基于智能复合结构的可靠性分布模式自动识别

 采用Vor onoi 向量对SOM 网络算法进行了改进, 提高了学习收敛速度。通过提取数据的统计特征,建立了可靠性分布模式自动识别样本。提出的智能自动识别模型分两层, 在SOM 网络层对概率分布模式进行自动聚类, 在支持向量机层对各聚类组进行分类学习和识别, 获得识别模型的双层记忆权值。最后采用模型对常用可靠性分布模式进行了自动识别研究。测试结果表明, 建立的可靠性分布模式自动识别模型是可行、有效的。     

基于卷积门控循环网络的滚动轴承故障诊断

针对许多基于深度学习的滚动轴承故障诊断方法在小样本数据集下诊断性能下降的问题,提出一种基于卷积门控循环神经网络的轴承故障诊断模型。该模型使用两层的卷积网络来从输入信号中提取特征,同时使用tanh函数作为激活函数,且池化层使用大池化核来进行重叠下采样。将所提取得到的高层特征连接到双向门控循环网络。合并循环网络正向和逆向的最后一个状态,并连接一层全连接层进行输出。选用凯斯西储大学的轴承故障数据集来验证模型在小样本数据集下的诊断性能,实验结果表明,相比于其他类型的模型,该模型在仅有20个训练样本的情况下依然保持97%的识别准确率。      

基于迁移学习的无人机对地面目标识别研究

无人机对地面目标的识别精度受到数据集少和目标小的影响。传统深度学习需要大量有标注的数据集,限制了在具有小样本下的无人机对地面目标识别领域的应用。将迁移学习的方法用于卷积神经网络VGG16,并修改VGG16网络最后的3个连接层;同时利用单样本数据增强法将UC Merced数据集扩大到原来的8倍,对其进行验证和对比分析。实验结果表明,基于迁移学习的VGG16网络对地面目标识别的准确度可达97.62%,相较于未使用迁移学习的VGG16网络模型,整体提高了23.53%。并且在相同训练参数的设置下,模型比SqueezeNet、AlexNet、Inceptionv3、MobileNet-v2以及EfficientNetb0模型验证精度提高了3.63%～17.38%,收敛速度最快,可基本满足对地面目标的识别。     

质量控制图智能识别中样本数据的预处理

神经网络在质量控制图的异常模式识别应用中,小参数样本之间的模式特征在强随机干扰下差别不明显,易造成误判。本课题采用了两种样本数据的预处理方法--模糊化处理和指数加权滑动平均(EWMA)处理,可明显改善小参数异常模式的被识别能力。     

基于迁移学习的小样本目标识别研究进展与展望

在空、天、海等复杂环境下的目标识别任务中，高质量的样本数据往往较少。特别是在干扰对抗环境下，某些特定领域的目标信息获取困难，可靠的标注数据较少。小样本问题对深度学习技术在目标识别任务中的应用提出了新的挑战。迁移学习为小样本不确定环境下的目标识别问题提供了新的研究思路。本文针对小样本目标问题，以机载雷达等空天传感器信息对海面目标识别为例，介绍了迁移学习的主要思路和方法，对迁移学习在海面目标识别问题中的应用现状进展进行了总结；分析和归纳了迁移学习在海面目标识别应用中的主要挑战。最后对可解释性及鲁棒性的海洋目标识别技术需求及未来发展方向进行了展望。     

小样本目标检测综述

小样本学习是指在样本数据不足或质量较低的情况下进行的深度学习训练和预测的方法.针对深度学习目标检测应用中可能会面对的样本数据不足的问题,分析了小样本目标检测的数学模型和误差来源,将适用于小样本目标检测的方法分成数据、模型和算法三个类别进行了归纳总结,简述了各个方案的缺点与不足,并枚举了近年来在小样本目标检测上的可行方法...     

基于稀疏自动编码器的发动机机载模型建模方法研究

为解决分段线性化机载模型精度不足的问题,提出并设计了基于稀疏自动编码器的大包线、具有10输入11输出的发动机机载自适应模型,该模型由稳态、动态两部分组合而成。首先基于一种新的相似准则进行建模所需样本数据的压缩,在保留主要信息的同时,大大降低了数据量及采样时间。用BP算法对简化后的样本数据进行了机载模型稳态部分的建模。针对机载模型动态部分所需样本数据量巨大、BP算法难以训练的问题,建立了基于稀疏自动编码器的动态机载模型。引入准稳态判断逻辑,在动态过程使用稀疏自动编码器的动态机载模型,在稳态过程使用基于BP算法的稳态机载模型。仿真结果表明,所建立的发动机机载模型具有优良的动稳态精度,且实时性好、存储量小,其中动态精度小于1%,稳态精度小于0.6%,一次模型计算时间不大于1ms,模型存储量不大于100kB。     

多尺度层级金字塔网络的无人机入侵检测方法

近年来,无人机的快速发展给众多领域带来便利,然而无人机入侵给机场安全带来了巨大的挑战。由于无人机目标小、背景复杂、飞行速度快等特点,现有的主流目标检测方法通常难以准确地识别出入侵的无人机,易产生误检漏检的现象。提出了多尺度层级金字塔网络的无人机入侵检测方法,同时利用特征融合模块赋予特征金字塔不同层级、不同尺度的图像语义信息,并通过网格删除和4-Mosaic数据增强技术,对小样本数据集进行扩充,有效地提高了模型的泛化性能。实验表明,方法较于目前最优的无人机检测方法性能提升了5.5%。     

可拓模式识别算法中经典域的确定方法

可拓模式识别算法是根据各个关联度的对比来识别最终的模式,其中经典域区间的确定有着重要的作用。传统的经典域确定方法是基于数理统计学的,不适用于小样本数据,数据不完全以及数据分布未知等复杂情况。引入Bootstrap算法,并和传统的方法对比,更精确地确定了经典域区间。     

基于支持向量回归机的数控机床 几何误差元素建模研究

针对数控机床几何误差元素建模时面临的误差样本数据少且呈非线性的问题,研究在小样本数据集非线性回归分析中具有独特优势的支持向量回归机,并基于此建立数控机床几何误差元素的预测模型。分析现有几何误差检测中常用的九线法所存在的测量选点难和计算累积误差等问题,提出增加每条测量线垂直方向直线度的测量和修正误差项计算模型的改进方法。以高斯径向基核函数为支持向量回归模型的核函数,运用交叉验证法,选取合适的模型参数,求解凸二次规划问题,进而建立几何误差元素的预测模型。以QLM27100–5X五轴龙门机床X轴为例,基于改进的九线法进行测量辨识得到几何误差样本数据,然后分别基于支持向量回归机和最小二乘法建立几何误差元素预测模型,对比两个模型的预测精度,结果显示,前者的预测均方差值MSE为0.0238,小于后者的0.072,验证了支持向量回归模型在小样本集下具有更高的预测精度。     

A fault diagnosis model based on weighted extension neural network for turbo-generator sets on small samples with noise

In data-driven fault diagnosis for turbo-generator sets, the fault samples are usually expensive to obtain, and inevitably with noise, which will both lead to an unsatisfying identification performance of diagnosis models. To address these issues, this paper proposes a fault diagnosis model for turbo-generator sets based on Weighted Extension Neural Network (W-ENN). W-ENN is a novel neural network which has three types of connection weights and an improved correlation function. The performance of the proposed model is validated against Extension Neural Network (ENN), Support Vector Machine (SVM), Relevance Vector Machine (RVM) and Extreme Learning Machine (ELM) based models. The results indicate that, on noisy small sample sets, the proposed model is superior to the other models in terms of higher identification accuracy with fewer samples and strong noise-tolerant ability. The findings of this study may serve as a powerful fault diagnosis model for turbo-generator sets on noisy small sample sets.     

基于变精度粗糙集的航空发动机故障诊断

针对航空发动机故障诊断的实际情况,在分析粗糙集理论缺陷的基础上,引入变精度粗糙集模型对专家给出的样本集进行处理,并将所得到的极小化决策算法用于历史故障样本集的分析,得到了高的识别率,从实践的角度证明了该方法的有效性以及小样本情况下所得决策算法的普适性.      

采用BP网络辨识航空发动机数学模型

运用ＢＰ网络和实测数据作为学习样本，对某型航空发动机的数学模型进行了辨识研究。辨识模型输出的结果与实测数据比较误差较小。这种方法收敛速度快、精度高，结果表明用ＢＰ网络辨识方法能够得到比较精确的发动机数学模型。     

An inverse design method for supercritical airfoil based on conditional generative models

Inverse design has long been an efficient and powerful design tool in the aircraft industry.In this paper, a novel inverse design method for supercritical airfoils is proposed based on generative models in deep learning. A Conditional Variational Auto Encoder(CVAE) and an integrated generative network CVAE-GAN that combines the CVAE with the Wasserstein Generative Adversarial Networks(WGAN), are conducted as generative models. They are used to generate target wall Mach distributions for the inve...     

Longitudinal parameter identification of a small unmanned aerial vehicle based on modified particle swarm optimization

This paper describes a longitudinal parameter identification procedure for a small unmanned aerial vehicle(UAV)through modified particle swam optimization(PSO).The procedure is demonstrated using a small UAV equipped with only an micro-electro-mechanical systems(MEMS)inertial measuring element and a global positioning system(GPS)receiver to provide test information.A small UAV longitudinal parameter mathematical model is derived and the modified method is proposed based on PSO with selective particle regeneration(SRPSO).Once modified PSO is applied to the mathematical model,the simulation results show that the mathematical model is correct,and aerodynamic parameters and coefficients of the propeller can be identified accurately.Results are compared with those of PSO and SRPSO and the comparison shows that the proposed method is more robust and faster than the other methods for the longitudinal parameter identification of the small UAV.Some parameter identification results are affected slightly by noise,but the identification results are very good overall.Eventually,experimental validation is employed to test the proposed method,which demonstrates the usefulness of this method.     

Correlated Clutter and Resultant Properties of Binary Signals

A simple Markov process model of binary, digitized radar clutter returns is assumed. Probability distributions for the number of hits in n observations are developed for small n with a binary parameter describing the process derived for Rayleigh distributed clutter. Tables of distributions are included, along with an example to show the effects of correlation on the false-alarm probabilities of a sliding-window detector.     

Scatter factor confidence interval estimate of least square maximum entropy quantile function for small samples

Classic maximum entropy quantile function method (CMEQFM) based on the probabil-ity weighted moments (PWMs) can accurately estimate the quantile function of random variable on small samples, but inaccurately on the very small samples. To overcome this weakness, least square maximum entropy quantile function method (LSMEQFM) and that with constraint condition (LSMEQFMCC) are proposed. To improve the confidence level of quantile function estimation, scatter factor method is combined with maximum entropy method to estimate the confidence inter-val of quantile function. From the comparisons of these methods about two common probability distributions and one engineering application, it is showed that CMEQFM can estimate the quan-tile function accurately on the small samples but inaccurately on the very small samples (10 sam-ples); LSMEQFM and LSMEQFMCC can be successfully applied to the very small samples;with consideration of the constraint condition on quantile function, LSMEQFMCC is more stable and computationally accurate than LSMEQFM; scatter factor confidence interval estimation method based on LSMEQFM or LSMEQFMCC has good estimation accuracy on the confidence interval of quantile function, and that based on LSMEQFMCC is the most stable and accurate method on the very small samples (10 samples).     

Bayesian gamma mixture model approach to radar target recognition

This paper develops a Bayesian gamma mixture model approach to automatic target recognition (ATR). The specific problem considered is the classification of radar range profiles (RRPs) of military ships. However, the approach developed is relevant to the generic discrimination problem. We model the radar returns (data measurements) from each target as a gamma mixture distribution. Several different motivations for the use of mixture models are put forward, with gamma components being chosen through a physical consideration of radar returns. Bayesian formalism is adopted and we obtain posterior distributions for the parameters of our mixture models. The distributions obtained are too complicated for direct analytical use in a classifier, so Markov chain Monte Carlo (MCMC) techniques are used to provide samples from the distributions. The classification results on the ship data compare favorably with those obtained from two previously published techniques, namely a self-organizing map and a maximum likelihood gamma mixture model classifier.     

试车数据驱动的涡轴发动机起动过程模型辨识

针对传统的采用解析法建立涡轴发动机起动过程模型复杂的问题,提出了一种基于变步长萤火虫算法优化的有外部输入的非线性自回归网络(CSFA-NARX)的涡轴发动机起动过程模型辨识方法。以涡轴发动机起动过程试车试验数据为数据样本,利用CSFA-NARX网络模型辨识得到涡轴发动机起动过程模型,并采用留一交叉验证方法对辨识模型的性能进行验证。结果表明:得到的辨识模型输出参数,如燃气发生器转速n_g、输出轴转速n_r和涡轮后温度T₄都较好地逼近了试车实测数据,各参数验证样本最大相对误差平均值分别为0.90%、1.51%、和2.01%;在相同训练与验证样本情况下,得到的辨识模型精度优于采用萤火虫算法优化的NARX网络(FA-NARX)、NARX网络和变步长萤火虫算法优化的BP网络(CSFA-BP)模型精度。