Explore the vast range of titles available.

The gear cracks of gear box are one of most common failure forms affecting gear shaft drive. It has become significant for practice and economy to diagnose the situation of gearbox rapidly and accurately. The extracted signal is filtered first to eliminate noise, which is pretreated for the diagnostic classification based on the particle filter of radial basis function. As traditional error back-propagation of wavelet neural network with falling into local minimum easily, slow convergence speed and other shortcomings, the particle swarm optimization algorithm is proposed in this paper. This particle swarm algorithm that optimizes the weight values of wavelet neural network (scale factor) and threshold value (the translation factor) was developed to reduce the iteration times and improve the convergence precision and rapidity so that the various parameters of wavelet neural network can be chosen adaptively. Experimental results demonstrate that the proposed method can accurately and quickly identify the damage situation of the gear crack, which is more robust than traditional back-propagation algorithm. It provides guidances and references for the maintenance of the gear drive system schemes.

Frequent and intense forest fires have posed severe challenges to forest management in many countries worldwide. Since human experts may overlook important signals, the development of reliable prediction models with various types of data generated by automatic tools is crucial for establishing rigorous and effective forest firefighting plans. This study applied recently emerged ensemble learning methods to predict the burned area of forest fires and the occurrence of large-scale forest fires using the forest fire dataset from the University of California, Irvine machine learning repository collected from the northeastern region of Portugal. The results showed that the tuned random forest approach performed better than other regression models did with regard to the prediction accuracy of the burned area. In addition, extreme gradient boosting outperformed other classification models in terms of its predictive accuracy of large-scale fire occurrences. The findings showed that ensemble learning methods not only have great potential for broader application in forest fire automatic precaution and prevention systems but also provide important techniques for forest firefighting decision making in terms of fire resource allocation and strategies, which can ultimately improve the efficiency of forest fire management worldwide.

This paper proposes a method to evaluate the degree of emotion being motivated in continuous music videos based on asymmetry index (AsI). By collecting two groups of electroencephalogram (EEG) signals from 6 channels (Fp1, Fp2, Fz and AF3, AF4, Fz) in the left and right hemispheres, multidimensional directed information is used to measure the mutual information shared between two frontal lobes, and then, we get AsI to estimate the degree of emotional induction. In order to evaluate the effect of AsI processing on physiological emotion recognition, 32-channel EEG signals, 2-channel EEG signals and 2-channel EMG signals are selected for each subject from the DEAP dataset, and different sub-bands are extracted using wavelet packet transform. k -means algorithm is used to cluster the wavelet packet coefficients of each sub-band, and the probability distribution of the coefficients under each cluster is calculated. Finally, the probability distribution value of each sample is sent as the original features into echo state network for unsupervised intrinsic plasticity training; the reservoir state nodes are selected as the final feature vector and fed into the support vector machine. The experimental results show that the proposed algorithm can achieve an average recognition rate of 70.5% when the subjects are independent. Compared with the case without AsI, the recognition rate is increased by 8.73%. On the other hand, the ESN is adopted for the original physiological feature refinement which can significantly reduce feature dimensions and be more beneficial to the emotion classification. Therefore, this study can effectively improve the performance of human–machine interface systems based on emotion recognition.

In the future, big data will become an efficient and useful means for improving urban planning, and machine learning can take city as a simplified and efficient system. We take full advantage of the benefits of new technology, but also clarify that city is not a machine, also cannot fully mechanically control the urban development. This study presents a methodology for identifying low-carbon travel block, which can be used to identify the built environment conducive to residents’ low-carbon travel. We chose the four elements of traffic survey—travel mode, travel time, travel purpose and travel frequency—as the framework to evaluate travel carbon emissions. Using the index data collected from “WeChat,” a popular social-media platform in China and questionnaire surveys, we conducted hotspot analysis of the spatial distribution of travel carbon emissions in GIS. We obtained a comprehensive carbon emissions and its spatial distribution through the superposition of hotspot density surface of different indexes. The results show that E block within the research area has the lowest travel carbon emissions. These results suggest some planning implications from three aspects—land use mode, road network and public service facilities: In the old urban district of Pucheng, the ratio of residential building area and other types’ building area should be “4:1–3:1”; and we should develop the travel model of bicycle, and the interval of bicycle lanes should be 350–450 m; The ratio of walking road to total road area should be 15–20%, and the width of road should be restricted. Coverage of transit site buffered for the radius of 150 m is 40–50%, coverage of shopping services buffered for the radius of 50 m is 45–60%, and coverage of recreational facilities buffered for the radius of 100 m is 50–70%. The results confirm that “functional mixing” and “dense road network” are beneficial to residents’ low-carbon travel proposed by the predecessors. At the same time, we found that not the higher volume rate is, the more favorable for low-carbon travel. Small cities have limited number of population and scattered distribution of professional posts, which are not suitable for the traditional mode of improving the volume ratio and the bus system. It is not that the higher the bus station coverage is, the better for residents to travel as low-carbon, and the high popularity of public transportation in small cities will increase the carbon emission of residents. The study provides a new way to evaluate the carbon emission assessment of blocks and provides a basis for block planning with low-carbon concept.

To realize real-time fault detection in power devices and enhance reliability of inverter circuits, this paper proposes a detection method based on Park’s transform algorithm and neural network. Park’s transform is applied to obtain the three-phase current base wave amplitude as the characteristic variable for fault detection. Faulty switch devices can be located using a backpropagation neural network in combination with simple logic analyses. The simulation results verify the effectiveness and the feasibility of the proposed method.

Named data networking (NDN) is a novel network architecture which adopts a receiver-driven transport approach. However, NDN is the name-based routing and source uncontrollability, and network congestion is inevitable. In this paper, we propose an adaptive congestion control protocol (ACCP) which is divided into two phase to control network congestion before affecting network performance. In the first phase, we employ the time series prediction model based on deep learning to predict the source of congestion for each node. In the second phase, we estimate the level of network congestion by the average queue length based on the outcomes of first phase in each router and explicitly return it back to receiver, and then the receiver adjusts sending rate of Interest packets to realize congestion control. Simulation experiment results show that our proposed ACCP scheme has better performance than ICP and CHoPCoP in terms of the high utilization and minimal packet drop in a multi-source/multi-path environment.

Fingerprint recognition technology is widely used as a kind of powerful and effective authentication method on various mobile devices. However, most mobile devices use small-area fingerprint scanners, and these fingerprint scanners can only obtain a part of the user’s fingerprint information. Besides, traditional fingerprint recognition algorithms excessively rely on the details of fingerprints, and their recognition performance has great limitations in mobile devices which can only get partial fingerprint images due to fingerprint scanners. This paper proposes a partial fingerprint recognition algorithm based on deep learning for the recognition of partial fingerprint images. It can improve the structure of convolutional neural networks, use two kinds of loss functions for network training and feature extraction and finally improve the recognition performance of partial fingerprint images. The experimental results show that the fingerprint recognition algorithm in this paper has a better performance than the existing fingerprint recognition algorithm based on deep learning on the problem of partial fingerprint classification and fingerprint recognition in the public dataset NIST-DB4 and self-built dataset NCUT-FR.

For the fault tree analysis, a basic event probability is often complicated. The probability is not constant and even can be represented by function. In order to analyze the system reliability and related characteristics, we represent the probabilities of the basic events by functions. The variables of the function are n influencing factors on the basic events. We extend the top event probability from the constant value to n  + 1-dimensional space considering n influencing factors, and the probability is n  + 1th dimension. Further research the n  + 1-dimensional space with related mathematical methods, and then, transform the system probability analysis into the problem of mathematic. The above ideas are the space fault tree (SFT). In SFT, component fault probability distribution replace basic event probability and system fault probability distribution replace top event probability. In this paper, we research the electrical system fault probability distribution and explain the related construction process. The main factors influencing the system are working temperature c and working time t . This paper constructs the three-dimensional fault probability distribution of the components and the system, and the probability importance and criticality importance of the components. With partial derivation of the system fault probability distribution by the c and t , we study the change trend of the fault probability. The optimal replacement schemes of components and the scheme considering the cost are obtained. The results show SFT is feasible and reasonable to analyze the fault probability of system under multi-factor influence and suitable for deep learning of the characteristics of the system reliability change.

Computational complexity and sample selection are two main factors that limited the performance of online sequential extreme learning machines (OS-ELMs). This paper proposes a new model that introduces the concept of hybrid kernel and sample selection method based on an online learning model using a membership function. In other words, an online sequential extreme learning machine based on a hybrid kernel function (HKOS-ELM) is presented. The algorithm only calculates the kernel function to determine the final output function, mostly solving the computational complexity of the algorithm. The hybrid kernel function proposed in this paper has the advantages of strong learning ability and good generalization performance of single kernel function. Based on the classification essence of the OS-ELM classification, the membership function is introduced into the sample selection to remove the noise point and the outlier point. The experimental results showed that the HKOS-ELM algorithm adding the membership degree with mixed kernel functions preserves the advantages of kernel functions and online learning and improves the classification performance of the system.

It is key indexes of worsted yarn quality such as worsted yarn strength index, etc., and it can well control worsted yarn quality by predicting yarn strength index, etc. Generally, it is generally used to predict yarn strength indexes such as multiple linear regression (MLR) algorithm, support vector machine (SVM) and backpropagation neural network (BPNN). This paper proposes a new neural network; it combines convolutional neural network (CNN) with general regression neural network (GRNN), which is written as the CNN–GRNN. It used 1900 sets of data to train CNN–GRNN, SVM and BPNN. It tested CNN–GRNN, MLR, SVM and BPNN with 10 sets of data. The CNN–GRNN neural network is the best accuracy among these four algorithms.