共查询到18条相似文献,搜索用时 62 毫秒
1.
针对传统红外图像的机载电路板芯片故障诊断法诊断率低且无法诊断动态故障的问题,本文提出了一种基于红外温度数据的改进麻雀搜索算法优化BP神经网络(Improved sparrow search algorithm-Back propagation neural networks, ISSA-BPNN)机载电路板芯片故障诊断方法。首先,提取红外热像仪采集的电路板芯片温度数据,建立电路板芯片升温过程中静态、动态、统计特征的特征模型;然后,利用Sine混沌映射初始化麻雀种群分布,利用Levy飞行策略改进发现者种群位置更新公式,将改进后的麻雀搜索算法优化BP神经网络的权值参数;最后,将温度特征模型输入到ISSA-BP神经网络进行训练和测试,从而完成电路板芯片故障诊断。实验采用航电系统电源电路板进行可靠性分析,实验结果表明,该方法在电路板不同工况下综合故障诊断率达到97.84%。 相似文献
2.
为了提高电路芯片故障诊断准确率,超参数设置的效率以及特征提取效率,提出一种基于时间模式注意力机制(TPA)的改进算数优化算法(IAOA)优化双向长短期记忆网络(BiLSTM)的电路故障诊断方法。首先,利用IAOA搜寻BiLSTM的最优超参数组合,提高模型诊断精度;然后使用TPA提取重要特征并分配权重,改善模型特征提取能力;最后,将红外摄像仪采集的红外温度数据输入到最优诊断模型中,实现电路芯片故障诊断。实验采用0~30 V可调稳压电源电路进行验证。结果表明,该模型对电路芯片故障诊断准确率高达9827,可实现对电路芯片的高准确率故障诊断。 相似文献
3.
电路板红外温度序列包含了丰富的故障类别信息,充分利用其局部与全局特征可以提高电路板故障诊断的准确率。为此,文中提出了一种由特征提取网络(Features Extraction Network,FEN)与关系学习网络(Relationship Learning Network,RLN)并行构成的可综合利用温度序列局部特征及特征间关系的电路板故障诊断模型。其中,FEN基于多尺度膨胀卷积(Multi-scale Dilated CNN,MDCNN)残差结构搭建,可在不增加训练参数的前提下构建多层次感受野,学习温度序列不同范围的空间特征;RLN基于嵌入长短期记忆网络的注意力机制(Long Short-Term Memory hybridized with Attention,LSTMwAtt)结构搭建,通过控制温度序列信息传递来学习特征重要性并分配权重,挖掘不同位置特征间的相关性。实验结果显示,所提模型在两个自建电路板温度序列测试数据集上的诊断性能优于同类型的FCN、MFCN、LSTM和LSTM-FCN,故障诊断准确率分别达到91.15%和96.27%,可实现对电路板故障的高准确率诊断。 相似文献
4.
5.
针对雷达高分辨距离像(HRRP)目标识别问题,传统方法只考虑样本的包络信息而忽略了距离单元间的时序相关性,该文提出了一种基于注意力机制的双向自循环神经网络模型。该模型将时域的HRRP数据通过滑窗分为正反两个序列,并将其分别通过两个相互独立的GRU网络进行特征提取,然后将同时刻提取到的特征进行拼接,从而利用了距离像双向的时序信息。考虑到不同时刻的序列对目标分类的重要性不同,通过注意力机制自适应地对各时刻隐层特征赋予不同的权值,最后根据加权求和后的隐层特征进行目标的识别与分类。实测数据实验结果表明,该文所提方法可以有效完成高分辨距离像的目标识别问题,并且在数据发生一定的时序偏移情况下,仍然可以准确找到目标区域。 相似文献
6.
本文为了提高Android恶意软件的检测效率,利用GRU模型解决标准RNN中出现的梯度消失问题和处理上下文具有长期依赖关系问题的能力,提出了基于GRU模型的Android恶意软件检测方法。对原始数据做标准化处理,将原始的数据集变化为特定尺寸的特征向量,可以用作深度学习网络模型的输入。使用Drebin数据集进行对照试验,对实验中特征向量进行降维处理,在全连接层实现归一化处理,最后在softmax分类,GRU层作为门控机制来保存代码数据间的依靠关系。对照实验结果表明,GRU模型与机器学习中的SVM模型以及单一的LSTM、DCNN模型相比,训练时间更短,检测结果中准确率、召回率、精确率、F1值都是最高的。 相似文献
7.
锂离子电池性能在衰退过程中呈现非平稳性和非线性,寿命预测往往被再生容量所干扰,衰退趋势难以捕捉,进而影响寿命预测。针对该问题,以容量为特征,构建一种基于麻雀搜索算法优化变分模态分解和门控循环单元的锂离子电池寿命预测方法。首先,利用麻雀搜索算法优化了变分模态分解的分解层数和惩罚因子,再通过优化了的变分模态分解算法将锂电池容量分解为若干分量,最后引入门控循环单元预测所分解的若干分量,将若干预测结果进行集成。通过NASA电池数据集对所提方法进行验证,并与两种模型相比较,结果表明,该方法相较于另两种方法预测精度平均提升了60%。 相似文献
8.
为提高电力负荷预测的准确性,采用灰狼优化(Grey Wolf Optimizer,GWO)算法对门控循环单元(Gated Recurrent Unit Neural Network,GRU)神经网络进行优化,并进行短期电力负荷预测.首先预处理数据并量化影响因素,然后搭建基于GWO超参数优化的GRU神经网络模型,最后与其... 相似文献
9.
10.
新闻推荐是根据用户的阅读习惯,为其推送更符合需求的内容,然而现有的方法仍存在特征学习不足的问题.针对此问题,提出了一种基于多通道CNN-BiGRU与多特征融合方法,主要由以下四部分组成:(1)词嵌入层.在词向量中融入实体嵌入向量,弥补单独仅使用词向量的不足,完成多通道词向量的构建;(2)多通道CNN-BiGRU模型.此... 相似文献
11.
12.
13.
模拟电路的固有特点使其故障诊断较数字电路困难.相对于BP网络,RBF神经网络具有最佳逼近性能且收敛快、无局部极小,可引入解决上述困难.根据具体电路,定义故障,选定测试点,确定网络结构,用Pspice获得训练样本,经过训练得到RBF网络.网络的输入为从测试点得到的输入向量,输出为对应的故障.为了验证网络的泛化性能,对每种... 相似文献
14.
15.
A novel method based on a fault dictionary that uses entropy as a preprocessor to diagnose faulty behavior in switched current
(SI) circuit is presented in the paper. The proposed method uses a data acquisition board to extract the original signal form
the output terminals of the circuit-under-tests. These original data are fed to the preprocessors for feature extraction and
finds out the entropies of the signals which are a quantitative measure of the information contained in the signals. The proposed
method has the capability to detect and identify faulty transistors in SI circuit by analyzing its output signals with high
accuracy. Using entropy of signals to preprocess the circuit response drastically reduces the size of fault dictionary, minimizing
fault detect time and simplifying fault dictionary architecture. The result from our examples showed that entropies of the
signals fall on different range when the faulty transistors` Transconductance Gm value varying within their tolerances of 5 or 10%, thus we can identify the faulty transistors correctly when the response
do not overlap. The average accuracy of fault recognition achieved is more than 95% although there are some overlapping data
when tolerance is considered. The method can classify not only parametric faults but also catastrophic faults. It is applicable
to analog circuits as well as SI ones. A low-pass and a band-pass SI filter and a Clock feedthrough cancellation circuit have
been used as test beached to verify the effectiveness of the proposed method. A comparison of our work with Yuan et al. (IEEE
Trans Instrum Meas 59(3):586–595, 2010), which used entropy and kurtosis as preprocessors, reveals that our method requiring
one feature parameter reduces the computation and fault diagnosis time. 相似文献
16.
Analogue electronic circuit diagnosis based on ANNs 总被引:1,自引:0,他引:1
Feed-forward artificial neural networks (ANNs) have been applied to the diagnosis of nonlinear dynamic analogue electronic circuits. Using the simulation-before-test (SBT) approach, a fault dictionary was first created containing responses observed at all inputs and outputs of the circuit. The ANN was considered as an approximation algorithm to capture mapping enclosed within the fault dictionary and, in addition, as an algorithm for searching the fault dictionary in the diagnostic phase. In the example given DC and small signal frequency domain measurements were taken as these data are usually given in device’s data-sheets. A reduced set of data per fault (DC output values, the nominal gain and the 3 dB cut-off frequency, measured at one output terminal) was recorded. Soft (parametric) and catastrophic (shorts and opens) defects were introduced and diagnosed simultaneously and successfully. Large representative set of faults was considered, i.e., all possible catastrophic transistor faults and qualified representatives of soft transistor faults were diagnosed in an integrated circuit. The generalization property of the ANNs was exploited to handle noisy measurement signals. 相似文献
17.