Search

The rapid development of information technologies like Internet of Things, Big Data, Artificial Intelligence, Blockchain, etc., has profoundly affected people’s consumption behaviors and changed the development model of the financial industry. The financial services on Internet and IoT with new technologies has provided convenience and efficiency for consumers, but new hidden fraud risks are generated also. Fraud, arbitrage, vicious collection, etc., have caused bad effects and huge losses to the development of finance on Internet and IoT. However, as the scale of financial data continues to increase dramatically, it is more and more difficult for existing rule-based expert systems and traditional machine learning model systems to detect financial frauds from large-scale historical data. In the meantime, as the degree of specialization of financial fraud continues to increase, fraudsters can evade fraud detection by frequently changing their fraud methods. In this article, an intelligent and distributed Big Data approach for Internet financial fraud detections is proposed to implement graph embedding algorithm Node2Vec to learn and represent the topological features in the financial network graph into low-dimensional dense vectors, so as to intelligently and efficiently classify and predict the data samples of the large-scale dataset with the deep neural network. The approach is distributedly performed on the clusters of Apache Spark GraphX and Hadoop to process the large dataset in parallel. The groups of experimental results demonstrate that the proposed approach can improve the efficiency of Internet financial fraud detections with better precision rate, recall rate, F1-Score and F2-Score.

State of charge (SOC) estimations are an important part of lithium-ion battery management systems. Aiming at existing SOC estimation algorithms based on neural networks, the voltage increment is proposed in this paper as a new input feature for estimation of the SOC of lithium-ion batteries. In this method, the port voltage, current and voltage increment are taken as inputs and the current SOC is used as output to train a neural network. Different from the adaptive filtering algorithm, which requires complex equivalent circuit parameter identification, this algorithm uses the voltage increment instead of the open circuit voltage (OCV); hence, the complexity of the SOC estimation algorithm is reduced, and the problem of inaccurate estimation caused by neural network algorithms without considering the internal structure of the battery is avoided. The experimental results show that compared with the traditional neural network algorithm, the neural network SOC estimation algorithm based on the voltage increment could improve the accuracy of SOC estimation.

In existing machine vision technology for fruit defects, the hue appears different, and the defect area is small due to the irregularity of illumination reflection from the surface incident light source, this makes it difficult to extract the defect area. Thus, we proposed an apple defect detection method based on the Fuzzy C-means Algorithm and the Nonlinear Programming Genetic Algorithm (FCM-NPGA) in combination with a multivariate image analysis. First, the image was denoised and enhanced through fractional differentiation. The noise points and edge points were removed, and the important texture information was preserved. Then, the FCM-NPGA algorithm was used to segment the suspicious defect graph. Finally, a method based on a multivariate image analysis strategy was used to detect the flaws of the apple’s suspicious defect map. The application results of 2000 images showed that the overall detection accuracy was 98%. Experiments show that the apple defect detection algorithm based on FCM and NPGA combined with multi-image analysis method is effective.

In view of the current sports training needs, the computer-aided teaching software Kinect is used as the image collector to collect moving images, and then the image is processed and feature extracted by the frame difference method and the Canny edge detection algorithm, and the horizontal line scanning algorithm is adopted. Complete the extraction of 25 different joint points of the human body. Experimental results show that the average correct recognition rate of the system for sports actions reaches 91.25%, and the correct evaluation rate reaches 95.9%, which can provide timely and effective feedback and evaluation, and play a certain role in sports training assistance.

Point cloud is an important 3D data structure, but its irregular format brings great challenges to deep learning. The advent of PointNet makes it possible to process irregular point cloud data by neural networks directly. As an extension of PointNet, PointNet++ can extract local features, which makes it perform better than PointNet in processing point cloud data. But in practice, it is common that the density of a point set usually varies with the location, which makes the computation overhead of PointNet++ very heavy. To deal with it, we propose an octree grouping-based network structure for PointNet++, named Octree-Grouping-PointNet++ (OG-PointNet++). It determines the point density by constructing an unbalanced octree for the point cloud, and groups point according to the point density. These point groups are assigned to different layers according to their density, and the local feature of each group is extracted by PointNet++. The global feature is obtained from the last abstract layer and used for classification and segmentation. The experiments show its competitive performance in many 3D tasks, such as object classification and semantic segmentation.

Multi-unit IGBT PIM (Power Integrated Module) is widely used in industrial transmission and household air conditioning due to its small size, light weight and low switching loss. More and more attention is paid to its reliability. Through a simulation study, the package structure of a PIM with a three-phase bridge rectifier, a chopper and a three-phase IGBT module was analyzed. Due to the limitation of package structure, chopper circuit works at the highest temperature. By comparing the temperature distribution of a SiC hybrid PIM using SiC Schottky barrier diodes instead of Si fast recovery diodes with the original Si-PIM, the chip layout and vertical packaging structure of the SiC hybrid PIM are optimized. The high thermal conductivity material such as graphene was applied to the PIM to improve the thermal reliability performance. After optimization, the maximum temperature of the module can drop by 24°C.

With the characteristics of high power density, high energy density and flexible configuration, electrochemical energy storage power station is widely used in power generation, transmission, distribution and power consumption, which effectively solves the problem of time and space distribution of electric energy. Based on the Ragone curve of energy storage battery, the design method of energy storage system is proposed in this paper. This paper introduces the drawing method of Ragone curve, and introduces the Ragone curve of commonly used energy storage lithium iron phosphate battery and lead-acid battery. Taking the given 20kW, 500kJ energy storage system design as an example, using the Ragone curve and the actual demand, combined with the battery power constraints, two kinds of energy storage battery configuration methods are obtained. This method avoids complicated derivation and calculation of electrochemical formula, and has certain practical application value.

In order to improve the accuracy of life prediction of lithium-ion battery, a life prediction method based on deep belief network and long short-term memory network is proposed. Combined with the capacity data of lithium-ion battery from the Prognostic Center of Excellence (PCoE) of NASA Ames Research Center, discuss the prediction accuracy of the network under different hidden layers, different hidden layer nodes and different optimizers, then choose the best for the fusion model. Combining the above two models, the experimental results show that: the fused model can effectively fit the decline trend of battery capacity, and the remaining useful life prediction error is only 1 cycle. Compared with using DBN or LSTM alone, the prediction accuracy of fusion model is better.

We proposed a temperature sensor by selectively infiltrating refractive index liquid (RIL) in the central airhole of a dual core photonic crystal fiber (PCF). The temperature dependent propagation characteristics of modes in the dual core PCF are analyzed numerically.

The three-phase four-wire three-level neutral-point-clamped shunt active power filter (NPC-SAPF) is considered as an effective way to compensate harmonics generated by nonlinear loads in power system. In this paper, a Lyapunov-Function-based control using the decoupled NPC-SAPF state-space mathematical model is presented. An affine nonlinear model of NPC-SAPF is established in the synchronous d-q-0 frame. Applying state feedback linearization method to the affine nonlinear model, a linearized and decoupled NPC-SAPF mathematical model is obtained. The globally asymptotically stable NPC-SAPF control system is built by choosing an appropriate Lyapunov candidate function and making sure that the function satisfies the required properties of Lyapunov stability theory. Compared with the classic proportional integral (PI) control method, the proposed control strategy has larger bandwidth and better compensation performance. The simulation results proved the validity of the proposed control strategy.

We report the investigation of a hybrid photonic crystal fiber (PCF) realized by filling liquid crystal to a twin-core PCF. Both the index-guiding and photonic bandgap guiding mechanism co-exists in the fiber, leading to interesting properties for sensing applications. This paper presents both simulation results of the hybrid PCF structure and shows preliminary experiment results of the temperature response of the hybrid PCF.

A simulation design was presented to optimise the model of long period grating based fibre devices. The simulation used a modified phase matching condition to better estimate the resonance wavelengths in the transmission profile. Experiments were conducted to verify the results obtained in the simulation. The experimental results matched closely with the simulations.

In this paper, for the first time Python was used to model the long period fibre grating devices and simulate the light propagating in the waveguide structures in order to design and optimize LPFG devices for sensing applications. It begin with resolving the effective refractive index of the LP01 core mode and different cladding modes, followed by calculating the mode electric and magnetic fields to determine the strength of coupling for core mode as well as between core mode and different cladding modes. After then, the transmission power spectrum within typical telecommunication wavelengths range was plotted. The spectrums and the wavelength shift under different surrounding refractive index was then plotted to analyze the SRI sensitivity of LPFG devices. Eventually, the results of Python and MATLAB will be compared for verification.

We propose a k-nearest neighbors (kNN) classification algorithm optimized by P systems in this article, called kNN-P, which can improve the performance of the original kNN classifier. A P system consisting of several cells is considered as its computational framework. Under the control of both evolution rules and communication rules, each cell determines the optimal set of k-nearest neighbors for a sample. The proposed kNN-P is evaluated on 18 benchmark datasets and compared with classical kNN algorithm and 8 recently developed improved algorithms. Comparison results demonstrate the availability and effectiveness of the proposed algorithm.

In this paper, we report an optical fiber Sagnac interferometer (OFSI)-based temperature sensor constructed by a selectively filled polarization-maintaining photonic crystal fiber (PM-PCF). The transmission spectrum of the OFSI is in sinusoidal form and is sensitive to temperature. Simulation results predicted a higher sensitivity by selective filling than nonselective filling. In experiments, we used an extremely low-cost process to realize the selective filling. A sensitivity of 2.58 $\hbox{nm}/^{\circ}\hbox{C}$ was achieved with an 11.7-cm-long PM-PCF. The sensitivity dependence on the infiltration length ratio was also investigated.

Battery Monitoring is very important for most of the electric vehicle(EV) and battery energy storage system (BESS) since the safety, operation and even the life of the passenger/operator depends on the battery system. Checking and controlling the status of battery within their specified safe operating conditions is exactly the major function of battery management system (BMS). The state of charge (SOC) is a critical parameter of a Li-ion battery, an accurate on-line estimation of the SOC is important for forecasting the EV driving range and BESS power dispatching. A widely used method to estimate SOC is based on coulomb counting, due to the uncertainty, including unit-to-unit variation, measurement noise, operational uncertainties, and model inaccuracy, it's difficult to estimate the SOC by using coulomb counting, an online inference of open-circuit voltage (OCV) is used to calibrate the SOC, however, a relatively flat OCV curve of LiFePO4 will lead to the error of SOC is large. In this paper, an algorithm using the derivation curve (dV/dQ vs Q) to calibrate the SOC on-line is bring out. The SOC estimation has been implemented using coulomb counting and derivation curve methods thereby eliminating the limitation of the stand-alone coulomb counting method. Finally, by utilizing the actual operation data, experiments and numerical table were conducted, show that this SOC calibrate algorithm based on characteristic curve has better robust performance of the practical application of LiFePO4 battery energy storage system.

We experimentally demonstrate a magnetic field sensor based on long-period fiber grating (LPG) and magnetic fluid. Our proposed sensor possesses a high sensitivity of ∼0.154 dB/Gauss and a low measurement threshold of ∼7.4 Gauss.

This paper proposes a kind of improved DV-Hop algorithm to solve the problem of low positioning accuracy in traditional DV-Hop algorithm. This method uses wavelet neural network to estimate the per-hop distance of the anchor nodes. In order to improve the positioning performance, the distance of per-hop is applied in the wireless sensor network. Simulation results show that, the localization error of the improved algorithm significantly reduced. And in the case of less anchor nodes, the superior positioning performance of the WNNDV-Hop algorithm is more obvious.

The long-term driving effect of the human capital on the national independent innovation was studied, taking the per capita domestic invention patents granted to the natives as a composite index of the national independent innovation strength, and taking the first principal component of four indexes — the proportion of R&D personnel in a country's population, the proportion of university and college student enrollment in a country's population, the enrollment rate of junior middle school graduates and the enrollment rate of senior high school graduates — as a composite index of the national human capital of a country. The Granger causality test result, based on the annual time series data of Japan during its innovation transformation period of 1965 to 2009, showed that the human capital was a long-term driving force for the national independent innovation, and its driving effect lasts for more than seven years. It was implied that a country's innovation transformation was the result of the long-term accumulation of its human capital.

We present an offline signature verification system using three different pseudo-dynamic features, two different classifier training approaches and two datasets. One of the most difficult problems of off-line signature verification is that the signature is just a static image while losing a lot of useful dynamic information. Three separate pseudo-dynamic features based on gray level: local binary pattern (LBP), gray level co-occurrence matrix (GLCM) and histogram of oriented gradients (HOG) are used. The classification is performed using writer-dependent Support Vector Machine (SVMs) classifiers and Global Real Adaboost method, where two different approaches to train the classifier. In the first mode, each SVM is trained with the feature vectors obtained from the reference signatures of the corresponding user and those random forgeries for each signer while the global Adaboost classifier is trained using genuine and random forgery signatures of signers that are excluded from the test set. The fusion of all features achieves the best result of 7.66% and 9.94% equal error rate in GPDS while 7.55% and 11.55% equal error rate in CSD respectively.

A review of photonic crystal fiber sensors with microbubble is presented. The work concentrate in three sensing fields: the first reports on temperature sensing; the second describes on strain sensing; the last part focuses on refractive index sensing and bio-sensing. The method of fabricating the microbubble is also briefly discussed.

We present hybrid propagation in a photonic liquid crystal fiber (PLCF) fabricated by selective infiltration of photonic crystal fiber (PCF) air channels with a nematic liquid crystal (LC). Temperature tuning of the position of the photonic bandgaps is also demonstrated.

Research on cross-media topic analysis and information retrieval methods, which utilize semantic of multimedia data to describe cross-media documents. As the emerging of multimedia data on Internet, text based methods cause the problem of inadequacy of semantic, so a uniform semantic presentation on different media data is very important for information retrieval system. Cross-media topic analysis and information retrieval methods are proposed in this paper. Firstly, generative method and visual topic learning algorithm are used to construct visual topic model and map visual data to text topics. This method can solve the problem of consistent semantic description of cross-media data. On this basis, cross-media topic tagging and integrated search are achieved. Using the proposed method, a food safety information retrieval system is established and experiment results also show its effectiveness.

A method for writer-independent off-line handwritten signature verification based on grey level feature extraction and Real Adaboost algorithm is proposed. Firstly, both global and local features are used simultaneously. The texture information such as co-occurrence matrix and local binary pattern are analyzed and used as features. Secondly, Support Vector Machines (SVMs) and the squared Mahalanobis distance classifier are introduced. Finally, Real Adaboost algorithm is applied. Experiments on the public signature database GPDS Corpus show that our proposed method has achieved the FRR 5.64% and the FAR 5.37% which are the best so far compared with other published results.

In this paper, the key features of Phonocardiogram (PCG) are extracted based on the slopes of envelop of Hilbert Transfer after relocating boundaries with energy envelope segmentation. In this attempt the overall accuracy of features extraction is found to be 91.95%. 25 significant clinical features are introduced, and chosen to make two-kind classification by SVM. In the results of two-kind classification, the overall accuracy is 91.3%, which is better than 85.23% accuracy in 100 features of Shannon Energy Envelope. The result shows that features including clinical signification is of signification for enhancing the accurate rate of Phonocardiogram classification.

We report on the fabrication of a miniaturized photonic crystal fiber (PCF) tip sensor for sensitive liquid refractive index measurement. The PCF tip sensor has a hybrid configuration of Fabry-Perot (FP) and modal interferometer. A short length (less than 1 mm) of PCF is spliced to a standard single mode fiber. An air bubble is created in the splice region forming the FP cavity. The air holes are fully collapsed on both sides of the PCF. The collapsed sections act as multimode fibers which excite and recombine higher order modes experiencing optical path difference (OPD) than that of the core mode and forming modal interference. The end of the PCF is modified to a solid sphere by fusion arc splicer. Experimental results show that the sensitivity of the probe can reach up to 21.4 nm/RIU (refractive index unit) to detect bulk liquid with refractive indices from 1.33 to 1.4.

Photonic crystal fibers (PCFs) offer innovative design and optical properties. In this paper, we measured the temperature dependence of a polarization-maintaining PCF based interferometer. The sensor head was fabricated by splicing PMPCF to single mode fiber (SMF) and temperature sensitivity of 15.4 pm/°C was obtained. The temperature measurements were carried out up to 1000°C.

We fabricate and demonstrate a refractometer based on photonic crystal fiber (PCF) modal interferometer. The streptavidin and biotin are used as analyte and receptor in the experiment. The fabrication of the sensor is carried out via micro-hole collapsing with straightforward fusion arc splicing technique. Both sides of a short length of PCF (∼3 mm) are spliced to conventional single mode fibers (SMF). The voids of the PCF are completely collapsed in the splice region, acting as mode splitter and combiner for the PCF modal interferometer. The bulk liquid sensitivity is about 70.2 nm/RIU (refractive index unit) based on the linear curve fit for refractive index ranging from 1.33 to 1.35. The transmission spectrum of the biotin modified PCF modal interferometer exhibits 0.8 nm wavelength shift after streptavidin and biotin binding occurs.

A method for writer-independent off-line handwritten signature verification based on grey level feature extraction and Real Adaboost algorithm is proposed. Firstly, both global and local features are used simultaneously. Secondly, dissimilarity vector is adopted. Finally, Real Adaboost algorithm is applied. Experiments on the public signature database GPDS Corpus show that our proposed method has achieved the FRR 5.64% and the FAR 5.37% which are the best so far compared with other published results.

In the process of the simulation design of the magnetoelectric(ME) tunable microwave resonator, we find that there are large errors between simulation results and experimental results when the effect of the demagnetizing field is ignored. Therefore, during the simulation process, a calculation model of the demagnetizing field is proposed based on the demagnetization effect theory of the ferromagnetic phase and a corresponding simulation analysis model is fabricated and characterized. When the magnetic field is parallel and perpendicular to yttrium iron garnet/ lead zirconium titanate (YIG/PZT) film surface, owing to the demagnetizing field effect, the simulation results of the ME microwave resonator agreed well with the experimental results. Based on this method, we can effectively predict the experimental results of the FMR frequency shift under the parallel and perpendicular magnetic field. Therefore, it is necessary to take into account the demagnetizing field effect during the simulation design of the ME microwave devices.

A theoretical model for the shift of ferromagnetic resonance (FMR) frequency by applying external magnetic field and electric field on magneto electric materials is proposed in this paper. This model is related to the geometry of YIG film and magneto electric(ME) constant of ferrite-piezoelectric layered materials. The insertion loss under different magnetic and electric field predicted by this model are quantitatively in good agreement with experimental results. According to this theoretical model, we found that the influence of the demagnetization field caused by the geometry of YIG film can not be ignored. Further more, we convert the electric field tuning on laminated ME materials into equivalent magnetic field tuning on ferrite layer, a numerical analysis model of microwave ME coupling filter including YIG/PZT (yttrium iron garnet/ lead zirconium titanate) layers is fabricated. For parallel magnetic field, simulation results are agreed well with experimental results. Then, according to this numerical analysis model, we predict the phenomenon of approximately linear magnetic tuning FMR frequency shift under perpendicular magnetic field.

We present a simple fabrication method for embedding gold nanoparticles (GNPs) in a microfiber with two main advantages. The GNPs are positioned within the microfiber securing maximum enhancement of the electrical field and protection of the GNPs from the surroundings; moreover incoupling losses to the microfiber are reduced significantly by integrating a single mode fiber in the device. The average enhancement factor of the electric field within one unit cell of the microfiber due to the GNPs is estimated to 17.4. This enhancement leads to an enhanced Kerr effect near the plasmon resonance frequency. The fabrication method is demonstrated without GNPs.

this paper proposes an efficient method for realizing the fault tolerance of CNC software with the introduction of artificial neural network (ANN) to the design filed of CNC software. In addition, function aspects (velocity, acceleration, chord error, real time, prediction accuracy) from the experiment on Non-Uniform Rational B-Spline (NURBS) interpolator based on ANN were evaluated in detail. Our experimental results showed that the NURBS interpolation based on ANN not only meet the requirements of function aspects, but also can realize fault tolerance technology, which may provide a new strategy in the improvement of the reliability of CNC software.

Based on the features of NC horizontal milling process, the 3D geometric model of virtual machine tool was established by Microsoft Visual C++6.0, MFC and OpenGL graphics library. The method of Construction of Solid Geometry (CSG) and Boundary Representation (B-Rep) were adopted, and some realistic effects treatments (including material, light, texture mapping, etc.) were applied. The established virtual machine tool can be zoomed, rotated and translated in the view area. The workpiece decomposition expression model was put forward, and Boolean subtraction was implemented in this procedure. At the same time, a tool solid model to simulate the real time process of cutting was established. Simulation example proves that this system runs reliably and avoids the inconvenience and material waste of other methods.

With the object-oriented modeling method, the information models based on featured manufacturing resources of the automobile gears were established. The models consist of the basic information of manufacturing equipments and the information of feature-based process capabilities. At the same time, the associated relationship of features among features, machining tools and technical equipments was established. The relationship model increases the searching efficiency of manufacturing equipments.