广义同余神经网络在旋转机械常见故障诊断中的应用

通过利用广义同余函数代替传统神经网络的激励函数,提出一种改进的广义同余神经网络的模型及算法。分析了广义同余神经网络的结构、激励函数、权值调整算法等方面并与传统BP神经网络的异同点进行了比较和研究。通过广义同余神经网络和传统BP神经网络对正弦函数的逼近性能比较,表明最新改进的广义同余神经网络收敛速度快,又具有传统BP神经网络稳定性好的优点。该模型和算法在旋转机械常见故障诊断中提高了收敛速度和诊断精度。

层次广义同余神经网络结构及运算特性分析

提出一种层次广义同余神经网络 ( HGCNN ) ,以有限环上层次整数余神经网络HGCNN为例 ,分析了该网络的运算特性。这种运算保持了神经网络的高度并行结构 ,能够实时完成有限环上的同余运算。给出

平均连接卷积神经网络

在图像分类研究领域,卷积神经网络以其强大的特征提取能力取得了巨大的成就。它有效利用了权值共享结构,大大降低了网络模型的复杂程度,从而成功避免了传统识别过程中复杂的特征提取和数据重建过程。但是当前深度卷积神经网络模型仍然存在有一些问题,比如:反向传播过程中梯度消失或梯度爆炸、训练中过拟合以及网络沉余性等问题。针对上述问题,提出了一种新型的卷积神经网络模型AverageNet。本模型运用"平均连接"的思想,通过适当的跳跃连接,一方面充分利用了前面的输入数据的特征,另一方面不会出现过度的连接现象,降低了模型的参数量,提高了训练效率。网络结构在OlivettiFaces人脸数据集上进行实验,相对于目前流行的算法能够在更少的计算成本下表现出极强的性能。

求解最大流问题的算法

为提高网络最大流问题的算法效率,通过减弱对最短增广链算法的约束,给出一种增载轨算法。将剩余网络替换成余网络,它不必记录分层剩余网络邻接矩阵;根据余网络的特点,进一步将余网络划分成若干个区域,降低算法的空间复杂度。实验结果表明,在BA无标度网络中该算法与最短增广链算法的计算结果相同,其运行效率比最短增广链算法高,为大规模网络最大流问题提供了较为高效的求解算法。

定流值比例的最小双费用流算法研究

现有最小双费用流算法只能求解网络的最大双流问题,并不能得到定流值比例。为此,提出了一种定流值比例的最小双费用流新算法,在求解最小双流和最小费用的基础上,在调整双流值保证定流值比例的同时得到最小费用流。所提出的新算法定义了余网络和费用差,以邻接矩阵为网络数据存储结构,使用Ford算法分别得到两费用的最短增广链,选择费用最小的增广链增广并求出其对应的费用差,从费用差最小的开始调整流值就得到定流值比例下的最小费用。应用该新算法构建定流值比例的最小双费用流算法的运输网络模型,就可以获得最优运输方案。逻辑推理和仿真实验结果均表明,所提出的算法可行、有效,能较好地解决稀疏网络以及复杂网络中定流值比例的最小双费用流问题。

求解最小费用流的一种新算法

网络最小费用流问题是经典的双目标优化问题,其中利用的图论方法主要有负费用回路算法和最小费用路算法。最小费用路(Busacker-Gowan)算法每次增广流值之前都需要搜索一次最小费用路径,导致算法复杂度偏高,并且该算法是在剩余网络的基础上进行增广,使得该算法在计算预定流值最小费用流时有点冗余。针对这些不足,提出了一种求最小费用流的新算法。该算法首先利用改进的Dijkstra算法一次搜索出所有的源点至汇点费用路径,并且在余网络中增广流值。由于余网络比剩余网络构造简单,所以最终提高了算法的时间效率。仿真实验表明,在ER随机网络中提出算法和经典算法的计算结果相同,并且提出算法不管是在稀疏网络还是非稀疏网络中其运行时间比经典算法都要少,同时更适用于稀疏网络。

Deep Spectrum Prediction in High Frequency Communication Based on Temporal-Spectral Residual Network

High frequency(HF) communication is widely spread due to some merits like easy deployment and wide communication coverage. Spectrum prediction is a promising technique to facilitate the working frequency selection and enhance the function of automatic link establishment. Most of the existing spectrum prediction algorithms focus on predicting spectrum values in a slot-by-slot manner and therefore are lack of timeliness. Deep learning based spectrum prediction is developed in this paper by simultaneously predicting multi-slot ahead states of multiple spectrum points within a period of time. Specifically, we first employ supervised learning and construct samples depending on longterm and short-term HF spectrum data. Then, advanced residual units are introduced to build multiple residual network modules to respectively capture characteristics in these data with diverse time scales. Further, convolution neural network fuses the outputs of residual network modules above for temporal-spectral prediction, which is combined with residual network modules to construct the deep temporal-spectral residual network. Experiments have demonstrated that the approach proposed in this paper has a significant advantage over the benchmark schemes.

Pinch Location of the Hydrogen Network with Purification Reuse

In the hydrogen network with the minimum hydrogen utility flow rate,the pinch appears at the point with zero hydrogen surplus,while the hydrogen surpluses of all the other points are positive.In the hydrogen purity profiles,the pinch can only lie at the sink-tie-line intersecting the source purity profile.According to the alternative distribution of the negative and positive regions,the effect of the purification to the hydrogen surplus is analyzed.The results show that when the purification is applied,the pinch point will appear neither above the purification feed nor between the initial pinch point and the purification feed,no matter the purification feed lies above or below the initial pinch point.This is validated by two case studies.

Redundancy Elimination in GPRS network

The mechanisms of TCP’s retransmission and reset will result in redundant packets. These redundant packets are often sent unnecessarily to the user over a slow last-hop link delaying useful traffic. This is a problem for wide-area wireless links, such as General Packet Radio Service (GPRS), because unnecessary transmissions waste already limited radio bandwidth, battery power at the mobile terminal and incurs monetary cost due to charging by data volume. The paper first describes a GPRS model, then discusses how to eliminate the redundant packets in GPRS network and presents the simulation results in Network Simulation 2 (NS 2). The more traffic is, the more the network can benefit. In heavy traffic, it can even get more than 30% improvement in throughput. Average delay and loss percent are also lowered.

An energy efficiency clustering routing protocol for WSNs in confined area

Wireless sensor networks (WSNs) are important application for safety monitoring in underground coal mines, which are difficult to monitor due to natural conditions. Based on the characteristic of limited energy for WSNs in confined underground area such as coal face and laneway, we presents an energy- efficient clustering routing protocol based on weight (ECRPW) to prolong the lifetime of networks. ECRPW takes into consideration the nodes' residual energy during the election process of cluster heads. The constraint of distance threshold is used to optimize cluster scheme. Furthermore, the protocol also sets up a routing tree based on cluster heads' weight. The results show that ECRPW had better perfor- mance in energy consumption, death ratio of node and network lifetime.

A sink moving scheme based on local residual energy of nodes in wireless sensor networks

In the application of periodic data-gathering in sensor networks,sensor nodes located near the sink have to forward the data received from all other nodes to the sink,which depletes their energy very quickly.A moving scheme for the sink based on local residual energy was proposed.In the scheme,the sink periodically moves to a new location with the highest stay-value defined by the average residual energy and the number of neighbors.The scheme can balance energy consumption and prevent nodes around sink from draining their energy very quickly in the networks.The simulation results show that the scheme can prolong the network lifetime by 26%-65%compared with the earlier schemes where the sink is static or moves randomly.

Prediction about residual stress and microhardness of material subjected to multiple overlap laser shock processing using artificial neural network

In this work,the nickel-based powder metallurgy superalloy FGH95 was selected as experimental material,and the experimental parameters in multiple overlap laser shock processing(LSP)treatment were selected based on orthogonal experimental design.The experimental data of residual stress and microhardness were measured in the same depth.The residual stress and microhardness laws were investigated and analyzed.Artificial neural network(ANN)with four layers(4-N-(N-1)-2)was applied to predict the residual stress and microhardness of FGH95 subjected to multiple overlap LSP.The experimental data were divided as training-testing sets in pairs.Laser energy,overlap rate,shocked times and depth were set as inputs,while residual stress and microhardness were set as outputs.The prediction performances with different network configuration of developed ANN models were compared and analyzed.The developed ANN model with network configuration of 4-7-6-2 showed the best predict performance.The predicted values showed a good agreement with the experimental values.In addition,the correlation coefficients among all the parameters and the effect of LSP parameters on materials response were studied.It can be concluded that ANN is a useful method to predict residual stress and microhardness of material subjected to LSP when with limited experimental data.

Node deployment strategy optimization for wireless sensor network with mobile base station

The optimization of network performance in a movement-assisted data gathering scheme was studied by analyzing the energy consumption of wireless sensor network with node uniform distribution. A theoretically analytical method for avoiding energy hole was proposed. It is proved that if the densities of sensor nodes working at the same time are alternate between dormancy and work with non-uniform node distribution. The efficiency of network can increase by several times and the residual energy of network is nearly zero when the network lifetime ends.

Adaptive Control of Flexible Redundant Manipulators Using Neural Networks

An investigation on the neural networks based active vibration control of flexible redundant manipulators was conducted. The smart links of the manipulator were synthesized with the flexible links to which were attached piezoceramic actuators and strain gauge sensors. A nonlinear adaptive control strategy named neural networks based indirect adaptive control (NNIAC) was employed to improve the dynamic performance of the manipulator. The mathematical model of the 4-layered dynamic recurrent neural networks (DRNN) was introduced. The neuro-identifier and the neuro-controller featuring the DRNN topology were designed off line so as to enhance the initial robustness of the NNIAC. By adjusting the neuro-identifier and the neuro-controller alternatively, the manipulator was controlled on line for achieving the desired dynamic performance. Finally, a planar 3R redundant manipulator with one smart link was utilized as an illustrative example. The simulation results proved the validity of the control strategy.

A NEW APPROACH TO DVC COMBINED WITH ROI EXTRACTION AND RESIDUAL VIDEO CODING FOR MULTIMEDIA SENSOR NETWORKS

The Wyner-Ziv distributed video coding scheme is characterized for its intraframe encoder and interframe decoder which can also approach the efficiency of an interframe encoder-decoder system. In Wyner-Ziv residual coding of video, the residual of a frame with respect to a reference frame is Wyner-Ziv encoded, which can reduces the input entropy and leads to a higher coding efficiency than directly encoding the original frame. In this paper, we propose a new approach of residual coding combined with Region Of Interest (ROI) extraction. Experimental results show that, the proposed scheme achieves better rate-distortion performance compared to conventional Wyner-Ziv coding scheme.

Genetic Algorithm-Based Redundancy Optimization Method for Smart Grid Communication Network

This paper proposes a redundancy optimization method for smart grid Advanced Metering Infrastructure(AMI) to realize economy and reliability targets.AMI is a crucial part of the smart grid to measure,collect,and analyze data about energy usage and power quality from customer premises.From the communication perspective,the AMI consists of smart meters,Home Area Network(HAN) gateways and data concentrators;in particular,the redundancy optimization problem focus on deciding which data concentrator needs redundancy.In order to solve the problem,we first develop a quantitative analysis model for the network economic loss caused by the data concentrator failures.Then,we establish a complete redundancy optimization model,which comprehensively consider the factors of reliability and economy.Finally,an advanced redundancy deployment method based on genetic algorithm(GA) is developed to solve the proposed problem.The simulation results testify that the proposed redundancy optimization method is capable to build a reliable and economic smart grid communication network.

Dynamic Bayesian Network Based Prognosis in Machining Processes

Condition based maintenance (CBM) is becoming more and more popular in equipment main-tenance. A prerequisite to widespread deployment of CBM technology and practice in industry is effective diagnostics and prognostics. A dynamic Bayesian network (DBN) based prognosis method was investigated to predict the remaining useful life (RUL) for an equipment. First, a DBN based prognosis framework and specific steps for building a DBN based prognosis model were presented. Then, the corresponding inference algorithms for DBN based prognosis were provided. Finally, a prognosis procedure based on particle filtering algorithms was used to predict the RUL of drill-bits of a vertical drilling machine, which is commonly used in industrial process. Preliminary experimental results are promising.

A Novel Transient Current-Based Differential Algorithm for Earth Fault Detection in Medium Voltage Distribution Networks

This paper presents a novel transient current differential algorithm for earth fault detection in unearthed （isolated） and compensated neutral medium voltage （MV） networks. The proposed algorithm uses the transient residual currents, which are very sensitive for earth faults detection. The transient values of residual currents are calculated for each feeder in the network and used as an earth fault indicator. The flow of residual currents is investigated. It is found that the residual current for the faulted feeder is equal to the summation of all residual currents for all other healthy feedersl Based on this investigation, a differential technique is proposed. A percentage restrain performance is proposed to ensure the selectivity and security of the algorithm. The transient algorithm is very sensitive for earth fault incidence. To apply the proposed algorithm, the residual currents can be measured easily by one sensor for each feeder with no need to voltage measurement. The proposed algorithm is less dependent on the fault resistance and the faulted feeder parameters. The network is simulated by ATP/EMTP program. Different fault conditions are covered in the simulation process： different fault inception angles, fault locations and fault resistances.

A multipath routing protocol for wireless sensor network for mine security monitoring

Underground mining is a hazardous industrial activity. In order to provide a safe working environment for miners, a Wireless Sensor Network (WSN) technology has been used for security monitoring. It can provide a wide range of surveillance with a relatively low cost. In this study, an Energy-Based Multipath Routing (EBMR) protocol is proposed, which considers residual energy capacity and link quality in choosing hops and routing paths. Hops and paths with a high residual energy capacity and link quality will have the best chance to be selected to transmit data packages. Since the EBMR stores several routes in the routing table, when the current path fails, another path will be chosen to fulfill the task immediately. In this way, EBMR improves reliability and decrease time latency. Compared to AOMDV and REAR, EBMR decreases time latency by 51% and 12%.

Distributed coordinated adaptive tracking in networked redundant robotic systems with a dynamic leader

Distributed coordinated control of networked robotic systems formulated by Lagrange dynamics has recently been a subject of considerable interest within science and technology communities due to its broad engineering applications involving complex and integrated production processes,where high flexibility,manipulability,and maneuverability are desirable characteristics.In this paper,we investigate the distributed coordinated adaptive tracking problem of networked redundant robotic systems with a dynamic leader.We provide an analysis procedure for the controlled synchronization of such systems with uncertain dynamics.We also find that the proposed control strategy does not require computing positional inverse kinematics and does not impose any restriction on the self-motion of the manipulators;therefore,the extra degrees of freedom are applicable for other sophisticated subtasks.Compared with some existing work,a distinctive feature of the designed distributed control algorithm is that only a subset of followers needs to access the position information of the dynamic leader in the task space,where the underlying directed graph has a spanning tree.Subsequently,we present a simulation example to verify the effectiveness of the proposed algorithms.