基于分式聚合过程神经网络的发动机气路参数偏差值预测

为解决航空发动机气路参数偏差值时间序列中突变值难以预测的问题,基于有理式函数具有更好的非线性逼近能力的理论,提出一种分式非线性聚合过程神经网络模型.该网络结构在隐层中存在一个过程神经元对偶层,通过分式非线性空间聚合的方式,分别实现信号对神经元的激励和抑制作用.根据采样点离散化的特点,采用离散Walsh变换对的内积运算替代积分算子,在简化计算过程的同时消除了数据拟合中的精度损失.采用基于离散Walsh变换LM算法进行网络训练,将训练好的模型应用在气路参数偏差值时间序列预测中.从预测结果可以看出,该模型对存在突变值的时间序列预测具有更高的效率和灵敏性.     

基于BP神经网络的多环槽磁流变减振器动力学模型辨识

磁流变减振器的输入输出具有很强的非线性关系,通常在进行结构分析时,需要对结构进行简化或线性化处理,因此理论上计算的十分准确的控制量,在实际中并不能达到满意的控制效果。该文采用BP神经网络对所设计的减振器进行正模型和逆模型辩识,避免了对结构进行理论建模的复杂性与不精确性,达到了很好的辨识效果。     

基于人工神经网络的电火花加工工艺建模方法研究

电火花加工是一个受多参数影响的复杂随机过程,很难建立一个适当的机理模型。用神经网络技术以铜加工SKD-Ⅱ为例建立了电火花加工工艺模型,并依据工艺样本中各参数数据的不同特点采用了不同的预处理方法,测试表明效果较好,所建模型能精确地预测出给定条件下的加工工艺参数,反映了该机床的加工工艺规律。另外,采用石墨加工SKD-Ⅱ工艺数据验证了该方法的通用性及有效性。     

Modeling of abrasive flow rotary machining process by artificial neural network

Abrasive flow machining (AFM) is one of the non-traditional machining processes applicable to finishing, deburring, rounding of edges, and removing defective layers from workpiece surface. Abrasive material, used as a mixture of a polymer with abrasive material powder, has reciprocal motion on workpiece surface under pressure during the process. In the following study, a new method of AFM process called henceforth abrasive flow rotary machining (AFRM) will be proposed, in which by elimination of reciprocal motion of abrasive material and the mere use of its stirring and rotation of workpiece, the amount of used material would be optimized. Furthermore, AFRM is executable by simpler tools and machines. In order to investigate performance of the method, experimental tests were designed by the Taguchi method. Then, the tests were carried out and the influence of candidate effective parameters was determined and modeled by artificial neural network (ANN) method. To evaluate the ANN results, they were compared with reported results of AFM. An agreement between our ANN results on predictions of AFRM material removal value and surface roughness was observed with AFM data. The results showed through AFRM, in addition to saving of abrasive material, surface finish is achievable same as AFM’s.     

A neural network solution for LOM process performance

A neural network (NN) modeling approach is presented for the prediction of laminated object manufacturing (LOM) process performance. A NN was developed using experimental data which were conducted on a LOM 1015 machine according to the principles of Taguchi design of experiments (DoE) method. The process parameters considered in the experiment to investigate LOM process performance were nominal layer thickness (NLT), heater temperature (HT), platform retract (PR), heater speed (HS), laser speed (LS), feeder speed (FS), and platform speed (PS). LOM process performance is divided in dimensional errors in X and Y directions (Ex and Ey), actual layer thickness (ALT), average surface roughness of vertical supporting frame (VSF-R_a), and tensile strength in X direction (TSx). It was found that NN approach can be applied in an easy way on designed experiments and predictions can be achieved, fast and quite accurate. The developed NN is constrained by the experimental region in which the designed experiment is conducted. Thus, it is very important to select parameters’ levels as well as the limits of the experimental region and the structure of the orthogonal experiment. The above analysis is useful for LOM users when prediction of process performance is needed. This methodology could be easily applied to different materials and initial conditions for optimization of other Rapid Prototyping (RP) processes.     

基于BP神经网络的PID控制器参数寻优

详细介绍BP神经网络对PID控制器参数寻优控制算法，用高斯核函数作为节点激励函数对系统进行控制。试验表明系统操作方便．安全可靠．控制效果好。     

Prediction of incremental sheet forming process performance by using a neural network approach

Although the incremental sheet forming process results are widely investigated from the literature review, much more efforts are required to increase the industrial applicability of the process; first of all, the material failure for complex shapes need to be clarified. According to this aim, in this work, a preliminary analysis is carried out to detect the factors that deeply affect the process performance when a part having a changing transverse section has to be manufactured. Subsequently, a neural network approach is utilized to implement a ??ready to use?? procedure to predict failure in complex shapes.     

基于过程神经网络的航空发动机性能参数预测

针对传统方法难以对性能参数进行有效预测的问题,提出一种基于过程神经网络的性能参数预测方法。为解决反向传播学习算法收敛速度慢、易陷于局部极小点等问题,开发了一种基于正交基函数展开的Leven-berg-Marquardt学习算法。为提高过程神经网络的泛化能力,从提高训练样本的质量和规模入手,研究了实际测量数据的预处理方法,并提出一种基于样条函数拟合和相空间重构理论的训练样本集构造方法。最后,将该方法用于某型航空发动机性能参数的预测,获得了满意的结果。     

Determination of process parameters in stereolithography using neural network

For stereolithography process, accuracy of prototypes is related to laser power, scan speed, scan width, scan pattern, layer thickness, resin characteristics and etc. An accurate prototype is obtained by using appropriate process parameters. In order to determine these parameters, the stereolithography (SLA) machine using neural network was developed and efficiency of the developed SLA machine was compared with that of the traditional SLA. Optimum values for scan speed, hatching spacing and layer thickness improved the surface roughness and build time for the developed SLA.     

间歇过程的神经网络模型

间歇过程通常具有非线性，时变和易燃易爆的特点，用常规的建模方法建立起模型比较困难，本文针对间歇聚丙烯过程，利用前馈神经网络建立其数学模型。首先根据实际系统的输入输出建立网络的结构。再用经验数据对网络进行训练，并用未参加训练的数据对网络进行测试，测试的最大误差是0.03MPa，这一误差在要求的范围之内。     

小样本下自校正卷积神经网络的滚动轴承故障识别方法

针对实际工程中因故障样本数据稀少而导致模型识别准确率不高的问题,提出了一种基于自校正卷积神经网络(SC-CNN)的滚动轴承故障诊断模型,并将其应用于小样本条件下的故障识别研究。首先,为减少不同信号的数据分布差异,在每个卷积层后添加BN算法;其次,利用自校正卷积学习信号的多尺度特征,提高模型获取有用故障特征的能力;然后,引入通道自注意力机制,建立通道特征信息之间的相关性,用于突出故障特征并抑制数据过拟合;再将少量训练样本输入到模型中进行学习;最后,将各类不同条件下的故障信号输入到训练好的SC-CNN模型进行识别分类,并在两个数据集上进行实验验证。结果表明,所提模型在信噪比为-4 dB的强噪声环境下,识别准确率分别为98.64%和99.83%,在变工况条件下,识别准确率分别为94.37%和99.64%,验证了SC-CNN模型在小样本条件下具有较强的鲁棒性和泛化性能。     

An improved artificial neural network for laser welding parameter selection and prediction

In the laser welding production, the selection and prediction of welding parameters is essentially important to guarantee weld quality. Artificial neural networks (ANN), which perform a nonlinear mapping between inputs and outputs, are an alternative approach for developing welding parameter forecasting model. In this paper, in order to speed up the convergence and avoid local minimum of the conditional ANN, genetic algorithm simulated annealing (GASA) based on the random global optimization is inducted into the network training. By means of GASA method, weights and threshold of neural networks can be globally optimized with short training time. Meanwhile, the gray correlation model (GCM) is used as a pre-processing tool to simplify the original networks based on obtaining the main influence elements of network inputs. The GCM–GASA–ANN method combines the complementary features of three computational intelligence techniques and owns very good applicability. Through the simulation and analysis of an orthogonal experiment, the proposed method can be proved to have higher accuracy and to perform better than the traditional ANN to forecast the laser welding parameters.     

模糊神经网络在UV-LIGA工艺优化中的应用

将模糊神经网络理论和算法应用于负性光刻胶(SU-8)加工高分辨率和高深宽比微结构的工艺研究,在正交试验的基础上对网络进行训练,建立了光刻图形质量与前烘时间、前烘温度、曝光量、后烘时间之间的预测模型。该模型采用五层前向模糊神经网络,学习算法为梯度下降法。进行了实验,实验结果表明,前烘温度与前烘时间对光刻质量影响最大。对120~340 μm厚的光刻胶,前烘温度取95℃,前烘时间100 min时,图形的相对线宽差最小;超声搅拌能缩短显影时间,显著改善图形质量,试验结果与计算结果十分吻合。将模糊神经网络应用于UV－LIGA工艺中,能实现光刻加工微结构的工艺参数优化。     

基于人工神经网络的板料弯曲成形工艺优化及回弹分析

结合人工神经网络所表现出来的良好特性,利用正交试验获得的数据作为神经网络的训练样本,建立输入为弯曲工艺参数、输出为回弹量的神经网络模型,并通过样本检验了ANN模型的准确性,从而缩短设定工艺参数的时间,在工艺参数取值范围内,采用ANN模型代替CAE软件模拟试验,结合正交试验法,对工艺参数进一步优化.结果表明:将神经网络与正交试验、数值模拟三者结合用于板料弯曲成形参数优化,可以缩短优化工艺参数的时间.提高工艺设计效率,并能获得比单纯使用正交试验和数值模拟方法更为优化的结果.     

基于神经网络的杂草识别试验研究

以苗期玉米和冬小麦为例,对比土壤、小麦、玉米、玉米田间杂草及小麦田间杂草的两种绿色增强因子统计图,利用过绿特征(2G-R-B)作为参数,结合BP神经网络,运用计算机图像处理技术,采用像素位置直方图法,识别出了杂草,并确定出了杂草位置、面积.实验结果表明,该方法可以准确识别出杂草,误差正确识别率高;小麦和玉米田问杂草识别时间短,可以满足实时性要求.     

Modal parameter identification by adaptive parameter domain with multiple genetic algorithms

Journal of Mechanical Science and Technology - The identification of aero-engine dynamic parameters is fundamental to establishing accurate dynamic models, which has a great effect on the accuracy...     

基于信息优化的加工过程神经网络控制

为了提高加工过程生产率和保证加工精度,以加工过程的恒切削力控制作为研究对象,将信息论原理和神经网络智能控制理论应用于加工过程控制,以信息熵作为加工过程智能控制系统的性能测度能统一各级性能指标,将神经网络作为加工过程控制输入和系统输出的信息传输通道,确定了神经网络基于信息优化的目标函数,推导出了信息优化的三层BP神经网络学习算法,提出了恒力切削过程中基于信息优化的神经网络控制系统框架.通过加工过程的仿真实例证明,与传统自适应神经网络控制方法相比,基于信息优化的神经网络控制方法收敛精确,速度快,振荡小,系统超调量小,具有较好的综合性能.研究结果为信息理论应用于加工过程控制提供了有效途径.     

Optimization of high-pressure die-casting process parameters using artificial neural network

High-pressure die casting is a versatile process for producing engineered metal parts. There are many attributes involved which contribute to the complexity of the process. It is essential for the engineers to optimize the process parameters and improve the surface quality. However, the process parameters are interdependent and in conflict in a complicated way, and optimization of the combination of processes is time-consuming. In this work, an evaluation system for the surface defect of casting has been established to quantify surface defects, and artificial neural network was introduced to generalize the correlation between surface defects and die-casting parameters, such as mold temperature, pouring temperature, and injection velocity. It was found that the trained network has great forecast ability. Furthermore, the trained neural network was employed as an objective function to optimize the processes. The optimal parameters were employed, and the castings with acceptable surface quality were achieved.     

加工过程的复合递归神经网络建模

针对BP神经网络的缺陷,在对角递归网络结构的基础上,提出了一种复合递归神经网络.BP算法收敛速度慢、产生局部极小点的原因之一是该算法采用了均方误差准则,为克服BP算法的不足,采用了一种广义熵方误差准则.把基于广义熵方误差准则的复合递归神经网络应用于加工过程的建模.仿真试验结果表明,复合递归神经网络建模具有比BP神经网络更快的收敛速度和更好的逼近效果.     

两种神经网络在注塑产品工艺参数确定中的应用

汽车外饰件的塑料化趋势对注塑模成型质量提出了更高要求.为解决传统CAE方法需多次试验才能得到较优工艺的缺点,以一汽车观后镜为研究对象,建立了基于人工神经网络的从注塑工艺参数到注塑翘曲量的非线性映射关系,并对比了两种经典的前馈神经网络(BP网络和RBF网络)的学习能力,从而实现用神经网络模型代替CAE软件获得注塑翘曲量.研究结果表明,该方法能有效地缩短优化工艺参数的时间,提高了工艺设计效率.