The Complex System Modeling Method Based on Uniform Design and Neural Network

In this paper, the method based on uniform design and neural network is proposed to model the complex system. In order to express the system characteristics all round, uniform design method is used to choose the modeling samples and obtain the overall information of the system;for the purpose of modeling the system or its characteristics, the artificial neural network is used to construct the model. Experiment indicates that this method can model the complex system effectively.

Application of experimental design techniques to structural simulation meta-model building using neural network

Neural networks are being used to construct meta-models in numerical simulation of structures.In addition to network structures and training algorithms,training samples also greatly affect the accuracy of neural network models.In this paper,some existing main sampling techniques are evaluated,including techniques based on experimental design theory, random selection,and rotating sampling.First,advantages and disadvantages of each technique are reviewed.Then,seven techniques are used to generate samples for training radial neural networks models for two benchmarks:an antenna model and an aircraft model.Results show that the uniform design,in which the number of samples and mean square error network models are considered,is the best sampling technique for neural network based meta-model building.

Seismic velocity inversion based on CNN-LSTM fusion deep neural network

Based on the CNN-LSTM fusion deep neural network,this paper proposes a seismic velocity model building method that can simultaneously estimate the root mean square(RMS)velocity and interval velocity from the common-midpoint(CMP)gather.In the proposed method,a convolutional neural network(CNN)Encoder and two long short-term memory networks(LSTMs)are used to extract spatial and temporal features from seismic signals,respectively,and a CNN Decoder is used to recover RMS velocity and interval velocity of underground media from various feature vectors.To address the problems of unstable gradients and easily fall into a local minimum in the deep neural network training process,we propose to use Kaiming normal initialization with zero negative slopes of rectifi ed units and to adjust the network learning process by optimizing the mean square error(MSE)loss function with the introduction of a freezing factor.The experiments on testing dataset show that CNN-LSTM fusion deep neural network can predict RMS velocity as well as interval velocity more accurately,and its inversion accuracy is superior to that of single neural network models.The predictions on the complex structures and Marmousi model are consistent with the true velocity variation trends,and the predictions on fi eld data can eff ectively correct the phase axis,improve the lateral continuity of phase axis and quality of stack section,indicating the eff ectiveness and decent generalization capability of the proposed method.

Efficient Training of Multi-Layer Neural Networks to Achieve Faster Validation

Artificial neural networks(ANNs)are one of the hottest topics in computer science and artificial intelligence due to their potential and advantages in analyzing real-world problems in various disciplines,including but not limited to physics,biology,chemistry,and engineering.However,ANNs lack several key characteristics of biological neural networks,such as sparsity,scale-freeness,and small-worldness.The concept of sparse and scale-free neural networks has been introduced to fill this gap.Network sparsity is implemented by removing weak weights between neurons during the learning process and replacing them with random weights.When the network is initialized,the neural network is fully connected,which means the number of weights is four times the number of neurons.In this study,considering that a biological neural network has some degree of initial sparsity,we design an ANN with a prescribed level of initial sparsity.The neural network is tested on handwritten digits,Arabic characters,CIFAR-10,and Reuters newswire topics.Simulations show that it is possible to reduce the number of weights by up to 50%without losing prediction accuracy.Moreover,in both cases,the testing time is dramatically reduced compared with fully connected ANNs.

A back-propagation neural-network-based displacement back analysis for the identification of the geomechanical parameters of the Yonglang landslide in China

Xigeda formation is a type of hundredmeter-thick lacustrine sediments of being prone to triggering landslides along the trunk channel and tributaries of the upper Yangtze River in China. The Yonglang landslide located near Yonglang Town of Dechang County in Sichuan Province of China, which was a typical Xigeda formation landslide, was stabilized by anti-slide piles. Loading tests on a loading-test pile were conducted to measure the displacements and moments. The uncertainty of the tested geomechanical parameters of the Yonglang landslide over certain ranges would be problematic during the evaluation of the landslide. Thus, uniform design was introduced in the experimental design,and by which, numerical analyses of the loading-test pile were performed using Fast Lagrangian Analysis of Continua(FLAC3D) to acquire a database of the geomechanical parameters of the Yonglang landslide and the corresponding displacements of the loadingtest pile. A three-layer back-propagation neural network was established and trained with the database, and then tested and verified for its accuracy and reliability in numerical simulations. Displacement back analysis was conducted by substituting the displacements of the loading-test pile to the well-trained three-layer back-propagation neural network so as to identify the geomechanical parameters of the Yonglang landslide. The neuralnetwork-based displacement back analysis method with the proposed methodology is verified to be accurate and reliable for the identification of the uncertain geomechanical parameters of landslides.

Investigation of Optimal Megnetic Properties in NdFeB Magnets by Artificial Neural Network

In order to study the effect of alloy component on magnetic properties of NdFeB magnets, the experiment schemes are carried out by the uniform design theory, and the relationship between the component and the magnetic properties is established by artificial neural network(ANN) predicting model.The element contents of alloys are optimized by the ANN model.Meanwhile, the influences of mono-factor or multi-factor interaction on alloy magnetic properties are respectively discussed according to the curves ploted by ANN model.Simulation result shows that the predicted and measured results are in good agreement.The relative error is every low, the error is not more than 1.68% for remanence Br, 1.56% for maximal energy product (BH)m, and 7.73% for coercivity Hcj.Hcj can be obviously improved and Br can be reduced by increasing Nd or Zr content.Co and B have advantageous effects on increasing Br and disadvantageous effects on increasing Hcj.Influence of alloying elements on Hcj and Br are inverse, and the interaction among the alloying elements play an important role in the magnetic properties of NdFeB magnets.The ANN prediction model presents a new approach to investigate the nonlinear relationship between the component and the magnetic properties of NdFeB alloys.

A New Weight Initialization Method Using Cauchy’s Inequality Based on Sensitivity Analysis

In this paper, an efficient weight initialization method is proposed using Cauchy’s inequality based on sensitivity analy- sis to improve the convergence speed in single hidden layer feedforward neural networks. The proposed method ensures that the outputs of hidden neurons are in the active region which increases the rate of convergence. Also the weights are learned by minimizing the sum of squared errors and obtained by solving linear system of equations. The proposed method is simulated on various problems. In all the problems the number of epochs and time required for the proposed method is found to be minimum compared with other weight initialization methods.

基于实验方案设计的卷积神经网络超参数优化方法

卷积神经网络是人工智能的重要组成部分,在自然语言处理、图像识别等领域表现优异。卷积神经网络模型超参数配置涉及训练策略,在卷积神经网络大模型优化方面起着至关重要的作用。现有超参数优化方法耗时耗力,遍历整个超参数空间,容易陷入局部最优解。首先,构建3个不同深度的自建卷积神经网络作为优化对象,以提高模型在验证集上的准确率为优化目标找到最佳的超参数配置。其次,考虑优化神经网络大模型的训练过程并提高模型性能的需求,提出一种基于实验方案设计的卷积神经网络超参数优化方法。最后,为了验证方法的有效性,依据均匀设计理念构建训练方案,生成超参数优化组合,进行主观经验生成训练方案的对比实验。结果表明:所提出的优化方法在收敛速度、准确率和计算效率上更具优势。该方法为实现卷积神经网络大模型的高效训练提供支持,具有良好的通用性,可以应用于不同规模的卷积神经网络训练任务。

应用人工神经网络预测室内全年动态采光

在建筑设计早期阶段,了解建筑形态参数与室内采光之间的关系对设计优化至关重要。本文采用多层感知器(Multilayer Perceptron,MLP)神经网络,以四种主要特征(室外遮挡情况、建筑形态特征、开窗设置、测点位置信息)作为MLP的输入参数,通过计算机模拟收集的数据来构建神经网络,预测室内的全年自然采光质量(UDI<100 lx、UDI 100~2000 lx、UDI>2000 lx)。研究结果显示多层感知器神经网络模型在测试集中的回归决定系数R 2为0.984,均方误差MSE为11.624,准确性较高。对神经网络进行权重分析的结果表明,外部遮挡物的高度和建筑进深对输出结果影响最为显著。而窗台底部的标高和测点距窗户的距离对输出结果UDI的影响较小。神经网络模型为建筑设计预测日光提供了一种新的智能方法,有助于辅助建筑早期的设计决策。

基于WPSO-BP和L-MBWO的多翼离心风机优化研究

针对多翼离心风机气动性能、噪声情况难以同时改进的问题,提出了一种基于变权重粒子群优化算法的反向传播神经网络风机性能预测模型(WPSO-BP),以及一种基于逻辑混沌初始化的多目标白鲸优化算法(L-MBWO),并将二者应用于多翼离心风机的优化设计中。首先,选取了叶片进出口角、倾斜蜗舌的最大蜗舌半径、叶片切除角度作为设计变量,把风机的全压、效率、声压级作为优化目标;然后,构建了WPSO-BP预测模型,以反映设计变量与优化目标之间的关系,定量分析对比了该模型与BP神经网络预测模型,预测值用于风机的性能优化;接着,将逻辑混沌初始化引入到白鲸优化算法(BWO),基于第三代非支配排序遗传算法(NSGA-Ⅲ)构建了L-MBWO优化算法;最后,在实验验证仿真可靠的前提下,将提出的预测模型和优化算法应用于风机优化,并对优化效果进行了综合分析。研究结果表明:优化后的风机全压增加了34.79 Pa,效率提高了0.67%,噪声降低了1.73 dB,实现了多个优化目标之间的平衡,有效改善了风机的综合性能,为多翼离心风机的优化设计提供了一种新思路。

Analytic design of information granulation-based fuzzy radial basis function neural networks with the aid of multiobjective particle swarm optimization

Purpose–The purpose of this paper is to consider the concept of Fuzzy Radial Basis Function Neural Networks with Information Granulation(IG-FRBFNN)and their optimization realized by means of the Multiobjective Particle Swarm Optimization(MOPSO).Design/methodology/approach–In fuzzy modeling,complexity,interpretability(or simplicity)as well as accuracy of the obtained model are essential design criteria.Since the performance of the IG-RBFNN model is directly affected by some parameters,such as the fuzzification coefficient used in the FCM,the number of rules and the orders of the polynomials in the consequent parts of the rules,the authors carry out both structural as well as parametric optimization of the network.A multi-objective Particle Swarm Optimization using Crowding Distance(MOPSO-CD)as well as O/WLS learning-based optimization are exploited to carry out the structural and parametric optimization of the model,respectively,while the optimization is of multiobjective character as it is aimed at the simultaneous minimization of complexity and maximization of accuracy.Findings–The performance of the proposed model is illustrated with the aid of three examples.The proposed optimization method leads to an accurate and highly interpretable fuzzy model.Originality/value–A MOPSO-CD as well as O/WLS learning-based optimization are exploited,respectively,to carry out the structural and parametric optimization of the model.As a result,the proposed methodology is interesting for designing an accurate and highly interpretable fuzzy model.

基于BAS-BP神经网络结合熵权法多指标优化金蕾复方提取工艺

目的通过正交试验与天牛须搜索算法(BAS)-BP神经网络对金蕾复方的提取工艺参数进行多指标优化。方法在单因素考察得到最佳醇提浓度的基础上,以料液比、提取时间、提取次数作为正交试验考察因素,运用熵权法计算木犀草素、山柰酚、当药宁及干膏得率的综合得分,再建立BAS-BP神经网络模型,以BAS进行寻优,预测最佳提取工艺。结果BAS-BP神经网络优化得到金蕾复方醇提工艺为料液比1∶10、提取0.5 h、提取3次,综合得分为96.3526;正交设计所得最佳工艺参数为料液比1∶10、提取0.5 h、提取3次,综合得分为90.9880。前者略优于后者但差异较小,结合生产实际确定金蕾复方的最佳提取工艺为料液比1∶10,提取0.5 h,提取3次。结论基于BAS-BP神经网络优选所得工艺参数提取效率高、稳定性良好,可为后续开发及质量控制提供参考。

均匀设计和BP神经网络结合G1-熵权法对比优化苗药五香血藤提取工艺

目的:优化五香血藤的提取工艺参数。方法:采用U 6(64)均匀设计,对乙醇质量分数、料液比、提取次数和提取时间进行考察,以五味子醇甲和五味子乙素的含量为评价指标,采用G1-熵权法赋权得到综合评分,通过均匀设计试验进行分析得到最佳提取工艺,再通过BP神经网络建模进行网络模型优化和目标寻优,确定五香血藤的最佳提取工艺参数并进行验证对比2种分析方法。结果:均匀设计分析和BP神经网络建模得到的最优提取工艺综合得分分别为104.22、110.30,故确定BP神经网络模型预测的参数为最佳提取工艺,即为料液比20倍,乙醇质量分数90%,超声50 min,超声提取2次。结论:优选出的五香血藤最佳提取工艺稳定可行,为今后五香血藤有效成分的提取方法和工艺改进提供理论参考。

微生物发酵培养基优化中的现代数学统计学方法

发酵培养基优化已经成为现代发酵工业的研究热点。如今很多现代数学统计方法也已经广泛地应用于微生物发酵培养基的优化工作中,其中以正交试计、均匀设计、响应面优化设计、人工神经网络等最为常用。在实际优化工作中,可以根据不同的研究目的,选用不同的优化设计,以达到最佳的效果。文中对这些常用现代数学统计学方法在培养基优化中的特点进行了总结,并介绍了它们之间的区别以及在实际工作中如何对它们进行选用。

基于经济截割的采煤机运动学参数优化研究

利用Matlab软件编制了采煤机以不同牵引速度、不同截割深度截割不同坚固性系数煤层时的载荷计算程序并生成载荷文本,采用均匀设计法对这些文本进行选择,作为刚柔耦合模型的外载,仿真后通过人工神经网络预测了其他工况下各关键零部件的可靠性。基于神经网络预测结果分析了煤层坚固性系数,采煤机牵引速度以及滚筒截割深度与采煤机工作可靠性的关系。并且在保证采煤机可靠工作的前提下,得到了采煤机经济截割曲线,以及相应的最优生产率。研究表明:该型采煤机截割坚固性系数为3的韧性煤时,推荐牵引速度为4.807 m/min,截割深度为550 mm,此时采煤机落煤率为257.8 t/h,其中,单滚筒理论最大落煤率为165 t/h。将虚拟样机技术与人工神经网络相结合能更快更好地解决工程实际中的多参数复杂优化问题。

遗传算法优化神经网络权值盲均衡算法的研究

将遗传算法与神经网络盲均衡算法相结合,提出了两段式优化神经网络权值的方案。首先利用遗传算法全局搜索能力强的特点优化初始权值,然后发挥BP算法局部搜索速度快的特点得到最佳权值。经计算机仿真表明,该算法与传统BP神经网络盲均衡算法相比,收敛速度加快,稳态剩余误差减小,误码率降低。

基于均匀设计及遗传神经网络的大坝力学参数反分析方法

将均匀设计理论、BP神经网络和遗传算法三者结合起来,应用于大坝力学参数反分析中。首先对基本遗传算法进行改进,使得改进后的遗传算法具有很好的全局和局部寻优能力,将它作为BP神经网络的学习算法,形成遗传神经网络。然后利用均匀设计方法设计大坝坝体和坝基的材料力学参数样本,通过有限元正分析得到坝体的计算位移样本,训练遗传神经网络映射坝体计算位移值与材料力学参数之间的复杂非线性关系。最后将实测位移值输入训练好的遗传神经网络,即可得到各参数的反演值。本文以清江隔河岩水电站重力拱坝为例,反演分析了坝体混凝土的弹性模量、线膨胀系数以及坝基主要岩体的弹性模量等参数。经检验、评价与对比验证,结果表明该方法可以大大地缩短反分析时间,提高反分析效率和准确性。

基于RBF神经网络的边坡稳定可靠度分析

RBF神经网络具有网络训练速度快、可以避免局部极小等优点。文章基于RBF神经网络理论,采用蒙特卡罗模拟法来进行结构的可靠度分析,研究了样本点的生成方法及样本数对可靠度分析结果的影响,并将基于RBF神经网络的蒙特卡罗模拟法应用于边坡的可靠度分析。计算表明:基于RBF神经网络的蒙特卡罗模拟法具有较好的计算效率和计算精度;样本点的生成方法和样本数对计算结果影响较大。与随机取样法相比,均匀设计法生成的样本点分布更均匀,由此样本点集训练生成的神经网络能更好的代替原功能函数,在相同的样本数时具有更高的计算精度;当计算精度相同时,均匀设计法比随机取样法需要生成的样本点少,计算效率高。

均匀设计法、神经网络和遗传算法结合在内高压成形工艺参数优化中的应用

内高压成形技术是以轻量化和一体化为特征的一种空心变截面轻体构件的先进制造技术。目前,内高压成形技术越来越受到人们的关注,特别是汽车制造企业。管材的内高压成形过程与很多因素有关,其中施加在管件内部的压力与轴向进给量之间的配比关系尤为重要,对两者的匹配关系进行优化是内高压成形面临的重要课题。传统的优化方法需要大量的模拟计算,耗时多且不易掌握。针对这一问题,该文提出了将均匀设计法、神经网络和遗传算法相结合进行参数优化,既利用了均匀设计试验的均匀可靠性,又运用神经网络的非线性映射、网络推理和预测功能,最后发挥遗传算法的全局优化特性,得出了最优结果,并直接为实际生产提供了可靠的参数依据。

基于SPSS和GIS的BP神经网络农用地适宜性评价

以中原城市群为研究区,在应用GIS基础上运用SPSS软件进行主成分分析,根据因子贡献率提取主导性强的综合因子,利用提取的主导因子进行系统聚类选择典型样本;依据因子的贡献率调整BP神经网络初始权值,建立评价模型,进行农用地适宜性评价方法的探讨。通过检验,改变初始权值建立的BP模型预测合格率为50%,相对误差最大值为18.9%;为了进一步改进BP网络,把第一次调整权值训练得到的网络权值作为下一次建模的初始权值,训练得到的BP模型预测合格率为100%,相对误差最大值为9.5%。其后运用层次分析法、主成分聚类法作为对照,验证了这种BP模型在农用地适宜性评价中的可靠性与精确性。最后根据所建立的BP模型预测获得中原城市群农用地适宜性分区图,各土地类型适宜性从西往东呈现由低到高的分带性,其中高度适宜区占总面积的32.4%;中度适宜、勉强适宜、不适宜区分别占总面积的28.9%、22.1%、16.6%。