基于WOA–BP神经网络的液滴铺展预测

目的 提高BP神经网络对电喷印过程中液滴铺展行为的预测能力。方法 提出一种鲸鱼优化算法(WOA)优化BP神经网络的液滴铺展预测模型。首先，采用相场方法建立电场作用下液滴铺展的数值模型，并通过实验验证仿真结果的准确性。然后，选取初始直径、撞击速度、接触角和电场强度作为神经网络的输入参数，将最大铺展直径作为神经网络的输出参数，利用鲸鱼优化算法优化神经网络中的初始权值和阈值，构建液滴铺展预测模型。最后，基于仿真结果对预测模型进行训练与测试，并将其与传统的BP神经网络模型进行对比分析。结果 相较于传统BP神经网络预测模型，WOA–BP神经网络预测模型的平均绝对误差、均方根误差分别降低了72.60%、77.60%，而平均绝对百分比误差则从15.029 3%减小为4.585 3%。结论 WOA–BP神经网络预测模型可以更好地预测液滴铺展，可为液滴铺展的预测提供新的方法。     

Robustness analysis of radial base function and multi-layered feed-forward neural network models

In this paper, two popular types of neural network models (radial base function (RBF) and multi-layered feed-forward (MLF) networks) trained by the generalized delta rule, are tested on their robustness to random errors in input space. A method is proposed to estimate the sensitivity of network outputs to the amplitude of random errors in the input space, sampled from known normal distributions. An additional parameter can be extracted to give a general indication about the bias on the network predictions. The modelling performances of MLF and RBF neural networks have been tested on a variety of simulated function approximation problems. Since the results of the proposed validation method strongly depend on the configuration of the networks and the data used, little can be said about robustness as an intrinsic quality of the neural network model. However, given a data set where ‘pure’ errors from input and output space are specified, the method can be applied to select a neural network model which optimally approximates the nonlinear relations between objects in input and output space. The proposed method has been applied to a nonlinear modelling problem from industrial chemical practice. Since MLF and RBF networks are based on different concepts from biological neural processes, a brief theoretical introduction is given.     

A calibration framework for discrete element model parameters using genetic algorithms

In this research, a universal framework for automated calibration of microscopic properties of modeled granular materials is proposed. The proposed framework aims at industrial scale applications, where optimization of the computational time step is important. It can be generally applied to all types of DEM simulation setups. It consists of three phases: data base generation, parameter optimization, and verification. In the first phase, DEM simulations are carried out on a multi-dimensional grid of sampled input parameter values to generate a database of macroscopic material responses. The database and experimental data are then used to interpolate the objective functions with respect to an arbitrary set of parameters. In the second phase, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is used to solve the calibration multi-objective optimization problem. In the third phase, the DEM simulations using the results of the calibrated input parameters are carried out to calculate the macroscopic responses that are then compared with experimental measurements for verification and validation.The proposed calibration framework has been successfully demonstrated by a case study with two-objective optimization for the model accuracy and the simulation time. Based on the concept of Pareto dominance, the trade-off between these two conflicting objectives becomes apparent. Through verification and validation steps, the approach has proven to be successful for accurate calibration of material parameters with the optimal simulation time.     

BP神经网络在扬声器异常音检测中的应用

提出一种采用人工神经网络判断扬声器是否存在异常音的方法。首先简单介绍了获取扬声器异常音曲线的方法和人工神经网络中的BP模型及其训练方法,并比较了基本BP算法和共轭梯度法两种训练方法的差异。再将所获得的异常音曲线作为人工神经网络的输入向量,将听音员的听测结果作为目标向量,并使用共轭梯度法进行网络的训练。最后通过已训练好的人工神经网络判断扬声器是否存在异常音。实验结果表明,该方法可替代传统的人工设置门限的方法,并可大幅降低扬声器异常音检测的虚警率。     

基于频率和阻尼分析的含分层损伤复合材料层合板的损伤诊断

提出了一种基于动力有限元分析和神经网络相结合的含分层损伤层合板的诊断方法。采用作者发展的含分层损伤层合板的动力有限元分析模型和方法,计算了具有不同分层长度损伤层合板的频率和模态阻尼值,以此建立样本库。应用反向传播BP神经网络训练和形成网络。典型含层间分层损伤层合板的仿真结果表明,采用对损伤变化较为灵敏的高阶模态阻尼作为网络的输入参数进行分层损伤诊断比常用的模态频率更为合理。本文中提出的是一种用于层合板的分层损伤诊断的有效和经济的方法。      

基于Rough集理论的模糊神经网络构造方法

提出了在模糊神经网络中使用Rough集理论进行网络结构设计的方法。由于Rough集理论有强大的数值分析能力,而模糊神经网络具有准确的逼近收敛能力和较高的精度,所以通过两者的结合,可以得到一种可理解性好、计算简单、收敛速度快的神经网络模型。这种网络构造方法的主要过程为：首先,利用Rough集理论对给定数据集进行规则获取;然后,根据这些规则构造模糊神经网络各层的神经元个数及相关参数初始值;最后,用BP算法迭代求出网络的各种参数,完成网络的设计。给出了一个二维非线性函数拟合的实例,进一步验证了方法的正确性。     

基于粗糙PSO-BP神经网络的冷链物流服务商选择研究

为帮助冷链食品生产企业快速选择最佳冷链物流服务商,在传统BP神经网络基础上融合粗糙集和粒子群算法,构建了粗糙PSO-BP神经网络模型。该模型利用粗糙集剔除原始数据中的冗余信息,使输入指标更加精简;采用粒子群算法代替梯度下降法对神经网络权重进行训练,使输出结果不易陷入局部极小值,增强网络泛化能力。通过算例验证该模型的有效性和可行性。结果表明,该模型在提高运算速度的同时,预测误差为BP神经网络模型的40.94%,预测结果更加准确可靠,为冷链食品生产企业快速选择最佳冷链物流服务商提供一种新的方法指导。     

The Prediction of Dynamic Coefficients for Tilting-Pad Journal Bearings Based on an AGA-BP Neural Network

To provide real-time dynamic coefficients of tilting-pad journal bearings( TPJBs) for the dynamic analysis of a rotor-bearing system accurately,an improved error back propagation( BP) neural network model is built in this paper.First,the samples are gained by solving the Reynolds equation with the finite differential method based on hydrodynamic lubrication theory.Secondly,the adaptive genetic algorithm( AGA) is applied to optimize the initial weights and thresholds of the BP neural network before training.Then,with a number of trial calculations,the optimum parameters for the neural network are obtained.Finally,an application case of the neural network is given as well as the results analysis.The results show that the AGA can efficiently prevent the training of the neural network from falling into a local minimum,and the AGA-BP neural network of dynamic coefficients for TPJBs built in this paper can meet the demand of engineering.     

BP神经网络在测试系统动态补偿中的应用

测试系统的非线性动态补偿是仪器技术的一个重要方面．采用BP神经网络对测试系统进行动态补偿．BP神经网络的结果决定于网络输入、隐层和输出节点．由于其非线性映射特性,BP神经网络完全能够反映测试系统的动态响应特性．采用了收敛速度较快的递推预报误差算法训练神经网络．试验结果表明,BP神经网络的特性完全能够满足测试系统的动态补偿要求．表明本文的方法是有效的．     

布沼坝露天矿爆破效果预测BP神经网络模型应用

针对布沼坝露天矿西帮抢险治理工程,进行了35次生产爆破试验,完成了爆破参数的初步优化。同时以孔距、排距、孔深和抵抗线等作为模型的输入因子,大块率、爆堆的前冲距离、爆堆的后冲距离和挖掘机的铲装速率等作为模型的输出因子,以现场爆破试验数据为训练样本,建立爆破效果预测的BP神经网络模型。通过对网络仿真结果和现场实测数据进行比较分析,表明BP神经网络模型能够比较准确地预测出布沼坝露天矿西帮治理工程的爆破效果。     

梁损伤小波包分析和神经网络识别

针对柔性悬臂梁裂缝损伤问题进行损伤位置和损伤程度的识别研究。首先用有限元法建立系统动力学模型。然后对系统的动力响应信号进行小波包分解,建立基于小波包能量谱的损伤指标。把损伤指标作为改进BP神经网络的输入特征参数,用分步识别方法进行损伤位置和损伤程度的识别。最后进行了数值仿真研究。仿真结果表明,利用小波包分析和改进的BP神经网络可以精确地识别出柔性梁的损伤位置和损伤程度。     

多维线性模型参数估计的神经网络方法

本文提出一种用于多维线性模型（ＡＲ，ＡＲＭＡ）参数估计的神经网络方法和相应的递归预测误差算法。本文在分析多输入、单输出，含一个隐含和多层神经网络的输入输出关系的基础上，提出了首先将理想输出Ｘｉ进行预畸变（Ｆ（Ｘｔ））作为神经网络的训练目标。当神经网络训练完毕后，网络的连接权就是待估计的线性模型参数。本文提出方法的优点是网络结构简单，估计结果准确。仿真模拟结果表明，本文所提出的神经网络方法估计多维线性模型参数是有效的。     

An intelligent parameter selection system for the direct metal laser sintering process

As one of the promising Rapid Prototyping (RP) processes, the Direct Metal Laser Sintering (DMLS) technique is capable of building prototype parts by depositing and melting metal powders layer by layer. Metal powder can be melted directly to build functional prototype tools. During fabrication, four important resulting properties of interest to the users are: the processing time, mechanical properties, geometric accuracy and surface roughness. By adjusting an identified set of process parameters, these properties can be properly controlled. The process parameters involve: the laser scan speed, laser power, hatch density, layer thickness and scan path. But the relationships between these parameters and their resulting properties are quite complicated. In many cases, the effects of different parameters on the resulting properties contradict one another. In this paper, an intelligent system to assist the RP user to choose the optimal parameter settings based on different user requirements is presented. For the accurate prediction of the resulting properties of the laser-sintered metal parts, a method based on the feed-forward neural network (NN) with backpropagation (BP) learning algorithm is described. Through experiments, some input–output data pairs have been identified. After continuous training by using the data pairs, this NN constructs a good mapping relationship between the process parameters and their resulting properties. The system developed can determine the most suitable parameter settings containing the process parameters and predict resulting properties from the database built based on different process requirements automatically. It is very useful to RP users for saving material cost and reducing processing time.     

基于BP神经网络的金属拉深件裂纹在线监测

论文运用设计的三层BP神经网络对采集到的10个声发射参数进行特征提取。通过对比不同隐含层神经元个数的BP神经网络的训练误差与训练次数,确定当隐含层神经元个数为13个时,BP神经网络的逼近效果较好,产生的网络误差最小。然后利用计算各声发射参数对表征裂纹信号灵敏度的大小,逐步删除各个声发射参数,降低模式识别时输入信号的维数。最后确定相对到达时间、幅度、能率、上升计数、持续时间和平均信号电平六个声发射参数能够有效地识别金属拉深件裂纹。本研究对于金属拉深件裂纹的在线监测具有理论和实际意义。     

Modeling of Magnetic Properties of Nanocrystalline La-doped Barium Hexaferrite

In this paper an artificial neural network (ANN) has been developed to compute the magnetization of the pure and La-doped barium ferrite powders synthesized in ammonium nitrate melt. The input parameters were: the Fe/Ba ratio, La content, sintering temperature, HCl washing and applied magnetic field. A total of 8284 input data set from currently measured 35 different samples with different Fe/Ba ratios, La contents and washed or not washed in HCl were available. These data were used in the training set for the multilayer perceptron (MLP) neural network trained by Levenberg–Marquardt learning algorithm. The hyperbolic tangent and sigmoid transfer functions were used in the hidden layer and output layer, respectively. The correlation coefficients for the magnetization were found to be 0.9999 after the network was trained.     

Development and validation of calibration methods for discrete element modelling

Discrete element method (DEM) is proving to be a reliable and increasingly used tool to study and predict the behaviour of granular materials. Numerous particle-scale mechanisms influence the bulk behaviour and flow of bulk materials. It is important that the relevant measurable input parameters for discrete element models be measured by laboratory equipment or determined by physical calibration experiments for rational results. This paper describes some of the bench-scale experiments that have been developed to calibrate the DEM simulations to reflect actual dynamic behaviour. Relevant parameters such as static and rolling coefficients of friction, coefficient of restitution and inter-particle cohesion forces from the presence of liquid bridges have been investigated to model the bulk behaviour of dry and moist granular materials. To validate the DEM models, the results have been checked against experimental slump tests and hopper discharge experiments to quantitatively compare the poured and drained angles of repose and solids mass flow rate. The calibration techniques presented have the capability to be scaled to model and fine tune DEM parameters of granular materials of varying length scales to obtain equivalent static and dynamic behaviour.     

A hybrid of back propagation neural network and genetic algorithm for optimization of injection molding process parameters

This paper presents a hybrid optimization method for optimizing the process parameters during plastic injection molding (PIM). This proposed method combines a back propagation (BP) neural network method with an intelligence global optimization algorithm, i.e. genetic algorithm (GA). A multi-objective optimization model is established to optimize the process parameters during PIM on the basis of the finite element simulation software Moldflow, Orthogonal experiment method, BP neural network as well as Genetic algorithm. Optimization goals and design variables (process parameters during PIM) are specified by the requirement of manufacture. A BP artificial neural network model is developed to obtain the mathematical relationship between the optimization goals and process parameters. Genetic algorithm is applied to optimize the process parameters that would result in optimal solution of the optimization goals. A case study of a plastic article is presented. Warpage as well as clamp force during PIM are investigated as the optimization objectives. Mold temperature, melt temperature, packing pressure, packing time and cooling time are considered to be the design variables. The case study demonstrates that the proposed optimization method can adjust the process parameters accurately and effectively to satisfy the demand of real manufacture.     

基于插值与神经网络的包装薄膜热封参数分析

目的研究塑料薄膜热封工艺中热封参数之间的非线性关系,建立一种可用于自动包装机热封过程的数学模型。方法通过实验采集样本数据,并用附加动量法训练BP神经网络,建立热封参数之间的非线性数学模型,最后通过神经网络预测热封时间,并采用插值算法建立目标热封强度下热封温度和热封时间之间的多项式数学模型。结果通过插值算法与神经网络的结合运用,较为精确地描述了热封温度和热封时间之间的数学关系,插值函数实现了神经网络模型的简化,两者误差较小。结论通过文中方法确定了包装材料热封参数之间的非线性关系,将其用于热封包装设备,可提高设备的智能化程度。     

基于BP神经网络的盲均衡器设计

为构建用于某语音信号传输系统盲均衡器的BP神经网络模型,编写了基于BP神经网络的盲均衡算法伪代码,计算了算法的时间复杂度,分析了BP神经网络输入层神经元个数、隐含层神经元个数和隐含层层数对盲均衡算法性能的影响,评估了基于Sigmoid的变步长算法、基于误差补偿的变步长算法和基于误差的变步长算法对基于BP神经网络的盲均衡器性能的改进效率,据此设计了一种含双隐层结构的BP神经网络盲均衡器,并对其性能进行了数值仿真分析,明确了其适用范围,为该语音信号传输系统设计提供了技术支撑。     

小波包能量谱和BP神经网络在波纹管压浆超声检测中的应用

针对小波分析在信号处理的局限性,将小波包分析和反向传播(Back Propagation,BP)神经网络相结合,提出一种基于小波包能量谱和BP神经网络的波纹管压浆超声检测方法。采用超声检测方法接收波纹管模型的回波信号,以小波包分解后各子频带的能量作为检测特征,当波纹管内部出现脱落时,检测特征会发生变化,最后将特征输入BP神经网络中进行分类识别。试验结果表明,该方法能够理想地实现波纹管内部缺陷的诊断,可为波纹管超声检测提供一定的技术支持。