Ensemble Learning Using Decorrelated Neural Networks

We describe a decorrelation network training method for improving the quality of regression learning in 'ensemble' neural networks NNs that are composed of linear combinations of individual NNs. In this method, individual networks are trained by backpropogation not only to reproduce a desired output, but also to have their errors linearly decorrelated with the other networks. Outputs from the individual networks are then linearly combined to produce the output of the ensemble network. We demonstrate the performances of decorrelated network training on learning the 'three-parity' logic function, a noisy sine function and a one-dimensional non-linear function, and compare the results with the ensemble networks composed of independently trained individual networks without decorrelation training . Empirical results show than when individual networks are forced to be decorrelated with one another the resulting ensemble NNs have lower mean squared errors than the ensemble networks having independently trained individual networks. This method is particularly applicable when there is insufficient data to train each individual network on disjoint subsets of training patterns.     

Feedforward Neural Network;Construction Algorithm;Maximum Likelihood Function;Quasi-newton Methods.

A new method for constructing a feedforward neural network is proposed. The method starts with a single hidden unit and more units are added to the hidden layer one at a time until a network that completely recognizes all its input patterns is constructed. The novel idea about this method is that the network is trained to maximize a certain likelihood function and not to minimize the more widely used mean squared error function. We show that when a new hidden unit is added to the network, this likelihood function is guaranteed to increase and this increase ensures the finite termination of the method. We also provide a wide range of numerical results. The method was tested on the n -bit parity problems and the spiral problem. It was able to construct networks having less than n hidden units that solve the n -bit parity problems for n = 4, 5, 6, 7 and 8. The method was also tested on some real-world data and the networks it constructed were shown to be able to predict patterns not in the training set with more than 95% accuracy.     

Shear layer modelling for bobbin tool friction stir welding

Abstract

This study presents an approach to model the shear layer in bobbin tool friction stir welding. The proposed CFD model treats the material in the weld zone as a highly viscous non-Newtonian shear thinning liquid. A customised parametric solver is used to solve the highly non-linear Navier–Stokes equations. The contact state between tool and workpiece is determined by coupling the torque within the CFD model to a thermal pseudomechanical model. An existing analytic shear layer model is calibrated using artificial neural networks trained with the predictions of the CFD model. Validation experiments have been carried out using 4 mm thick sheets of AA 2024. The results show that the predicted torque and the shear layer shape are accurate. The combination of numerical and analytical modelling can reduce the computational effort significantly. It allows use of the calibrated analytic model inside an iterative process optimisation procedure.     

Genetic tool monitor (GTM) for micro-end-milling operations

Almost all existing tool condition monitoring methods require either the critical parameters of models which are experimentally found or the self-learning algorithms that are trained with existing data. Genetic Tool Monitor (GTM) is proposed to identify the problems by using an analytical model for micro-end-milling operations and genetic algorithm. The current version of the GTM is capable to monitor the micro-end-milling operations without any previous experience and is able to estimate symmetrical wear and local damages at the cutting edges of a tool. Genetic algorithms (GA) are found as a promising health monitoring tool if an expression exists and the necessary computational time is allowable in that particular application. GTM generates meaningful information about the ongoing operation and allows the establishment of rules based on the operators' experience.     

Bootstrapping with Noise: An Effective Regularization Technique

Bootstrap samples with noise are shown to be an effective smoothness and capacity control technique for training feedforward networks and for other statistical methods such as generalized additive models. It is shown that noisy bootstrap performs best in conjunction with weight-decay regularization and ensemble averaging. The two-spiral problem, a highly non-linear, noise-free data, is used to demonstrate these findings. The combination of noisy bootstrap and ensemble averaging is also shown useful for generalized additive modelling, and is also demonstrated on the well-known Cleveland heart data.     

基于有限元方法的动平衡机的设计

首先基于有限元建立了转子的动力学模型,然后给出了动平衡机硬支承的设计方法,同时利用有限元方法计算得出了系统的固有频率和各机构间的阻尼系数;然后分析了现有标定方法的特点,提出了优化的标定方法,该方法优于现有标定方法;最后根据处理后振动信号的特点,结合已有数字信号处理算法,提出一种更精确不平衡量提取方法,以提高动平衡机的灵敏度和精度。实验证明,该系统具有高效、高精和稳定的特点。     

High capacity,small world associative memory models

Models of associative memory usually have full connectivity or, if diluted, random symmetric connectivity. In contrast, biological neural systems have predominantly local, non-symmetric connectivity. Here we investigate sparse networks of threshold units, trained with the perceptron learning rule. The units are given position and are arranged in a ring. The connectivity graph varies between being local to random via a small world regime, with short path lengths between any two neurons. The connectivity may be symmetric or non-symmetric. The results show that it is the small world networks with non-symmetric weights and non-symmetric connectivity that perform best as associative memories. It is also shown that in highly dilute networks small world architectures will produce efficiently wired associative memories, which still exhibit good pattern completion abilities.     

基于Faster R-CNN的刨花板表面缺陷检测研究

为了克服刨花板表面缺陷人工目视检测的局限性,实现对多种缺陷准确、实时检测,提出一种基于Faster R-CNN的检测方法。运用从工厂生产现场获取的各种表面缺陷图,制作成一个包含3566张刨花板表面缺陷图像数据集,其中主要包括胶块、水印、砂痕、杂物、粗刨花5种缺陷类型。通过用该数据集对Faster R-CNN在ZF、VGG16和ResNet101不同特征提取网络下的不同锚点(Anchor)设置模型分别进行训练、验证和测试,并对比了不同参数对检测精度的影响。结果显示,该方法能有效检测刨花板表面缺陷,且模型在ResNet101作为特征提取网络时准确率最高。在对训练好的Faster R-CNN模型的鲁棒性进行评估和验证中,模型对122张新图像的5种缺陷类型进行检测,测试的5种缺陷类型识别率分别为92.31%、91.84%、90.57%、96.88%和95.24%,平均检测率为93.37%,测试结果表明该方法能为基于机器视觉刨花板表面缺陷检测系统提供良好支撑。     

STUDY ON CONSTITUTIVE RELATIONSHIP OF Ti-5Al-2Sn-2Zr-4Cr-4Mo ALLOY AND SIMULATION OF HOT FORMING PROCESSES

.~theSofar,finiteelementmethodhaswidelybeenusedinmetalfoeingprocesses,whichcanhelptoshortenthedevelopmentcycleandreducetheproductcosts.ConstitutiverelationshipisabridgebetweenthedefonnationbehaViorofmaterialsandallkindsOfthermomechanicalparameters,anditisalsoapresupPOsitiontothesimulationOfmetaldeformationprocessesbyusingfiniteelementmethod.Fwhermore,itisusuallynon--linearandcomplex,owingtoavallationinstructUredabingplasticdefonnation,Particularlyoccultinginsupendloys.FOrmanyyears,researche…     

基于机器视觉的金属板材表面波纹度检测方法

常规波纹度测量方法存在检测效率低、检测精度不高等问题,为了实现金属板材表面波纹度的快速准确检测,提出了一种非接触式的表面波纹度无损检测方法,该方法采用改进的中值滤波实现高效快速的图像去噪,使用霍夫变换实现图像的自动旋转,提取了表面图像基于灰度共生矩阵的纹理特征参数,并进行相关性分析。结果表明,对比度、能量、相关和熵等特征参数与波纹度之间存在较好的一致性。以对比度、能量、相关和熵等为输入参数,构建了基于BP神经网络的表面波纹度检测模型,试验验证该方法能够实现金属板材表面波纹度的高效、高精度测量,检测误差仅为49.0%。     

Extracting dimensional information from steel reinforcing bars in concrete using neural networks trained on data from an inductive sensor

A new method is presented for extracting dimensional information from steel bars using images generated by an inductive sensor. The technique is based on the application of two feedforward backpropagation neural networks; one to estimate bar depth and the other to estimate bar diameter. Both of the networks have been trained on a set of data that consists of the peak parameters of six different bars scanned at 41 different bar depths. These input and target data must be pre-processed to obtain a good network generalisation. By testing the two networks with a completely different set of data, accurate performance has been obtained. Real, two-dimensional scan data have then been applied to both of the networks and the bar dimensional parameters have been extracted successfully. The advantage of the neural network method for extracting information is that it continues to operate reliably for very deep bars, for which the signal strength is severely attenuated and manifests a poor signal-to-noise ratio. Depth and diameter measurements have been obtained for bars located down to 58 mm, with errors that satisfy the requirements of the BS 1881 standard. At a depth of 40 mm, these measurements yield an error of ±4%, and this decreases as the depth reduces; in other words, the extracted bar diameter is within the requirements of the DIN 488 standard.     

Reducing the model generation effort for the virtual commissioning of control programs

This paper deals with the field of developing and deploying highly automated manufacturing systems. Especially the stage of the control programming suffers from few applicable verification methods, as existing approaches still include drawbacks. The virtual commissioning, as one example, requires additional effort in creating simulation models to being usable. This work presents a method, which reduces the effort of generating simulation models through an automated reuse and transformation of existing development documents. In a detailed case study, a software prototype implementing the method was executed to derive a simulation model of an existing set of information. The used sources, their transformation and the results are described to give a qualified impression of the method’s possibilities.     

Analysis of Hidden Representations by Greedy Clustering

The hidden layer of backpropagation neural networks (NNs) holds the key to the networks' success in solving pattern classification problems. The units in the hidden layer encapsulate the network's internal representations of the outside world described by the input data. this paper, the hidden representations of trained networks are investigated by means simple greedy clustering algorithm. This clustering algorithm is applied to networks have been trained to solve well-known problems: the monks problems, the 5-bit problem and the contiguity problem. The results from applying the algorithm to problems with known concepts provide us with a better understanding of NN learning. These also explain why NNs achieve higher predictive accuracy than that of decision-tree methods. The results of this study can be readily applied to rule extraction from Production rules are extracted for the parity and the monks problems, as well as benchmark data set: Pima Indian diabetes diagnosis. The extracted rules from the Indian diabetes data set compare favorably with rules extracted from ARTMAP NNs terms of predictive accuracy and simplicity.     

基于遗传算法的Al-5%Cu合金电脉冲孕育处理参数优化

以实验为基础,利用神经网络和遗传算法优化Al-5%Cu合金的电脉冲孕育处理工艺参数。神经网络的输入参数为脉冲电压、脉冲时间和电脉冲孕育处理时熔体温度,输出参数是合金凝固组织的晶粒度。在神经网络训练的基础上,采用遗传算法优化神经网络的输入参数。结果表明,神经网络和遗传算法的组合建模获得了较好的优化结果。     

Lack of combinatorial productivity in language processing with simple recurrent networks

An astronomical set of sentences can be produced in natural language by combining relatively simple sentence structures with a human-size lexicon. These sentences are within the range of human language performance. Here, we investigate the ability of simple recurrent networks (SRNs) to handle such combinatorial productivity. We successfully trained SRNs to process sentences formed by combining sentence structures with different groups of words. Then, we tested the networks with test sentences in which words from different training sentences were combined. The networks failed to process these sentences, even though the sentence structures remained the same and all words appeared on the same syntactic positions as in the training sentences. In these combination cases, the networks produced work–word associations, similar to the condition in which words are presented in the context of a random word sequence. The results show that SRNs have serious difficulties in handling the combinatorial productivity that underlies human language performance. We discuss implications of this result for a potential neural architecture of human language processing.     

PCM控制的新型气液位置伺服系统的研究

本文提出了一种新型气液缸和气液位置伺服系统。该系统采用气液连动控制,充分发挥液压与气动各自的优点,大大提高了位置控制精度,同时克服单纯气动的缺点。在气液回路上,利用两位三通阀、“ＰＣＭ”阀组及油开关阀实现分区控制和准确定位,解决了快速性和定位精度之间的矛盾,通过分析和实验研究表明,此方案是可行的。     

Improving the Generalization Ability of an Artificial Neural Network in Predicting In-Flight Particle Characteristics of an Atmospheric Plasma Spray Process

This paper presents the application of the artificial neural network into an atmospheric plasma spray process for predicting the in-flight particle characteristics, which have significant influence on the in-service coating properties. One of the major problems for such function-approximating neural network is over-fitting, which reduces the generalization capability of a trained network and its ability to work with sufficient accuracy under a new environment. Two methods are used to analyze the improvement in the network’s generalization ability: (i) cross-validation and early stopping, and (ii) Bayesian regularization. Simulations are performed both on the original and expanded database with different training conditions to obtain the variations in performance of the trained networks under various environments. The study further illustrates the design and optimization procedures and analyzes the predicted values, with respect to the experimental ones, to evaluate the performance and generalization ability of the network. The simulation results show that the performance of the trained networks with regularization is improved over that with cross-validation and early stopping and, furthermore, the generalization capability of the networks is improved; thus preventing any phenomenon associated with over-fitting.     

Recognition of chatter with neural networks

Chatter deteriorates surface finish, reduces tool life, and damages machine tools. A chatter development prediction procedure is proposed for the cylindrical turning of long slender bars. The procedure uses two synthetically trained neural networks to recognize the harmonic acceleration signals and their frequency, and based on these observations, the future vibration characteristics of the system are estimated. The developed neural networks are capable of identifying 98% of the harmonic signals with over 90% certainty and estimate their frequencies with less than ±5% error from very short data sequences (only 11 sampled points). The accuracy of the neural networks is equivalent to time domain time series method based approaches; however, the proposed procedure can be implemented very quickly by using commercially available neural network hardware and software, and can use the new neural network chips to make the estimations very quickly by using parallel processors. The validity of the chatter prediction procedure is also demonstrated on the experimental data.     

Estimation of Transformation Temperatures in Ti–Ni–Pd Shape Memory Alloys

The present study focused on estimating the complex nonlinear relationship between the composition and phase transformation temperatures of Ti–Ni–Pd shape memory alloys by artificial neural networks (ANN). The ANN models were developed by using the experimental data of Ti–Ni–Pd alloys. It was found that the predictions are in good agreement with the trained and unseen test data of existing alloys. The developed model was able to simulate new virtual alloys to quantitatively estimate the effect of Ti, Ni, and Pd on transformation temperatures. The transformation temperature behavior of these virtual alloys is validated by conducting new experiments on the Ti–rich thin film that was deposited using multi target sputtering equipment. The transformation behavior of the film was measured by varying the composition with the help of aging treatment. The predicted trend of transformational temperatures was explained with the help of experimental results.     

基于多模态集成卷积神经网络的数控机床齿轮箱故障诊断

针对数控机床齿轮箱在实际工作环境中负载多变且噪声干扰大、传统神经网络难以充分提取信号中的故障特征等问题,提出一种多模态集成卷积神经网络（MECNN）用于数控机床齿轮箱故障诊断。该方法将多模态融合技术与多个卷积神经网络结合,利用快速傅里叶变换方法将时域信号转换成频域信号;利用时域信号和频域信号对2个卷积神经网络进行训练,使模型能够分别从时域和频域2个角度提取特征,再将浅层特征融合;最后,将融合后的特征输入到卷积神经网络中进行故障特征的深度挖掘,并进行故障诊断。使用东南大学的齿轮箱数据集进行验证,设计了2种特征融合的方法并进行了对比。实验结果表明：在噪声下,MECNN模型用于故障诊断的准确性和鲁棒性均优于单一的时域CNN和频域CNN。