This paper aims to explore the law of vessel traffic flow on the “21st Century Maritime Silk Road”. Based on the Automatic Identification System data of the “21st Century Maritime Silk Road” in 2018, the time series model is used to analyze the vessel traffic flow of three situations of cargo ships, tankers and cargo ships-tankers. The research shows that: 1) The changes in vessel traffic flow can be fitted and predicted by using the differential autoregressive moving average model (ARIMA). 2) Three situations have the same optimal selection model ARIMA (1, 1, 2). The results show that the current vessel traffic flow of “21st Century Maritime Silk Road” can be predicted by the traffic flow in the last two time periods. Three situations all have relatively accurate prediction results, and the prediction effect of the ARIMA (1, 1, 2) model on the single ship traffic flow is better than that of the mixed ship traffic flow.

To study the fault probability distribution of target event (TEFPD) of fault modes in space fault network (SFN), the distribution characteristics are determined under different conditions, and the determination methods of TEFPD under different conditions are proposed. The research objects are unit fault evolution process and all event induction + target event process. According to the relationships between cause events and result events, the analysis methods are divided into comparative form method and inheritance form method. According to the existence of events in fault modes, the processing methods of fault probability distribution can be divided into maximum method and average method. Considering the influence of many factors, the fault probability distribution of event is introduced into the analysis, and TEFPD is obtained under various conditions. TEFPD is obtained through a simple fault mode. Finally, the significance of TEFPD characteristics obtained by various methods is summarized.

The purpose of this paper is to illustrate a new scheme for a new medical high-precision constant current electric stimulator system. Methods: According to the design requirements of stimulation voltage of high-voltage and stimulation current of high-precision, the power supply adopts push-pull boost and then double-voltage rectification to obtain a stable 360 V high voltage. The 12 bit D/A and op amp constant current source circuit in the circuit could ensure that a high-quality pulse width is less than 10 μs and a maximum constant current is 100 mA. As for the high human safety factor, the control signals, the control signals, the feedback signals and the communication signals in the circuit are isolated by digital isolation, and the power supply is shielded by DC-DC power supply. At the same time, the over-current protection circuit, the pulse width limiting circuit and the pulse width error detection circuit are designed. Results: According to the high-amplitude voltage, high-precision current and high-safety factor of the human body, a system can be designed, which can be used as an electric stimulator for evoked potential or as a functional neuromuscular electrical stimulator. Conclusion: The feasibility and reliability of the system are proved by the hardware circuit simulation diagram and the actual test results. Concurrently, the intensity, frequency and pulse width of the stimulation current reach higher precision, the electric stimulation evoked potential waveform is clear, and the noise is small. Significance: The scheme of this new type of medical high-precision constant current stimulator system has been proved to meet the requirements of clinical medical examination.

High-resolution magnetic resonance images are of great significance for medical diagnosis. A convolutional neural network with multi-resolution learning is proposed for magnetic resonance image (MR) super-resolution. The network is an improved deep residual network, which involves residual units for feature extraction, a deconvolution layer for multi-resolution up-sampling, and a multi-resolution learning layer. The proposed network performs the super-resolution task in the low-resolution space, which can accelerate the network. Multi-resolution upsampling is put forward to integrate multiple residual unit information and to accelerate the network. Multi-resolution learning can adaptively determine the contributions of these upsampled high-dimensional feature maps to high-resolution MR image reconstruction. Experiment results indicate that the proposed method can achieve a good super-resolution reconstruction performance for magnetic resonance images, which is superior to the state-of-the-art deep learning methods.

Under the environment of e-commerce, fully understanding the consumer’s interest change process and predicting their purchase intention are the problems that personalized recommendation systems need to solve. To solve this problem, clickstream data with huge potential information has been widely studied and applied due to its availability and accuracy of prediction. Therefore, in order to mine the consumer’s interest change process reflected by the user’s click behavior and predict the consumer’s purchase intention, click stream data is used, based on interest drift theory, to establish an effective model for measurement and prediction on multi-attribute decision-making method (MADM) of preference ranking organization methods for enrichment evaluations (PROMETHEE). As a result, a multi-session consumer purchase intention evaluation system consisting of three primary indicators and eight secondary indicators is established, providing a practical evaluation method for consumer purchase intention prediction.

In order to solve the deficiency in capturing key information in key areas and the high cost of image annotation in the existing trademark retrieval methods, this paper proposes an unsupervised trademark retrieval method based on attention mechanism. The method applies the attention module to both the spatial dimension and the channel dimension of the feature map layer in the neural network of instance discrimination. Through assigning weights to each channel and learning the spatial transformation parameters, the unsupervised network improve its ability of extracting feature. We further validate the effectiveness of our method on the public trademark datasets and the experiments demonstrate that the proposed method in the paper is better than the traditional trademark retrieval methods, and even surpasses some supervised trademark retrieval methods.

With the wide application of space-to-earth observation technology in surface deformation monitoring, there is an increasing demand for the management and display of deformation monitoring results. On the one hand, effective management is needed to deal with the rapidly increasing data amounts from growing numbers of measuring frequencies and monitoring technologies (such as GPS and InSAR). On the other hand, the demonstration effect of the results obtained from the monitoring needs to be more intuitive and humane. In order to solve these two problems, based on the technology,this paper develops an online display system of deformation monitoring results throughOpenLayersmapframeworkand ASP.NET technology.In this system, InSAR data can be visualized by presenting point cloud data in a scale and GPS velocity data can be visualized by generating directed line segments in a scale.By using database to standardize the storage of monitoring data, this system has realized the efficient scientific management of the deformation monitoring results and greatly improved the efficiency of the displayeffect, editing ability and management efficiencyof the deformation monitoring results.

The stability of solutions of a class of impulsive stochastic differential equations was studied by using the fixed point theory and Lyapunov stability theorem and sufficient conditions for the exponential stability in the mean square of the equation are obtained.

In order to break through the performance limitations of hand-craft feature point detectors, a new data-driven unsupervised feature point detection network based on differential feature response graphs is proposed. The network calculates the differential output using convolution kernels of different scales, uses the absolute value of the differential response map to select a large number of feature points and judges the importance of these feature points. The detector is trained by using various image transformations such as rotation, illumination, and blur to obtain corresponding feature invariance, making the network more suitable for small-scale data set training. Through experiments on three commonly used data sets, and qualitative and quantitative comparison analysis with existing classic algorithms, the experimental results show that the proposed unsupervised feature point detection network based on the differential response graph can complete the feature point detection well. The network locates object edges more accurately, requires less training time, and has less dependency on dataset size. The network has optimized and improved feature point detection performance.

R245fa is one of the most commonly working fluids used in organic Rankine cycle (ORC) system. Investigating its heat transfer flow characteristics is of great importance in guiding the design and operation of ORC system. An experimental test rig is established for the test of the in-tube heat transfer and flowing of the organic working fluids. The experimental research for the boiling heat transfer of R245fa in a horizontally smooth tube is conducted. Firstly, the influence of mean quality and mass velocity on the average flow boiling heat transfer coefficient and wall temperature distribution along the tube are obtained and analyzed; then, experimental results are compared with three classical heat transfer correlations prediction results, with an average deviation of 38%, 39%, and 20%. The high temperature evaporation experimental results provide a basic database for the correction of heat transfer correlation and guidance for the engineering design practices.

A battery management system (BMS) is designed to manage 15 lithium battery cells by using STM32 microcomputer. Firstly, a battery monitoring chip BQ76940 is used to design measurement unit, a switched-capacitor-based circuit is used to design active equalization unit. Then, the SOC of battery is estimated by the current integration method and open-circuit voltage method, while the internal resistance augmentation failure of battery is detected on-line by battery equivalent model and parameter identification method. The design of both hardware and software of the BMS is introduced in detail. And experimental results for measurement unit, equalization unit and internal resistance identification are provided. The results show that the battery management system can collect the voltage, current and temperature of the battery, balance the single battery with different voltage, and effectively identify the internal resistance of the battery, and it has the advantages of high measurement accuracy, well equalization performance and online detection of the internal resistance for batteries.

Most natural soils have the characteristics of transverse isotropy. Using the method of block processing in the same model to discuss the change of passive earth pressure of transverse isotropic soils, based on the principle of least squares and the theory of fabric tensor, the two methods of one-step and multi-step displacement applications are studied to calculate the angle change between the direction of the major principal stress of the soil and the normal direction of the sedimentary surface. New soil strength parameters c_i and φ_i in each block are obtained, and then the change of passive earth pressure under soil anisotropy is analyzed. When reaching the passive limit equilibrium state in the transverse isotropy soil, the research shows that the passive earth pressure obtained by applying multi-step displacement method, which is close to the theoretical solution of Coulomb's passive soil pressure, is smaller than that of applying one-step displacement method. The change of earth pressure is more significant when the sedimentary surface being oblique than that of the sedimentary surface being horizontal. The differences in passive earth pressures obtained under the four type yield criteria are consistent with theoretical analyses.

Zhong's house is a well-preserved example of the enclosed house of the form of “nine halls and eighteen yards” in Qing Dynasty in Qingyuan area. It has typical Hakka enclosed house features in terms of architectural form and construction techniques, which is of great significance to the study of the humanistic historic culture in Qingyuan area. Based on the analysis of the structural damage caused by social and historical factors and the establishment of the digital model of the building by means of surveying and mapping tools and photography technology, this paper analyzes the layout features and defense features and records the detailed characteristics of the building of the enclosed house called Zhong’s house in Qingyuan through on-the-spot investigation. To provide appropriate strategies for the protection and sustainable development of the building for activate the building and continue its vitality.