Search

The vibration-based analysis is an effective technique for planetary gearbox fault diagnosis, but a difficult task is how to accurately identify fault features from noisy vibration signals. In this paper, a nonconvex wavelet thresholding total variation (WATV) denoising method is proposed for planetary gearbox fault diagnosis, which combines wavelet-domain sparsity and total variation (TV) regularization. The TV regularization algorithm is employed to modify the retained wavelet coefficients so that the occurrence of oscillations caused by wavelet thresholding is suppressed. To overcome the underestimation shortcoming of L1-norm regularization, nonconvex penalty function regularization is used to strongly promote the sparsity of estimation while guaranteeing that the global optimal solutions are obtained even though the objective function is nonconvex. Then, the split augmented Lagrangian shrinkage (SALSA) method is developed to solve the nonconvex WATV denoising problem. Two experimental studies are performed to verify the performance and effectiveness of the proposed method. Comparisons with the soft thresholding and basis pursuit denoising (BPD) methods show that the proposed method can accurately estimate the fault features from vibration signals, which means that the proposed method is an effective and promising tool for planetary gearbox fault diagnosis.

To solve the problem that the traditional feature selection methods, such as PCA and LDA, are unable to get the nonlinear relationship between characteristics. Deep belief networks cannot eliminate the noise and missing value, which affect the accuracy of the software defect prediction (SDP) model. Not only the methods of feature selection, but data preprocessing and learning algorithm can also affect the precision of the defect prediction model. This thesis uses deep belief networks and SVM to construct an SDP model (DBN-SVM) to increase prediction precision. Using denoising autoencoders and SVM to build an SDP model (DA - SVM), compared with the DBN - SVM, DA - SVM model not only improves the prediction precision, but also enhance the robustness of the model. The thesis also proposes an SDP model framework which includes data preprocessing, feature selection and learning algorithm.

For traditional fault diagnosis methods, it is usually necessary to manually extract features and then classify them by pattern recognition. Convolutional neural networks (CNN) can self-learn features from raw data and classify faults. The convolutional neural network is sensitive to the interval points of the training set signals, and the increased interval points will reduce the accuracy of the model for fault identification. Therefore, we propose a multi-scale batch normalized convolutional neural network model (MSBNCNN), with different convolution steps at each scale. The model improves the training speed while reducing the sensitivity to the training set signal interval points, reducing the number of training sets, and making the model diagnosis fault more stable.

Traffic congestion in the urban area has become a serious social problem. In order to alleviate congestion and reduce economic costs, it is important to predict the current and future traffic flows of road network nodes. Unfortunately, there are some challenges that make this work difficult. Firstly, the traffic flow is changeable and difficult to predict. Secondly, there are massive traffic data, but the quality of data is not high. For instance, when the traffic lights are just turning green, the speed of the vehicle can't represent the speed of the road vehicle. To this end, this paper proposes a traffic congestion prediction and detection algorithm based on data analysis. One of the most important parts of our algorithm is a short-term prediction method to predict the future traffic flow, which is called SRBDP. Moreover, we designed data characteristics and use clustering methods and a small amount of human intervention to determine the historical data congestion situation. Finally, we demonstrate the effectiveness of our algorithm through a real traffic data set.

The early wear fault signal of diesel engine is weak, and the fault can be reflected better by analyzing the acceleration signal, while the signal is unstable and cyclical. In order to analyze the signal characteristics better, a method of feature level enhancement is proposed. At first, change the acceleration signal into steady state signal by order resampling. Then, the time-frequency image of high resolution is obtained by synchrosqueezed wavelet transform. Put the time-frequency matrix map to the polar coordinates to enhance the fault feature. By extracting the texture features of the polar coordinates, the classification and diagnosis of faults are classified. Simulation and example signal show that the proposed method can achieve the accurate classification of fault signals.

With the rapid increasing impact of the mobile Internet on people's daily life, accessing the Internet via mobile devices is showing an obvious trend, which brings a large amount of network traffic at the same time. This paper analyzes HTTP traffic characteristics based on real world data collected from a commercial Internet Service Provider (ISP) in charge of some regions of China. We investigate the variation characteristics of HTTP traffic over time. Five common types of HTTP messages are also presented with their percentages in real network environment. In Particular, after the analysis of the distributions of HTML reply lengths and HTTP header numbers, we see the relationship between them can be described as linearity, which can be much instructive to some engineering works. Furthermore, we find the tag is not contained in a majority of HTML messages by extracting their HTML heads in accordance with certain rules. In this paper, we design a data acquisition system applied in the mobile Internet which can evaluate the load of network traffic by capturing and processing packets real-timely and efficiently.

As real-time mobile multimedia becomes more and more popular in modern life, Quality of Service (QoS) has drawn more and more attention. IEEE 802.11 has been widely deployed to provide broadband wireless network access due to its low cost, simplicity and high bandwidth capacity. However, its poor network availability and frequent network disconnection in a mobile scenario challenge real-time services. This paper presents a caching and pre-fetching mechanism based on the usage of Content Centric Network (CCN) in mobile environments. The idea is to anticipate disconnections and prefetch the content that a user is expected to access in the near future. Hence, when to trigger the caching and pre-fetching module is an essential part of the objective. This paper focuses on the design and implementation of a disconnection predictor algorithm. The method is to predict the time before a terminal disconnects from an Access Point (AP) by monitoring Received Signal Strength Indicator (RSSI). This paper evaluates and compares the used algorithms by experiments.

In this paper, supervisory predictive control (SPC) considered as a improved generalized predictive control is developed for coordinated control system of 1000MW once through boiler. This strategy has the abilities of self-learning and self-adaption of GPC and the characteristics of simple construction, high reliability and strong robustness of conventional PID controller. What's more the objective function includes not only the performance of the system but also the economic indicators, so this algorithm could achieve more satisfied performance on keeping steady state, tracking capability and disturbance rejection as well as the economics. The simulations investigate that the SPC is better than a traditional PID controller both on the solving nonlinearity, coupling, delay and uncertain factors problems and economics optimization.

Permutation entropy and fractal theory are new methods which are used to study irregularity and nonlinearity of a complex system in nature. Principle of Permutation Entropy (PE) and Correlation Dimension (CD) is introduced in this paper firstly, their differences in describing the complexity of vibration signal were analyzed and the selection methods of algorithm parameters were put forward, then taking the whole life vibration data of a bearing as an example, PE and CD algorithm were applied to detect abnormal characteristics respectively, and their difference on abnormal detection are compared with each other. It is concluded that the result of two detection methods is basically coincident, but PE has advantages in computational efficiency, which is much more suitable for online monitoring and will play an important role in preventive maintenance of mechanical equipments.

Permutation Entropy (PE) is a new subject which talks about the scrambling and non-linearity of complex system, which has been widely studied in recent years. This paper aims to introduce the basic algorithm of PE firstly, then verify PE using simulated signal, which shows that PE is feasible for fault diagnosis. Finally, the whole life vibration data of a rolling bearing is taken as an example, comparing with variation characteristics of mean square value obtained from the vibration signal, it is well proved that the early abnormity character of vibration signal could be successfully detected by the PE early before the fault occurred. The algorithm of PE was very simple and effective for early fault diagnosis, which supported the feasible idea for the on-line fault diagnosis, so this method will play a key role in the predictive maintenance.

The new high voltage electric energy meter CGDS-12W which integrates the voltage sensor, current sensor and metering device, display unit and remote communication unit together is introduced. The line voltages and currents are converted into micro voltage and current signals (in mV and mA) by the specially designed sensors. The converted signals are fed into a multiplier where a DC voltage signal proportional to the average line power is produced. The DC voltage is converted into a series of pulses by a v/f converter. Then the electric energy is measured by counting the number of pulses in a certain time period. The Changyi Power Administration Bureau (CPAB) installed 12 CGDS-12W equipments in April 2006, and in the mean time 12 conventional electromagnetic meters were also installed aside to make comparison. One year's operation has proved the new equipment's advantages over the conventional meters both in reliability and accuracy. The cost of new equipments is also significantly lower.