the big data analysis platform reads the data stream in Kafka in real time and transmits it to the multidimensional operator developed with Flink as the framework for real-time Calculate, and uses WebSocket technology to push the calculation results to the front end in real time, and perform visual display, enabling real-time monitoring and analysis of data flow.The experimental findings suggest that the real-time big data analysis platform based on the integration of Flink and Kafka can accurately, promptly, and clearly portray the real-time data flow process and visual analysis.

A power information vulnerability detection system has the problem of a low network survival rate. Therefore, a power information vulnerability detection system based on network traffic identification technology is designed. In the hardware part, the memory is configured as a synchronous interface and 4 DRAMs, and the level input and output power of the chip are kept in a state compatible with LVTTL levels; in the software part, the power information network indicators are obtained, the importance of network elements is reflected through a mesoscope, the failure characteristics of components are extracted using network traffic identification technology, a critical area is delineated, and the port protocols and association libraries of the passages are matched to optimize the system vulnerability detection function. Through analysis of the simulation results, it can be obtained that the network survival rate of the power information vulnerability detection system in this paper is 17.657% and 17.931% higher than that obtained by the other two power information vulnerability detection systems, respectively, indicating that the designed power information vulnerability detection system is more effective when fully integrated with network traffic identification techniques.

This paper introduces the types, internal structures and working principles of common non-electricity protection devices, and theoretically analyzes the cooperation between non-electricity protection devices. For pressure relief devices, the requirements of relevant standards for pressure relief valves are introduced, matters needing attention in field use of pressure relief valves are summarized, and factors affecting the performance of pressure relief valves are analyzed. Finally, using the calibration platform, through a large number of experiments, the influence of external environment temperature, caliber and pressure release times on pressure release performance parameters was studied. The results show that the average opening pressure at high temperature (95°C) is significantly lower than the average opening pressure of 10 times at normal temperature. At the same time, with the increase of ambient temperature, the aging opening pressure and the average opening pressure will obviously decrease, while the average closing pressure will increase. In addition, under normal temperature environment, the test values of aging opening pressure, average opening pressure and average closing pressure have little change. For pressure relief valves with different calibers, under the same rated opening pressure, the calibers of the pressure relief valves are different, which has little influence on the actual opening pressure. Through the research on the operation reliability of qualified pressure relief valves, it is found that the opening pressure after 50 operations and the first measured opening pressure and closing pressure data after 500 operations are relatively stable.

This paper describes an event of zinc oxide arrester breakdown at low voltage side due to single-phase grounding at medium voltage side of 220kV transformer. Combined with the system wiring mode, equipment damage, system background and test data, the causes of fault and the transmission mechanism of overvoltage are studied in detail, It is concluded that the single-phase grounding short circuit on the medium voltage side leads to the displacement voltage of the neutral point, and the overvoltage is transmitted to the low voltage side through the capacitive coupling between the transformer windings, resulting in the breakdown of the arrester. The correctness of the conclusion is verified by calculation, and reasonable rectification measures and suggestions are put forward to effectively limit the generation of overvoltage on the low-voltage side of the transformer and ensure the safe and stable operation of the power grid.

Imaging circuits of converting light signals to electrical signals is designed for different remote sensing satellites with different sensors in space applications. Based on a TDI CCD sensor, this paper presents the detailed design and implementation of imaging circuits which is more suitable for use on commercial satellites. This paper utilizes industrial-grade electronics to achieve imaging circuits, which becomes effective solutions to reduce the costs of remote sensing satellites. The signal processors receive analog output signals of TDI CCD to convert digital signals. The core processor, FPGA, receives and processes digital signals them from the signal processors and delivers to transceivers. The transceivers receive the image data from FPGA and deliver them to the host computer. Moreover, a schematic of the completed board of TDI CCD imaging circuits is presented. Finally, to verify the performance of the imaging circuit, the HCTF of TDI CCD was measured on this CCD imaging circuit. The measured result is consistent with the product manual, which shows that the imaging circuits perform well.

For the acquired visible and infrared missile-borne images, in order to provide accurate and reliable image data for the subsequent processing of images to achieve complementary information, we carried out visible and infrared missile-borne image registration research. Firstly, the shape context is used to extract the rough edge features of the image. Then we improve the SIFT gradient definition to overcome the difference in image gray value, and combine with the coherent point drift algorithm to increase the number of correct matching points to achieve global image registration. Experimental results show that, compared with the existing registration methods, the proposed method has a good visual effect of registration, greatly improves the correct registration points and registration accuracy, and can better solve the problem of visible infrared missile-borne image registration.

When the wireless charging system charges the lithium battery, in order to reduce the size of the receiver. With the increase of the charging power of the lithium battery, the charging strategy of the lithium battery should be changed accordingly, so as to improve the performance of the lithium battery. According to the characteristics of the wireless charging system and the charging strategy of the lithium battery, the whole system is dynamically adjusted based on the fuzzy logic control principle, so that the system can provide the required charging power for the lithium battery under different charging strategies. In this paper, the circuit model of the wireless charging system and the equivalent model of the lithium battery are established respectively, and the fuzzy logic control strategy is proposed. Finally, the accuracy of the control strategy is verified by simulation, which provides a reference for the dynamic adaptive adjustment of the wireless charging system.

The inducer is a part with a complex curved structure and is widely used in the energy, aerospace and petrochemical industries. To improve the efficiency of the inducer machining, a 5-axis machining path was prepared using the CNC machining programming software NREC MAX-PAC. The NC code was exported to VERICUT CNC simulation software through a customized post-processing program by IMSpost to verify the validity of the CNC program. Compared with the original machining time, the actual machining time is now shortened by 43.18% after optimization, which realizes the efficient milling of complex inducer structure.

With the development of science and technology, surveying and mapping technology has also advanced by leaps and bounds. More and more people in related industries have begun to pay more attention to the application of drone tilt photogrammetry to the surveying and mapping industry, which has greatly improved the progress of surveying and mapping projects. This photography technology has gradually become an indispensable part of the surveying and mapping industry due to its advantages of speed, efficiency, flexibility, and low cost. This article is aimed at the integration of premises and land project, using a drone platform equipped with 5 lens cameras and shooting from 5 directions at the same time, that is, through 1 vertical shooting lens and 4 tilting lenses to obtain ground surface images, to achieve a large overlap of multiple angles For viewing angle, the requirement for overlap is that the heading overlap is at least 60%, and the side overlap is 30%. This article uses ContextCapture Center Master software to perform three-dimensional modeling of the acquired images and obtain orthophoto results. The three-dimensional model is loaded into the EPS software for collection. The buildings, eaves, balconies and other buildings in each parcel of land are collected, and they are finally obtained. This technology greatly improves the work efficiency of the real estate integration project.

Refinement, intelligence, and standardization are the core requirements of today’s urban construction management. From the current situation, it is impossible to meet the needs of future urban construction and management by relying on a single technology. The integration of GIS and BIM provides new opportunities for urban construction management. GIS provides a basic framework for the construction and management of today’s cities, and BIM provides an accurate information model of a single building for urban construction management. The integration of GIS and BIM will bring new ideas and methods to city construction and management. A large number of high-precision BIM models provide detailed building information for city management and provide richer information for the construction of city 3D models. Although the fusion research of GIGIS and BIM has clear directions and urgent needs, the fusion research work of the two is still in the initial and exploratory stage.

This paper introduces the items and principles of transmission line parameter test, finds out the interference factors through the detailed analysis of the field test results, and puts forward the relevant precautions.

In this paper, ISE-TCAD, a device simulation software, is used to carry out simulation experiments on silicon-based trench MOS Schottky (TMBS) diodes with different trench dimensions under the condition of forward biases and obtain a series of experiment data associated with the negative resistance. Simulation results show that TMBS diodes under forward biases produce the negative resistance effect. The negative resistance effect intensifies linearly with the increase of mesa width and mitigates wavily with the increase of trench depth. The coupling effect of trench oxidization layer on electric field attenuates the change in device field strength, leading to the fact that TMBS devices present a weaker negative resistance effect than trench barrier Schottky (TBS) diodes which have the same parameters, but are free of oxidization layer. The study result provides a feasible basis for the effective application of TMBS negative resistance features to the design and production of corresponding semiconductor devices and circuits.

As a typical framework in ensemble learning, Stacking creates a multi-layer learner by integrating the learning results of different individual learners. Despite a general improvement in accuracy, multiple layers in stacking results in high computation cost. In addition, some individual learners do not perform better than others in single-model prediction task. To solve the above drawbacks of traditional stacking, this paper proposes an improved Stacking algorithm based on the concept of “weighted attribute”. For each individual learner, a corresponding weight will be calculated based on its prediction performance on training data. Weights will then be used as evaluation criteria for model performance. Models with larger weights will be selected as attributes for meta-learner. When improved Stacking algorithm is applied on the same dataset, experiment results indicate a slight improvement in prediction accuracy and a relatively large reduction in computation cost. This demonstrates practical application value of the improvement.

The negative resistance characteristics of Trench MOS Barrier Schottky (TMBS) diodes were studied by simulation and theoretical calculation. A reasonable simulation model was established according to the experimental results of Schottky Barrier Diodes (SBD), which were utilized to simulate the TMBS diodes. A theoretical calculation model of TMBS diodes under forward bias was constructed based on existing SBD theoretical analysis to systematically understand the simulation results. Both simulation and theoretical calculation demonstrate that the negative resistance characteristics of TMBS diodes under forward bias are caused by the accumulation of minority carriers at the epitaxial region and the accumulation of majority carriers at the epitaxial region and oxide layer.

Face super-resolution is a branch of the field of super-resolution. It is mainly aimed at the reconstruction of face images and distinguishing SR from general images. Face geometry prior information is used to optimize the face super-resolution network, which can generate high-resolution face images with better visual quality from low-resolution face images. In order to further improve the visual quality of reconstructed images, an improved face super-resolution network is proposed. In this paper, the key module of FSRNet is improved and new loss function is added to achieve a better face super-resolution network. Specifically, our job is to:(1) In the generated rough SR face image, we input low-resolution(16 x16) face image, then use the Deconv convolution enlarge images.(2)By introducing heatmap loss, facial attention loss and adversarial training to reduce the artifacts of FSRNet network. (3) We divide the network into two-step training, first train coarse SR network, get SR images quickly, and then we train the rest of the network. The final output is a super-resolution face with high visual quality.

In large scale distributed storage systems, erasure code is a basic technology that provides high reliability at low cost. Compared with traditional redundancy technology, erasure coding technology has low redundancy and high flexibility. Therefore, it is a good choice for distributed systems to construct Pyramid codes that are flexible and suitable for a variety of application scenarios, but its disadvantage is that the data recovery time is still long. To address the above problems, in this paper, we propose an Active Fault-Tolerant Pyramid (AFTP) based code, which dynamically adjusts the length of the group in the Pyramid code and the original data block correlation with redundant blocks by using the hard disk fault prediction model based on the decision tree to reduce the length of the group of data blocks in a potentially faulty hard disk, which can be used for multiple hard disk failures. All read and recovery operations are performed within the group, reducing recovery time without adding additional storage overhead. To verify the validity of the AFTP code, we conduct intensive experiments on the distributed storage system based on Ceph. The results show that, compared with Basic-Pyramid (BP), the recovery time of AFTP code is reduced by 8%-64%, and compared with the commonly used classic block codes, the recovery time of AFTP code is reduced by 11%-52%.

At present, the distributed storage systems are being adopted by more and more Internet companies because of their reliability and efficiency. In the distributed storage systems, some fault-tolerant technologies are often used to ensure the data reliability. Since the erasure code technology can store more data with lower storage space, it is often used as one of fault-tolerant technologies by some distributed storage systems. However, the existing erasure code technology has many problems. The large amount of repair overhead and I/O overhead consumed to recover data after data loss has become the biggest obstacle in actual use. This paper starts from the RS code, and constructs a new packet repair SLRC code by using the coding idea of stripe set, which can effectively reduce the repair cost required for single-node data recovery. At the same time, for the structure of SLRC code, a new data storage method, Curve storage, can effectively reduce the extra I/O overhead required for SLRC code degraded read operation. The experimental horizontal comparison with the LRC code and the RS code shows that the new method can effectively reduce the single node repair overhead and the additional I/O overhead required for the degraded read operation.

This paper provides an overview of our recent atomic-scale studies of semiconductor heterostructures, based primarily on combinations of zincblende compound materials grown by molecular beam epitaxy. Interfacial strain due to lattice mismatch inevitably causes growth defects to be introduced. Analysis of defect type and distribution using image filtering allows residual strain to be estimated. Exploratory investigations using aberration-corrected electron microscopy, which enables individual atomic columns to be resolved, are also described.