Search

In this article, the fault diagnosis problem for power converters is considered. Given that the existing fault diagnosis models rarely address the problems of the data noise and the new faults that are never emerged in the database, thus, an optimized fault diagnosis model for power converters based on temporal convolutional network (TCN) is proposed. Our contributions include the following: 1) unknown faults can be efficiently distinguished with an optimized classifier; 2) the proposed model has good robustness and reliability under noisy environment without any subsidiary predenoising algorithm; and 3) it can realize adaptive feature extraction, and the parameters are small. Experimental results on a three-phase voltage inverter platform demonstrate that the proposed approach is efficient and can be adaptively applied to various real applications.

Objective: The aim of this study was to investigate the effect the motion-correction reconstruction algorithm SnapShot Freeze (SSF) on the diagnostic performance of coronary computed tomography angiography (CCTA) using high-resolution, wide-detector CT. Methods: CCTA was performed using an electrocardiography-triggered auto-gating protocol in 62 patients. All images were reconstructed using both standard algorithm and motioncorrection algorithm (SSF). According to the American Heart Association (AHA) guidelines for coronary angiography, the coronary artery was divided into 15 segments. The coronary artery segments were evaluated using a 4-point Likert scale to determine the image quality. Arteries with a score greater than 1 were further evaluated by two expert radiologists blinded to conditions. The diagnostic performance of SSF was compared in the patients with high heart rate (HR ≥ 75 beats per minute) compared to low HR (< 75 beats per minute). Results: At the segmental level, the Likert score for segments 1–4 and 6–14 were higher in SSF compared to standard algorithm (p < 0.05). The proportion of images that was improved by SSF was greater for the high HR systolic phase imaging (66.7%) than for the low HR diastolic phase imaging (34.3%, p < 0.05). Conclusion: The SSF algorithm improves coronary CTA images acquired using a 256-row wide detector CT, compared to standard algorithm, particularly in patients with high HR.

When defending the intruding links of optical fiber network, the intruding signals are not classified, which leads to the low accuracy of feature extraction and the poor effect of intruding defense. For this reason, the inter-correlation characteristic is used to process the fiber signal. The time-domain eigenvectors, conjugate signals and feature sets of the intrusion signals in the optical fiber network are extracted from three aspects: wavelet domain, frequency domain and time domain by using the two-threshold method and the short-time energy. The classification model is designed by the neural network algorithm, the feature vectors of the intrusion signal are input into the classification model, and the error function is determined by the negative logarithmic likelihood function. The network transmission signal vulnerability is obtained by comparing the data metadata characteristics in the process of network data transmission. Big data is adopted to quantitatively analyze the data interception and monitoring risks in the transmission process, establishing the network transmission security mechanism, completing the quantitative assessment of data transmission risks, and realizing the security defense of network intrusion link situational awareness. Experimental results show that the proposed method has high accuracy in feature extraction and intrusion detection.

Violence detection from surveillance video is a challenging and attractive task. This paper introduce a new violence detection using binocular stereo vision. We use the sparse stereo matching method to extract the feature points of both rectified images and obtain the vision disparity of the point. The 3D coordinates of the points are calculated through the standard 3D measurement theory. To describe the spatio-temporal property, we extract features aligned with the trajectories to characterize depth information (Three-dimensional motion vector), appearance (histograms of oriented gradients) and motion (histograms of optical flow). In order to obtain the discriminative feature, this paper adopts sparse coding scheme and support vector machine (SVM) to classify the feature vector as normal or abnormal.

Sidelobe reduction is important to improve the quality of through-the-wall radar imagery, and adaptive sidelobe reduction (ASR) is the widely used method for that. However, similar to coherent factor (CF)-based methods, the major defect of ASR is to suppress the weak targets in the imaging scene. Moreover, in near-field radar applications, including through-the-wall imaging, ASR cannot achieve the global optimum. In most letters about ASR, the three crucial parameters, the filter order, the interpolation factor, and the constraint threshold, are selected empirically. This letter analyzes their relationships and value ranges and then proposes a modified ASR (M-ASR) method, which has a more reasonable strategy of parameter assignment. Both simulative and experimental results validate the feasibility of M-ASR in the near-field application and its improvement in weak target detection.

Human action recognition has many applications including design of human–machine system. Identifying the interaction between the driver and the vehicle information system is necessary to accurately identify the driver’s intention and improve the stability of the vehicle. A machine learning-based framework for interaction action recognition of drivers was proposed in this chapter. Several dimensionality-reduction algorithms (PCA, Isomap, LLE, LE) for interaction action recognition are compared in this chapter. The test sequence is mapped into a low-dimensional space through these dimensionality-reduction algorithms, and traditional classifiers (naive Gaussian, logistic regression, SVM, Kneighbors, DecisionTree) were trained in order to test the effect of dimensionality-reduction. Results show that “LLE+SVM” achieves the highest precision rate.

With the rapid development of intelligent transportation, the public transport intelligence dispatching system is playing an increasing important role. However, the traditional statistical analysis method can not meet the requirements of current bus passenger flow analysis and prediction. The realization of bus system optimization based on large data plays an important role in solving the shortage of traditional statistical methods, and it is the key to realize the short period prediction of bus traffic. Based on the linear regression forecasting method, this paper uses MATLAB-R2016a as the system development environment, and compares the three forecasting methods of linear regression forecasting model, grey prediction model and time series prediction model, and gives the test to realize the bus system optimization technology. Through the forecast of the passenger flow in the short period of a bus line in Nantong, the optimization of the bus line is perfected and the traffic congestion is alleviated by the dispatch center.

Deep learning has strong abilities in finding and expressing characteristics of pictures. Recent years, with the arrival of big data era and the development of computers, deep learning has made great breakthroughs and become the focus of the field of computer vision. First the history and classification of deep learning are presented. This thesis also introduces the basic theory of typical deep learning models on computer vision, which include convolutional neural network, recurrent neural network and generative adversarial network. And then summarizing the research situations and progress of deep learning on image classification, image detection, image segmentation as well as video recognition and prediction. Finally, the development and trend of deep learning in the field of computer vision are analyzed. The combination of convolutional neural network and recurrent neural network will be a good choice for video recognition and prediction, which still has a big gap between human beings cognition. And it is the generative adversarial network which has strong ability to generate new samples based on the potential distribution will play an important role in computer vision.

Aiming at air defense forces operation supported by space information, the paper designs the simulation of air defense forces operation supported by reconnaissance information, early warning information, communication information, navigation and orientation information, and weather information. The paper lays the foundation of simulation research on air defense forces operation supported by space information.

A novel digital predistortion (DPD) technique is proposed to linearize ultra-broadband millimeter wave (mmWave) power amplifiers (PAs) by only employing very limited bandwidth resources for the Tx, feedback loop (FB), and baseband (BB). Compared to the conventional methods, the proposed method will comprehensively reduce the bandwidth requirements for the whole system, which will make the linearization affordable for mmWave PAs. To validate the proposed idea, a 4-carrier 320 MHz modulated signal was employed to excite a mmWave PA with the center frequency of 41 GHz. Experimental results have proven that the proposed method can effectively realize the PA linearization with very narrow Tx/FB/BB bandwidth, which largely extends the capability of DPD to the forthcoming 5G era.

In this paper, a millimeter wave digital pre-distortion (DPD) platform for linearization of power amplifiers (PAs) is presented for the first time. The platform covers a frequency range from 40.5GHz to 43.5GHz and supports wideband applications up to 800MHz. Noise floor of the proposed platform is about −74dBm, which is comparable with the DPD platform built up by instruments. Experimental validation is carried out with a balanced PA driven by an orthogonal frequency division multiplexing (OFDM) signal with bandwidth of 160MHz and peak-to-average power ratio (PAPR) of 7.8dB. Measured results show that after DPD adjacent channel leakage ratio (ACLR) can reach −45dBc with an improvement about 12dB.

Violent detection from video is a hot topic which has wide application. The aim of this paper is to design a novel feature descriptors called motion gradient location and orientation histogram (MoGLOH), which encode not only the local appearance but also explicitly models local motion. Our proposed MoGLOH is composed of two part of information. The first part is the gradient location and orientation histogram (GLOH) describing the spatial appearance, and the second part is an aggregated histogram of optical flow with a log-polar location grid named Optical Flow Orientation Histogram (OFOH) which indicate the movement of feature point. To eliminate the feature noise, the non-parametric Kernel Density Estimation (KDE) is employed on the MoGLOH descriptor. The theoretical analysis demonstrates the proposed algorithm performs robustly and favorably.

In this paper, a camera ranging method for agricultural robot localization in arable farming is described. Three calibration objects in different directions in the field are set. The distances between the robot are calculated, and the calibration objects are used the camera to obtain the robot position. The benchmarking procedure in Matlab includes image binarization of the calibration object and the curve fitting method was used to establish the relationship between the pixels number of the calibration object and the actual distance. The verification results show that the relative error of ranging is less than 1.5%. Furthermore, the camera ranging method for agricultural robot localization is proved to be easy and feasible.

Due to the uncertainty of the factors that influence the network service time and other characters of college student, Bayesian Network is used to model this kind of system. Different algorithms are used for learning Bayesian Networks in order to compare several models. It is suggested that researchers can use Bayesian Networks to explore the potential relationship between variables of complex social problems. The result indicates that learning target and family closeness degree are the key variables which influenced college students network service time. Origin of student and family economy didnt influence college students network service time directly. Schools and community should strengthen the education of college students life planning and communication with parents.

A design of mechanical arm for laser weeding robot is described. The long term objective of the project is the replacement of chemical weeding by laser weeding. The distinguishing feature of this design is the use of a mechanical arm, which is equipped with a laser generator and a control center consisted of a single chip microcomputer and two driver hybrid ICs. The mechanical arm comprises two stepping motors vertically linked together. The robot moves ahead one step, the working area of laser generator is divided into 10×10 grids. Through the inverse operation of mechanical arm, the motor displacement angle of each grid is calculated. Mechanical arm turns to designated position if single chip gets weeds position information. The result of the design is a key executive mechanism of laser weeding. The operation error of mechanical arm is analyzed. The arm can point the laser beam towards the target weeds with small error. After the order of the execution areas is optimized, the length of arm moving path is shorter. Analysis of the running track shows that the sequence of execution area is feasible and reasonable.

In order to improve the personalized mobile network services, some researchers have employed the social relationship into the acquisition of mobile user's needs. But, when mobile users belong to different communities, their impacts on other mobile users are different. Therefore, in this paper, we propose an improved division of mobile social network based on the mobile user's behaviors. Firstly, we propose a computation of trust including the direct trust and indirect trust based on communications between mobile users. Then, we construct the mobile social network according to the obtained trusts. Afterward, we propose an improved method to divide the mobile social network based on cohesive subgroups. Finally, we perform experiments using the MIT real dataset. Experimental results show that we can get more accurate subgroup division.

This paper attempts to present the structures and characteristics of variable frequency low-speed high-torque squirrel cage induction machine for elevator (VFLHSCIM). The structures of rotor sheet and stator winding are analyzed. And how the rotor resistance influences the performance of the motor is presented. Moreover, a prototype is tested, and the experimental results are analyzed.

As friction is the main nonlinear resistance in the flight simulator servo system which can lead to overall flight simulator performance degradation. The mathematical model of flight simulator servo system was created based on a three-axis servo system and the LuGre friction model was analyzed. Exponential approach law was introduced into variable structure control and the robust controller was realized. The simulation result shows the sliding mode controller has superior position tracking and speed tracking performance for low speed servo system with friction compensated effectively.

Thermocouple sensor for temperature measurement has been widely used, however, the increase of precision is constrained due to the shortcoming of hardware based or table look up method, especially with nonlinear adjustment and cold end compensation. A new method was presented to compensate nonlinearity and cold-side-offset for signal processing of thermocouple with RBF neural networks. The structure of RBF neural networks was proposed and optimized with genetic algorithm, the principle of temperature measurement with thermocouple was analyzed and the neural networks model for signal conditioning was created. The simulation experiments show that the algorithm can improve network generation ability and high accurate compensation and nonlinear adjustment for cold-side-offset was realized effectively.

Inverter and related system were widely used in power electrical system and motor drive system to enhance the reliability and efficiency, the faults with various types and difficult to isolate with traditional techniques, so a new method based on neural network was presented. The neural-point clamped three level invert systems were analyzed and fault features were created by harmonica spectral analysis. The neural network was designed with algorithm programmed, with the fault diagnosis as inputs of neural network, by neural networks adaptive self-learning and take the outputs as judgment of fault types and then the faults occurred in inverter system was isolated . The new technique proposed in this paper was a theoretical foundation for invert system fault diagnosis and practical application for motor driver system, the simulation shows this method is effective and can be widely into inverted fault diagnosis system and relevant fault diagnosis system.

A nonlinear sliding mode control method is presented for single inverted pendulum position tracking control based on RBF neural networks. The neuro sliding mode controller was designed with RBF neural networks and the stability of the proposed control scheme is proved by Lyapnouv theorem, also the control scheme is applied to the nonlinear inverted pendulum system, simulation studies shows the methods is effective and can applied into nonlinear control system.

As mobile welding robot are widely applied into industry and the weld quality are highlighted, the more flexible control strategy for welding robot are needed, but these control systems are often hard to control by traditional method, so a nonlinear sliding mode control method based on variable structure control was presented for mobile robot control was designed. Sliding mode control (SMC) is a special nonlinear control method which have quick response, insensitive to parameters variation and disturbance, online identification for plants are not needed, its very suitable for nonlinear system control, but in reality usage, the chattering reduction and elimination is key problem in SMC. In this paper, the kinematics models of mobile welding robot was established, the switching function was designed and the stability was analyzed by Lyapnouv theorem, also the sliding mode controller was designed with uniform reaching law, by simulation we can get the methods that presented in this paper is effective and can applied into linear or nonlinear control system.

Sliding mode control is a typical nonlinear control strategy which was easy realized and with strong robustness, in this paper, a neuro sliding mode controller was designed with RBF neural networks and the stability of the proposed control scheme is proved by Lyapnouv theorem. For the chattering of sliding mode control are often derive from switching gain, the gain was adjusted with neural networks with RBF networks’ output, the algorithm with fixed gain and adaptive gain are all proposed, also the control scheme is applied to a nonlinear system, simulation studies shows the methods is effective and can applied into nonlinear control system.

In this paper, the bio-electrochemistry mechanism of the microbes in the dairy is analyzed, we construct the bio-sensor of the conduction micro-current with the oxidation- reduction reaction between the strong oxidation pole and the extra coenzyme, devise the quick-quantitative testing instrument. Compared with the traditional method of plate cultivating germs, this method is more quickly, conveniently, timely.

Short-term load forecasting in power system is necessary for management and contrl of power system. A new method for short-term load forecasting was presented based on neural networks optimized by genetic algorithm (GA) is proposed in this paper, short-term load forecasting model for power system was setup as sample sets for Elman neural network(Elman NN), with GA's optimizing and Elman NN's dynamic feature, the higher forecasting pricision was realized and the simulation indicates the method is feasible and effective.

The artificial neural networks with structure self-organizing algorithm was used in quality testing of gear wheel, the structure self-organizing training algorithm was proposed, the application in quality testing of gear wheel was analyzed and the simulation shows the method is effective and can applied into mechanism quality testing system with the automation realized.

The controlled plant can be tackled with intelligent control always a complex system and often can not be easily controlled with traditional control strategy, single intelligent control maybe inadequate for some complex system. So the integrated intelligent control is needed and often can combine with traditional control, with take advantages respectively. And intelligent control is scramble for it. In this paper, an integrated intelligent controller proposed with hierarchical structure, the principles of hierarchical was analyzed, the expert controller, fuzzy controller and neural network PID controller are designed, the integrated controller cooperate with each other by expert ratio divider, and applied to a control plant with the good control effect was proved.

It is widely believed that Shor’s factoring algorithm provides a driving force to boost the quantum computing research. However, a serious obstacle to its binary implementation is the large number of quantum gates. Non-binary quantum computing is an efficient way to reduce the required number of elemental gates. Here, we propose optimization schemes for Shor’s algorithm implementation and take a ternary version for factorizing 21 as an example. The optimized factorization is achieved by a two-qutrit quantum circuit, which consists of only two single qutrit gates and one ternary controlled-NOT gate. This two-qutrit quantum circuit is then encoded into the nine lower vibrational states of an ion trapped in a weakly anharmonic potential. Optimal control theory (OCT) is employed to derive the manipulation electric field for transferring the encoded states. The ternary Shor’s algorithm can be implemented in one single step. Numerical simulation results show that the accuracy of the state transformations is about 0.9919.

For packet-based transmission of data over a network, or temporary sensor failure, etc., data samples may be missing in the measured signals. This paper deals with the problem of iterative learning controller design for linear discrete time systems with missing measurements. The missing measurements will happen at any sample time, and the probability of the occurrence of missing data is assumed to be known. The iterative learning controller is developed to guarantee the convergence of the tracking error and with quadratic performance index. A numerical example is provided to demonstrate the validity of the proposed design approach.