The presented investigation on fault diagnosis of ball bearings compares the conventional method using FFT spectra with a data-driven approach using Support Vector Machines (SVMs). Three different cases of bearings (one healthy and two faulty bearings with different crack thickness) were used as experimental cases. The experimentally obtained time-domain acceleration signals were converted to the frequency-domain and de-noised using optimal wavelets selected based on relative magnitudes of Shannon entropy and energy values. The dominant peak was identified for each case and was subsequently compared with the characteristic bearing frequencies evaluated theoretically. The wavelet transformed time-domain experimental data was also used to train the SVM classifier. Also, the effect of statistical tools such as Principal Component Analysis (PCA) and Zero-phase Component Analysis (ZCA) on the classification accuracy of normal SVM and wavelet feature extraction-based SVM have been investigated.

Rolling bearings are the main components of modern machinery, and harsh operating environments often make them prone to failure. Therefore, detecting the incipient fault as soon as possible is useful for bearing prognostics and health management. However, the useful feature information relevant to the bearing fault contained in the vibration signals is weak under the influence of the noise and transmission path. The useful feature information is even submerged in the noise. Thus, it becomes difficult to identify the fault symptom of rolling bearings in time from the vibration signals. Stochastic resonance (SR) is a reliable method to detect the weak signal in intense noise. However, the effect of SR depends on the adjustment of two parameters. Cuckoo Search (CS) is a heuristic novel optimization algorithm that can search the global solution quickly and efficiently. Thus, CS is utilized to optimize the two parameters in this paper. Local signal-to-noise ratio (LSNR) is used to evaluate resonance effect. Two bearing fault datasets were used to confirm the effectiveness of SR optimized by CS. SR methods optimized by particle swarm optimization (PSO), genetic algorithms (GA), firefly algorithm (FA), and ant colony optimization (ACO) are also used to detect the bearing fault signal in the two datasets. The analysis results state SR optimized by CS can find better LSNR than SR optimized by other algorithms no matter if it is in the same iterations or in the same computation time, thereby making the fault feature more obvious.

In medium and high capacity transformers where current rating is high and the number of turns is low, the low voltage (LV) winding is generally of the helical type. These helical windings have very large magnitudes of electromagnetic forces during a short circuit. This is due to the inherent asymmetry of helical structure. The objective of this work is to use the finite element method to compute the radial and axial components of short circuit forces and identify areas of high stresses in the windings. This can be used to find the likely reason of transformer failure during a short circuit. For this work, a 3-phase power distribution transformer of 11kV/433V, 630kVA rating is considered. The effect on short circuit forces of the tapping in the center of HV winding is also studied.

The feature-based visualization method can separate important areas for users from flow field data, which can better highlight the feature structure. However, most of the current feature extraction methods are only applicable to single typical features, and they need complex mathematical analysis. Based on the above reasons, this paper proposes a universal feature visualization method, recognizes demand in the region of flow data, shows the characteristics of structure protruding from the global visual effect in the design of a multi-dimension parallel convolution kernel that contains the recognition model, and further puts forward the method of feature visualization based on a convolutional neural network. Compared with the classical three level BP neural network model, our model gets a high accuracy rate. We verify the effectiveness of the method and solve the problem of insufficient expansion of existing methods.

In the era of rapid development of multi-media information, forgery detection has become an important research field of digital image security. This paper proposes a new method to detect the forged image generated by deep learning. First, the feature matrix is constructed through extracting each pixel value of channels a and b in Lab color space for the real and the forged image training set, respectively, which is used to fit the Gaussian Mixture Model (GMM) distribution. Then, the Expectation Maximization (EM) adaptation algorithm is used to re-fit the GMM for test image using the obtained GMM parameter as prior information. Finally, the similarity between two GMM is calculated for forgery detection. Experiments show that the proposed method is more accurate than the traditional SVM for forgery detection.

With strong heterogeneity and the limited computing ability of IoT nodes, this dissertation proposes a Trust Authorization Model based on detection feedback in IoT that is combined with the current trust model of IoT as well as implement storage and other tasks. By calculating and storing the cluster head node along with its strong ability to facilitate the data transmission and search for energy consumption, it prevents the local network from being limited by the computing power of the device. In terms of trust calculation, the threshold value is based on the recommendation. At the same time, the BP neural network algorithm with self-learning function is periodically detecting the interactive data stream, detecting the attack nodes, quickly implementing the response measures, and meeting the actual situation of unmanned IoT of mass devices. Simulation results show that this model has lower energy consumption than other similar models, has good coping ability for attacks such as malicious recommendation and malicious slander, and has a higher detection rate and response rate to attack nodes.

Nowadays, remote sensing images have been widely used in civil and military fields. But, because of the limitations of the current imaging sensors and complex atmospheric conditions, the resolution of remote sensing images is often low. In this paper, a superresolution reconstruction algorithm based on the deep convolution neural network to improve the resolution of the remote sensing image is proposed. First, this algorithm learned a series of features of the mapping between high and low resolution images in the training phase. This mapping is expressed as a kind of deep convolutional neural network; the trained network is a series of parameter optimization for super-resolution reconstruction of remote sensing image. Experimental results show that the superresolution algorithm proposed in this paper can keep the details subjectively and improve the evaluation index objectively.

Software rejuvenation modeling of the virtualized Cloud Server has been studied. A software rejuvenation policy on the virtual machines and the virtual machine monitor has been proposed in order to ensure high availability of the virtualized Cloud Server. The multi-component system, composed of the virtual machines and the virtual machine monitor, which are structurally dependent, has been reduced to the multiple two-component systems. The state-control-limit-based rejuvenation policy has been proposed and the stationary probability density of the two component system state has been derived. Furthermore, the stationary unavailability of the virtualized Cloud Server has been modeled. Numerical experiments have verified the correctness of the probability density function and the feasibility of the rejuvenation policy. The state-control-limit-based rejuvenation policy leads to lower unavailability of the virtualized Cloud Server in comparison with the lifetime-based rejuvenation policy.

In this article, an intelligent public transportation system based on ZigBee 3.0 technology is proposed after researching conditions of recent public transportation systems that mostly adopt GPS or Beidou satellite positioning technology, 3G/4G communication technology and GIS technology. This system includes the principle and design of system architecture, design of intelligent public transportation system network, the solution of moving-bus positioning, the auto-announce function and design of intelligent public transportation stop board. Contrasted with recent bus systems, this system has the virtue of low construction cost, low running cost, low implementation difficulty and high intelligent level. ZigBee 3.0 protocol is compatible with Wi-Fi, too. Without additional ZigBee chip in smart phones, the intelligent public transportation system based on ZigBee 3.0 can interconnect with smart phones directly. It will break the barrier between citizens and intelligent public transportation systems, and the real intelligence can be realized.

This paper proposes a new method for estimating the player’s and ball’s 3D position information from monocular broadcast videos. For players, the homography between image and playfield is used to estimate their positions. By analyzing the geometry relation between the ball, its “virtual” shadow and camera position, we derive equations for estimating the flying ball’s 3D position. Moreover, we propose a method to predict the flying plane if it cannot be determined from images. This method designs a new cost function, which arrives at the minimum when the predicted flying plane is reasonable. This method has at least two merits. One is that it can estimate the flying ball’s position without referring to other objects with known height; the other is that only one assumption is made and the camera is in a fixed position. Experimental results are satisfying.

The existed parallel collaborative filtering algorithm based on co-occurrence matrix (CMCF) consumes a lot of time in the construction of co-occurrence matrixes and calculation of matrix multiplication. It also ignores the role of neighboring users, so it will influence the accuracy of recommendation. In order to solve this problem, this paper proposes the improved parallel collaborative filtering algorithm (IPCF) and its implementation on spark. The experimental results show that the improved parallel collaborative filtering algorithm in this paper has better running efficiency and higher recommendation accuracy.

Distributed Denial of Service (DDoS) attacks on the cloud computing platform has become one of the key issues affecting cloud security. According to the sources of security threat of cloud computing platform, construct secure routing alliance, filter and resist DDoS from the route of cloud user to cloud computing center, design data forwarding mechanism and fault nodes replacement mechanism. The strategy of secure overlay services is combined with the structural characteristics of the ubiquitous routing platform to defend against DDoS attacks. The Chord ring is improved, the nodes are divided according to the distance in the physical network, and the Chord algorithm is avoided repeatedly ignoring the forwarding of physical paths. Since the original Chord algorithm is applied to the P2P network, in order to make it more suitable for the hierarchical physical topology, only the first three jumps of the Chord algorithm's query steps are taken. Fault nodes replacement mechanism uses virtual machine technology to convert nodes in the network into a large number of virtual nodes and serve as backup nodes in the security structure in time to replace the attacked nodes with backup nodes to minimize the impact of attacks on the nodes. The simulation results show that with the increase of the number of nodes, the data passing rate of the secure routing alliance can exceed 90% and the pass rate can be guaranteed to be over 35% when the number of attack nodes is large, which ensures data security and the availability of the transmission paths.

With the development of remote sensing technology and the diversification of sensors, remote sensing image data reveals the trend of “three features” -- high resolution, hyper spectral and multi-temporal. As the increasing demand of remote sensing information, considerable amounts of data will be acquired, transmitted and stored in various remote sensing applications, which, without doubt, sets higher requirements for data processing. To solve the above problems, according to the feature of compressed sensing theory, which original image can be reconstructed by low sampling data, we develop a new method of Remote Sensing Image Block Compressed Sensing Reconstruction Based on Augmented Lagrangian Method TV. It represented remote sensing image sparsely by means of block sampling and joint sparse representation model. Besides, it also combined the total Variation and Augmented Lagrangian method to optimize the solution and implemented the algorithm of the model. Finally, it created a remote sensing image with low distortion. Furthermore, it also increased efficiency in data transmission and reduced data storage. Simulation test results confirm the validity of algorithm proposed in this paper and also suggest that it can achieve better effects of a distinct advantage in PSNR, which is remote sensing image reconstruction, in comparison with other algorithms.

The paper presents a CFD numerical simulation method for ship engine room ventilation. First, through the discretization of the fluid governing equations, apply the basic physical model of ship engine room established by GABIT Software to lay out the engine room outlet according to the air supply and then divide the meshes. After the physical model is established, import the FLUENT and then reasonably choose the boundary conditions, solving methods and solving precision. Finally, obtain the optimal scheme by the example of researching the airflow velocity, temperature and humidity distribution under different ventilation schemes, and compare the characteristics of various schemes. The method presented in the paper has a strong significance of theoretical analysis and practical guidance for optimizing the ventilation of the ship engine room.

A self-adaptive fuzzy proportional integral derivative (PID) control method based on genetic optimization is proposed to solve the problem of low precision and low anti-jamming capabilities of the brushless direct current (DC) motor control system of electric vehicles. A double closed-loop speed control system model of the drive motor is established based on an analysis of the mathematic model of a permanent magnet brushless DC motor. Adaptive fuzzy PID control is introduced. The fuzzy membership function is optimized by the genetic algorithm and referred to as the optimized adaptive fuzzy PID control method. The design and simulation of the system are realized by using MATLAB/Simulink. Results show that in the same environment, the genetic algorithm with adaptive fuzzy PID control has better dynamic and static performance than ordinary and fuzzy PID. It has a good speed and anti-interference ability in a typical city driving environment.

A traditional state-space model (SSM) only contains one observation equation. There are some restricted conditions when using the traditional SSM to describe the evolution process between a state indicator and multiple observation indicators instantaneously. In order to solve this problem, this paper puts forward an SSM that has multiple observation equations, which can be applied to multiple observation indicators with different sampling times. The modeling process and parameters evaluation approach of the proposed SSM are studied and given. A simulation study is conducted to indicate advantages of the proposed SSM when sampling times and observation equations are not the same for different observation indicators. Simulation results show that the proposed SSM is more accurate than the traditional SSM in system degradation prediction.

In the process of attribute reduction, the importance degree of a condition attribute is generally measured by means of the dependence degree between the condition attribute and the decision attribute set. If the dependence degree of the condition attribute is 0, we generally think that the condition attribute does not affect the decision results of the decision table and can be directly deleted form the condition attribute set. However, to some extent, it cuts off the connection of the condition attribute and other attributes, resulting in a great loss of valuable information in the decision table. Therefore, based on the fact that the dependency degree of the condition attribute set is more credible than the dependency degree of a single condition attribute, this paper researches the dependency degree of the condition attribute set and puts forward a calculation method for dependency degree of condition attribute set using a discernibility matrix. This paper also presents and proves a theorem to improve the proposed method. The proposed method can quickly get the discernibility matrix and can directly calculate the dependency degree of the condition attribute set. The theoretical analysis and the simulation experiment comparison results all show that the proposed method has better effectiveness and lower time complexity.
Submitted on July 17, 2017; First revised on October 27, 2017; Second revised on November 21, 2017; Accepted on December 21, 2017
References: 20

An external factor is an added condition beyond the normal load of the system affecting the system’s success. Each external factor has an influence on system components, which may lead to system failure. To decrease the effects of different states of the external factors on system components, multistate protections are established to protect the components from the effects of the external factors. The protection states are dependent on the external factor state. When a protection is at different states, the protection may be damaged. In this paper, considering the joint effect of the external factors and protections on system components, a measure of states of the protections on the system reliability is introduced to identify which protection level has the most important influence on the system success. Based on the importance values of different protections, appropriate actions can be applied in system components to improve the system protection and reliability and reduce the effects of external factors. At last, an example of a computer server system is used to demonstrate the proposed method.

In order to solve the problem of incorrectly decode the ADS-B signal and the missing aircraft information caused by overlap and interference, in the information transmission process of 1090ES ADS-B signal, a multi overlapping 1090ES ADS-B signal separation algorithm based on RLS algorithm is proposed. After a comprehensive analysis of the ADS-B signal to a plurality of base stations, the algorithm applies RLS blind source separation and recovers the source signals of ADS-B, thereby improving signal decoding accuracy and the dynamic performance monitoring for aircraft. The paper has verified the signal separation, including signals based on two level overlapped and three level overlapped and signals with noise, and verified the feasibility of RLS algorithm in ADS-B signal separation by MATLAB simulation test.