Fault Diagnosis of Rolling Bearings Based on WPE by Wavelet Decomposition and ELM

The fault diagnosis classification method based on wavelet decomposition and weighted permutation entropy (WPE) by the extreme learning machine (ELM) is proposed to address the complexity and non-smoothness of rolling bearing vibration signals. The wavelet decomposition based on ‘db3’ is used to decompose the signal into four layers and extract the approximate and detailed components, respectively. Then, the WPE values of the approximate (CA) and detailed (CD) components of each layer are calculated and composed to be the feature vectors, which are finally fed into the extreme learning machine with optimal parameters for classification. The comparative study of the simulations based on WPE and permutation entropy (PE) shows that the classification method of seven kinds of signals of normal bearing signals and six types of fault states (7 mils and 14 mils) based on WPE (CA, CD) with the number of nodes in the hidden layers of ELM determined by the five-fold cross-validation has the best performances, the training accuracy can reach 100%, and the testing accuracy can reach 98.57% with 37 nodes of the hidden layer by ELM. The proposed method using WPE (CA, CD) by ELM provides guidance for the multi-classification of normal bearing signals.     

Rapprochement of artificial intelligence and dynamics

It is proposed that continuous time is in effect discretized in the brain by dynamic pattern recognition mechanisms in neurons. Time discretization is required to support formal computations in continuous time systems consisting of a large number of components. The ability to perform formal computations is necessary if the system is to execute high level algorithms of the type used in present day artificial intelligence. The weakness of such algorithms is that they work efficiently only when the forms of patterns and objects presented to them are highly constrained. The dynamic mechanisms which discretize the brain's time line also serve to code patterns into constrained forms suitable for high level processing.     

Mathematics learning support and engagement in first year engineering

This paper considers the effects of both free optional mathematics learning support and engagement on the mathematics performance in a foundation mathematics subject of a cohort of engineering students entering university with poor mathematical skills. New engineering students were directed to either a foundation or standard mathematics subject based on the results of a placement test. For students in the foundation subject, it was found that high levels of learning support were associated with greater improvement over the semester. Some form of learning support was used by 57.9% of the students, a reasonably high proportion of the cohort. Some factors for this high level of use of learning support are considered. One possible factor, the placement test, appears to have had a positive effect. Engagement in the subject activities as measured by tutorial attendance and learning management system use was found to have a positive association with final mark. Students who utilized a high level of learning support were more likely to be more engaged with the subject, making it impossible to draw conclusions about improvements being solely due to the use of learning support.     

Gaussian kernel approximation algorithm for feedforward neural network design

A Gaussian kernel approximation algorithm for a feedforward neural network is presented. The approach used by the algorithm, which is based on a constructive learning algorithm, is to create the hidden units directly so that automatic design of the architecture of neural networks can be carried out. The algorithm is defined using the linear summation of input patterns and their randomized input weights. Hidden-layer nodes are defined so as to partition the input space into homogeneous regions, where each region contains patterns belonging to the same class. The largest region is used to define the center of the corresponding Gaussian hidden nodes. The algorithm is tested on three benchmark data sets of different dimensionality and sample sizes to compare the approach presented here with other algorithms. Real medical diagnoses and a biological classification of mushrooms are used to illustrate the performance of the algorithm. These results confirm the effectiveness of the proposed algorithm.     

A correlative classifiers approach based on particle filter and sample set for tracking occluded target

Target tracking is one of the most important issues in computer vision and has been applied in many fields of science, engineering and industry. Because of the occlusion during tracking, typical approaches with single classifier learn much of occluding background information which results in the decrease of tracking performance, and eventually lead to the failure of the tracking algorithm. This paper presents a new correlative classifiers approach to address the above problem. Our idea is to derive a group of correlative classifiers based on sample set method. Then we propose strategy to establish the classifiers and to query the suitable classifiers for the next frame tracking. In order to deal with nonlinear problem, particle filter is adopted and integrated with sample set method. For choosing the target from candidate particles, we define a similarity measurement between particles and sample set. The proposed sample set method includes the following steps. First, we cropped positive samples set around the target and negative samples set far away from the target. Second, we extracted average Haar-like feature from these samples and calculate their statistical characteristic which represents the target model. Third, we define the similarity measurement based on the statistical characteristic of these two sets to judge the similarity between candidate particles and target model. Finally, we choose the largest similarity score particle as the target in the new frame. A number of experiments show the robustness and efficiency of the proposed approach when compared with other state-of-the-art trackers.     

Evaluating continuous chirality measure as a 3D descriptor in chemoinformatics applied to asymmetric catalysis

Continuous Chirality Measure (CCM) is a computational metric by which to quantify the chirality of a compound. In enantioselective catalysis, prior work has postulated that CCM is correlated to selectivity and can be used to understand which structural features dictate catalyst efficacy. Herein, the investigation of CCM as a metric capable of guiding catalyst optimization is explored. Conformer-dependent CCM is also explored. Finally, CCM is used with Sterimol parameters to significantly improve the performance of Random Forest models.     

InRout - A QoS aware route selection algorithm for industrial wireless sensor networks

Wireless sensor networks are a key enabling technology for industrial monitoring applications where the use of wireless infrastructure allows high adaptivity and low cost in terms of installation and retrofitting. To facilitate the move from the current wired designs to wireless designs, concerns regarding reliability must be satisfied. Current standardization efforts for industrial wireless systems lack specification on efficient routing protocols that mitigate reliability concerns. Consequently, this work presents the InRout route selection algorithm, where local information is shared among neighbouring nodes to enable efficient, distributed route selection while satisfying industrial application requirements and considering sensor node resource limitations. Route selection is described as a multi-armed bandit task and uses Q-learning techniques to obtain the best available solution with low overhead. A performance comparison with existing approaches demonstrates the benefits of the InRout algorithm, which satisfies typical quality of service requirements for industrial monitoring applications while considering sensor node resources. Simulation results show that InRout can provide gains ranging from 4% to 60% in the number of successfully delivered packets when compared to current approaches with much lower control overhead.     

基于深度学习的大规模电磁信号识别

近年来,很多高质量的数据集支撑了深度学习在计算机视觉、语音和自然语言处理领域的快速发展.但在电磁信号识别领域仍缺乏高质量的数据集,为促进深度学习在电磁信号识别中的应用,本文基于广播式自动相关监视(ADS-B)建立了一个大规模的真实电磁信号数据集.首先设计了一个自动数据收集和标注系统,在开放和真实的场景中自动捕获ADS-...     

基于支持向量机的病毒程序检测方法

支持向量机是一种对于小样本具有良好学习性能的机器学习方法.本文将支持向量机方法用于病毒程序的检测中,可以改善其它方法在先验知识较少情况下的推广能力的问题.仿真实验结果看出,该方法在训练样本数相对较少的情况下,仍然具有较高的检测率和正确率,同时也具有较低的虚警率.