Greyscale based learning in BPNN for image restoration problem

A new method of back propagation learning with respect to the problem of image restoration which is named as greyscale based learning in back propagation neural networks ( BPNN) is investigated. It is observed that by using this method the value of mean square error ( MSE) decreases significantly. In addition,this method also gives good visual results when it is applied in image restoration problem. This method is also useful to tackle the inherited drawback of falling into local minima by reducing its effect on overall system by bifurcating the learning locally different for different grey scale values. The performance of this algorithm has been studied in detail with different combinations of weights. In short,this algorithm provides much better results especially when compared with the simple back propagation algorithm with any further enhancements and without going for hybrid solutions.     

Improved BP Neural Network for Transformer Fault Diagnosis

The back propagation （BP）-based artificial neural nets （ANN） can identify complicated relationships among dissolved gas contents in transformer oil and corresponding fault types, using the highly nonlinear mapping nature of the neural nets. An efficient BP-ALM （BP with Adaptive Learning Rate and Momentum coefficient） algorithm is proposed to reduce the training time and avoid being trapped into local minima, where the learning rate and the momentum coefficient are altered at iterations. We developed a system of transformer fault diagnosis based on Dissolved Gases Analysis （DGA） with a BP-ALM algorithm. Training patterns were selected from the results of a Refined Three-Ratio method （RTR）. Test results show that the system has a better ability of quick learning and global convergence than other methods and a superior performance in fault diagnosis compared to convectional BP-based neural networks and RTR.     

一种递归型模糊神经网络用于动态系统辨识的研究

针对静态模糊神经网络对动态系统辨识精度低的特点,在T-S模糊神经网络标准结构基础上,通过在输入层与状态层间加入可以记忆暂态信息的递归层,一种新的T-S递归型模糊神经网络(TSRFNN)被提出,来提高对动态系统的辨识能力.同时,给出了参数的动态BP学习算法.通过仿真实验,证明提出的TSRFNN对动态非线性系统的辨识比传统静态模糊神经网络(TFNN),具有更快的网络收敛速度,更高的辨识精度,更适合于动态系统的辨识.     

Correlation of the positive stress ratio effects on fatigue crack propagation rate of aluminium alloys based on a new parameter da/dS

In-situ SEM （Scanning Electron Microscope） observation of fatigue crack propagation in aluminium alloys reveals that crack growth occurs in a continuous way over the time period during the load cycle. Based on this observation, a new parameter da/dS is introduced to describe the fatigue crack propagation rate, which defines the fatigue crack propagation rate with the change of the applied stress at any moment of a stress cycle. The relationship is given between this new parameter and the conventional used parameter da/dN which describes the fatigue crack propagation rate per stress cycle. Using this new parameter, an analysis has been performed and a model has been set up to consider the effect of the applied stress ratio on the fatigue crack propagation rate. The obtained results have been used to correlate the published test data and a good correlation has been achieved. This method is very easy to use and no fatigue crack closure measurement is needed, therefore this model is significant in engineering application.     

基于神经网络的电力通信网风险评估方法

提出了一种基于神经网络的电力通信网风险评估算法--基于二分法的学习速率自适应BP（back propagation）神经网络算法. 该算法在网络训练过程中使用二分法调整学习速率,使得学习速率在训练过程中不断向最优化方向自动调整. 仿真结果表明,收敛速度、误差精度和训练时间等算法性能得到了优化.     

基于改进Fisher准则的深度卷积神经网络识别算法

为了有效利用深度学习技术自动提取特征的能力,并解决当训练样本量减少或者迭代次数降低时识别性能急速下降的问题,提出了基于Fisher准则的深度学习算法.该方法在前馈传播时,采用卷积神经网络自动提取图像的结构信息等特征,同时利用卷积网络共享权值和池化、下采样等方法减少了权值个数,降低了模型复杂度;在反向传播权值调整时,采用了基于Fisher的约束准则.在权值的迭代调整时既考虑误差的最小化,又同时让样本保持类内距离小,类间距离大,从而使权值能更加快速地逼近有利于分类的最优值,当样本量不足或训练迭代次数不多时可有效地提高系统的识别率.大量的实验结果证明:该基于Fisher准则的混合深度学习算法在标签样本不足或者较少训练次数的情况下依然能达到较好的识别效果.     

油漆调色系统的改进算法

根据现有CAID系统中色彩设计和色彩生产难以协调统一的问题以及油漆厂的实际需求，在原有的“基于神经网络的油漆调色系统”基础上，创建了BP网络样本数据库，找到了一种适用于实际应用的有效的BP改进算法。研究测试的结果表明，BP改进算法能够弥补原有算法精度速度不够理想的不足，解决了电脑调色的问题。     

一种用于医学图像恢复的紧小波标架算法

为更好地消除噪声,保留细节信息,根据图像和噪声的小波系数在频域呈现的不同特性,提出了一种基于区域的消噪方法.将该消噪算法插入到小波标架算法中,在消除噪声的同时恢复了部分丢失的系数.实验结果表明,在没有增加计算复杂度的情况下,无论是峰值信噪比还是视觉效果都有了明显改善.     

基于机器学习的无线信道簇的提取与轨迹追踪

针对时变无线信道抽头簇的提取和轨迹追踪提出了一种新方法：首先在时延-幅度维上采用反向传播（BP）神经网络对无线信道冲激响应（CIR）进行去噪,然后利用k-means聚类算法对有效抽头信号进行分簇,再用基于密度的空间聚类（DBSCAN）算法去除各个簇峰值抽头中的异常值,最后采用多项式拟合对去除异常值后的簇峰值抽头进行拟合,得到其时间变化轨迹.经过仿真和实测数据验证,该方法得到的簇峰值时间变化轨迹与根据几何关系得到的结果一致.     

浅表地层缺陷的动力探测技术

为了对浅表地层缺陷进行探测,本文将以往对材料特性的反演,扩展到对浅层地表空洞缺陷的反演,并采用遗传算法作为反演优化算法,在时域下建立路面结构动态反演的方法,并编制了动力反演程序.在对地表下既有空洞的情况进行正演模拟的基础上,对在不同工况下空洞位置和大小进行了模拟反演分析和工程试验分析.通过对理论数据和实测数据进行分析,验证了反演程序的合理性和有效性.结果表明：利用动力对浅层地表空洞缺陷（大小、位置等）探测具有一定的适用性,为检测浅层地表缺陷提供了新的技术路径.     

遗传算法构建的神经网络及在机械工程中的应用

在分析遗传算法和神经网络优点的基础上,采用遗传进化的方式自动获得神网络的结构、权值和阈值．提出了构建神经网络模型参数的遗传算法分区编码方案,构建了适应度函数并依据个体适应度值的大小动态调整隐层节点及连接权个数的方法,给出了整体算法过程．采用该方法构建的神经网络计算两自由度的机械手参数,并通过实例仿真与常规凭经验构建网络结构及采用BP学习算法相比较,采用遗传算法构建的神经网络具有仿真精度高、占用资源少、计算效率高等优点．     

基于混合算法人工神经网络的回采巷道锚杆支护参数预测

回采巷道锚杆支护参数的确定, 一直是煤巷支护设计的重点和难点． 本文尝试应用误差逆传播和模拟退火混合算法人工神经网络预测锚杆支护参数, 结果表明这一方法具有快速、简便、实用和智能化特点, 可大大提高锚杆支护参数选择的科学性、合理性和有效性．     

Seam tracking control for mobile welding robot based on vision sensor

To solve the seam tracking problem of mobile welding robot, a new controller based on the dynamics of mobile welding robot was designed using the method of backstepping kinematics into dynamics. A self-turning fuzzy controller and a fuzzy-Gaussian neural network (FGNN) controller were designed to complete coordinately controlling of cross-slider and wheels. The fuzzy-neural control algorithm was described by applying the Gaussian function and back propagation (BP) learning rule was used to tune the membership function in real time by applying the FGNN controller. To make the tracking more quickly and smoothly, the neural network controller based on dynamic model was designed, which utilized self-learning and self-adaptive ability of the neural network to deal with the partial uncertainty and the disturbances of the parameters of the robot dynamic model and real-time compensate the dynamics coupling. The results show that the selected control input torques make the system globally and asymptotically stable based on the Lyapunov function selected out; the accuracy of the proposed controller tracing is within ±0.4 mm and can satisfy the requirements of practical welding project.     

FNN的服务型制造企业供应商选择

针对服务型制造条件下供应商选择的新要求和复杂性,将模糊理论与人工神经网络有效结合起来,提出基于模糊神经网络（FNN）的供应商选择策略．利用模糊系统较强的知识表达能力,将神经网络的输入袁示为模糊隶属度,解决定性指标定量化及模糊性问题;利用神经网络较强的自适应、自学习能力和泛化能力,克服供应商评价过程的主观随意性、复杂非线性和动态特性带来的不利影响,从而实现对供应商的实时动态评价;通过选择合适的算法,设计相应的BP神经网络,进行实例仿真．仿真结果表明,该方法能够克服供应商选择评价过程中的主观随意性大、算法复杂以及定性指标定量化问题,并实时动态再现专家评价结果,其最大相对误差不超过0．4841％．     

An ICPSO-RBFNN nonlinear inversion for electrical resistivity imaging

To improve the global search ability and imaging quality of electrical resistivity imaging(ERI) inversion, a two-stage learning ICPSO algorithm of radial basis function neural network(RBFNN) based on information criterion(IC) and particle swarm optimization(PSO) is presented. In the proposed method, IC is applied to obtain the hidden layer structure by calculating the optimal IC value automatically and PSO algorithm is used to optimize the centers and widths of the radial basis functions in the hidden layer. Meanwhile, impacts of different information criteria to the inversion results are compared, and an implementation of the proposed ICPSO algorithm is given. The optimized neural network has one hidden layer with 261 nodes selected by AKAIKE's information criterion(AIC) and it is trained on 32 data sets and tested on another 8 synthetic data sets. Two complex synthetic examples are used to verify the feasibility and effectiveness of the proposed method with two learning stages. The results show that the proposed method has better performance and higher imaging quality than three-layer and four-layer back propagation neural networks(BPNNs) and traditional least square(LS) inversion.     

EMD遗传神经网络方法

针对BP(back propagation)神经网络搜索速度慢、容易陷入局部最小的缺陷,提出了经验模态分解(EMD)遗传神经网络方法,首先用对带噪的信号进行分解,得到信号的各阶本征模函数分量,每个本征模函数分量对应着一个能量不同的频段,即一种故障特征,将各频段能量的特征向量作为优化神经网络的输入样本;其次用遗传算法对神经网络的初始权值和阈值进行优化.利用EMD遗传神经网络方法对滚动轴承多类故障信号进行分析,可提高故障识别能力.     

嵌岩桩竖向承载力预测的遗传BP神经网络模型研究

将遗传算法的全局搜索能力和BP神经网络的高次非线性能力,应用于嵌岩桩竖向承载力研究中。两种算法的结合,避免了BP神经网络收敛于局部极值,使神经网络更快找到全局最优解;并利用济南地区的现场实测资料,建立起了预测嵌岩桩单桩竖向承载力的遗传BP神经网络模型。通过该模型可以建立起高应变动力测试法与静载荷试验之间的内在联系,证明了高应变动力测试法在本地区的可靠性较高。     

自适应深度卷积神经网络模型构建方法

针对传统卷积神经网络(CNN)模型构建过度依赖经验知识、参数多、训练难度大等缺点,同时鉴于复杂多类问题的CNN模型构建策略的重要价值,提出一种自适应深度CNN模型构建方法.首先,将初始网络模型的卷积层和池化层设置为仅含一幅特征图;然后,以网络收敛速度为评价指标,对网络进行全局扩展,全局扩展后,根据交叉验证样本识别率控制网络展开局部扩展,直到识别率达到预设期望值后停止局部网络学习;最后,针对新增训练样本,通过拓展新支路实现网络结构的自适应增量学习.通过图像识别实验验证了所提算法在网络训练时间和识别效果上的优越性.     

一种基于改进型递归神经网络的数据校核方法

对于无法得到数学模型的过程，我们提出了基于改进型自递归神经网络的动态过程数据校核方法，并基于递归网络结构推导出一种BP（Back Propagation－反向传播）训练方法，由此方法而得到的数据校验结果更准确无误。     

改进BP算法及其在滚动轴承故障诊断中的应用

对每次权值和阈值的调整均采用固定不变的学习率,是导致传统ＢＰ算法收敛速度慢的一个主要原因。本文从提高收敛速度及精度出发,对改进ＢＰ算法进行了深入研究,在ＢＰ算法中引入统计思想,给出相关系数定义。基于相关系数,采用变学习率策略,提出两种学习率自适应调整算法,并将其具体应用于滚动轴承的故障诊断中。试验证明,此改进算法的收敛速度比传统ＢＰ算法显著提高。