CN118395364B - Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network - Google Patents
Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network Download PDFInfo
- Publication number
- CN118395364B CN118395364B CN202410865137.7A CN202410865137A CN118395364B CN 118395364 B CN118395364 B CN 118395364B CN 202410865137 A CN202410865137 A CN 202410865137A CN 118395364 B CN118395364 B CN 118395364B
- Authority
- CN
- China
- Prior art keywords
- data
- fault diagnosis
- discriminator
- distribution
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000012549 training Methods 0.000 claims abstract description 82
- 238000012545 processing Methods 0.000 claims abstract description 39
- 238000009826 distribution Methods 0.000 claims description 86
- 230000003595 spectral effect Effects 0.000 claims description 22
- 230000004913 activation Effects 0.000 claims description 19
- 238000013527 convolutional neural network Methods 0.000 claims description 16
- 230000007246 mechanism Effects 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 14
- 238000011176 pooling Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 10
- 238000013139 quantization Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000009827 uniform distribution Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 54
- 238000010586 diagram Methods 0.000 description 25
- 238000000354 decomposition reaction Methods 0.000 description 22
- 230000000717 retained effect Effects 0.000 description 12
- 230000000875 corresponding effect Effects 0.000 description 11
- 238000001228 spectrum Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000010354 integration Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000005096 rolling process Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 101100012902 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) FIG2 gene Proteins 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000001965 increasing effect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 101001121408 Homo sapiens L-amino-acid oxidase Proteins 0.000 description 1
- 102100026388 L-amino-acid oxidase Human genes 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000002040 relaxant effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0475—Generative networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/094—Adversarial learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Acoustics & Sound (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
本发明提供基于改进EEMD和生成对抗网络的旋转机械故障诊断方法及系统,属于旋转机械故障诊断技术领域,获取振动信号,通过改进EEMD算法将振动信号分解为多个本征模态分量,形成一维振动信号;将一维振动信号通过灰度处理转化为二维图像,并构建单通道二维图像训练数据集;建立旋转机械故障诊断模型,并设定目标函数;基于鉴别器损失及生成器损失训练生成对抗网络,并扩充故障数据,利用真实训练数据及生成训练数据训练分类器实现故障诊断。本发明结合改进EEMD和生成对抗网络进行数据不平衡情况下故障诊断,可以对旋转机械关键部件在复杂工况下所训练的诊断模型具有较好的稳定性、鲁棒性。
The present invention provides a rotating machinery fault diagnosis method and system based on improved EEMD and generative adversarial network, belonging to the technical field of rotating machinery fault diagnosis, obtaining a vibration signal, decomposing the vibration signal into multiple intrinsic mode components through an improved EEMD algorithm to form a one-dimensional vibration signal; converting the one-dimensional vibration signal into a two-dimensional image through grayscale processing, and constructing a single-channel two-dimensional image training data set; establishing a rotating machinery fault diagnosis model, and setting an objective function; training a generative adversarial network based on discriminator loss and generator loss, and expanding fault data, and using real training data and generated training data to train a classifier to achieve fault diagnosis. The present invention combines improved EEMD and generative adversarial network to perform fault diagnosis under data imbalance, and can have good stability and robustness for the diagnostic model trained for key components of rotating machinery under complex working conditions.
Description
技术领域Technical Field
本发明属于旋转机械故障诊断技术领域,尤其涉及基于改进EEMD和生成对抗网络的旋转机械故障诊断方法及系统。The present invention belongs to the technical field of rotating machinery fault diagnosis, and in particular to a rotating machinery fault diagnosis method and system based on improved EEMD and generative adversarial network.
背景技术Background Art
目前,旋转机械广泛应用于风力发电机、内燃机等大型传动机组,长期处于高负载、强冲击等极端环境下工作,其齿轮、轴承等旋转机械关键部件容易产生故障且齿轮、轴承引发的停机时间较长,因此,监控旋转机械的运行状态,尽早发现旋转机械内部潜在故障并及时维护,对旋转机械实际工业生产正常运维有重要意义。At present, rotating machinery is widely used in large transmission units such as wind turbines and internal combustion engines. It works under extreme environments such as high load and strong impact for a long time. Its key components such as gears and bearings are prone to failure and the downtime caused by gears and bearings is relatively long. Therefore, monitoring the operating status of rotating machinery, discovering potential internal faults of rotating machinery as early as possible and maintaining them in time are of great significance to the normal operation and maintenance of rotating machinery in actual industrial production.
旋转机械故障诊断涉及数学、计算机、机械等多学科知识,传统故障诊断流程包括信号采集、信号特征提取、诊断模型搭建、训练、测试、结果分析等步骤。伴随着人工智能的发展,基于机器学习算法可以将特征提取、诊断模型、训练、测试、结果融合成一体,形成了基于机器学习的旋转机械故障诊断方法。其中,受益于计算机技术以及GPU性能的快速发展,深度学习凭借其强大的特征提取能力以及出色的故障分类性能被广泛应用于旋转机械故障诊断中。但在实际工业场景中旋转机械大多运行于健康状态下,旋转机械齿轮、轴承等关键部件的健康数据、故障数据存在数据不平衡问题,难以训练出泛化性能好、故障诊断准确率高的故障诊断模型。生成对抗网络通过生成器和判别器的对抗博弈,能够生成高质量、逼真的数据,以其独特的对抗机制和强大的生成能力,在旋转机械数据不平衡下故障诊断已展现出强大性能,解决了深度学习在旋转机械不平衡数据集下诊断性能差的问题,放宽了深度学习诊断方法对于训练数据的要求,生成对抗网络大大提高了模型的泛化能力,因此基于生成对抗网络对旋转机械在不平衡数据下进行故障诊断十分重要。Rotating machinery fault diagnosis involves multidisciplinary knowledge such as mathematics, computer science, and mechanics. The traditional fault diagnosis process includes signal acquisition, signal feature extraction, diagnostic model construction, training, testing, and result analysis. With the development of artificial intelligence, feature extraction, diagnostic models, training, testing, and results can be integrated into one based on machine learning algorithms, forming a rotating machinery fault diagnosis method based on machine learning. Among them, benefiting from the rapid development of computer technology and GPU performance, deep learning has been widely used in rotating machinery fault diagnosis with its powerful feature extraction capabilities and excellent fault classification performance. However, in actual industrial scenarios, most rotating machinery operates in a healthy state. There is a data imbalance problem in the health data and fault data of key components such as rotating machinery gears and bearings, making it difficult to train a fault diagnosis model with good generalization performance and high fault diagnosis accuracy. Generative adversarial networks can generate high-quality and realistic data through adversarial games between generators and discriminators. With its unique adversarial mechanism and powerful generation capabilities, it has demonstrated strong performance in fault diagnosis under unbalanced rotating machinery data, solving the problem of poor diagnostic performance of deep learning under unbalanced rotating machinery data sets and relaxing the requirements of deep learning diagnostic methods for training data. Generative adversarial networks have greatly improved the generalization ability of the model. Therefore, it is very important to perform fault diagnosis on rotating machinery under unbalanced data based on generative adversarial networks.
基于生成对抗网络进行故障诊断虽然可以扩充数据量少的类别数据,解决数据不平衡的问题,但生成对抗网络容易出现模式崩溃的问题,需要对生成器、鉴别器、分类器进行协同训练保证诊断模型训练效果及诊断性能。且旋转机械齿轮、轴承等关键部件在复杂工况下易受噪音等激励因素干扰,其产生的振动信号通常具有非线性、非平稳、非周期的特点,难以从直接获取的振动信号中提取出关键的整体、局部故障特征,在振动信号整体、局部故障特征不明显的情况下作为训练集输入到生成器、鉴别器中,生成数据质量低、特征冗余复杂,且生成器、鉴别器的训练难度大,同时生成数据结合真实训练数据输入到分类器中,分类准确率低,无法训练出诊断准确率高的诊断模型。因此,实现旋转机械在复杂工况下提取振动信号的关键整体、局部故障特征,生成质量高、信息丰富、特征明显的振动信号,并基于该振动信号,对旋转机械故障进行诊断是当前亟待解决的技术问题。Although fault diagnosis based on generative adversarial networks can expand the data of categories with small amounts of data and solve the problem of data imbalance, generative adversarial networks are prone to mode collapse. It is necessary to train the generator, discriminator, and classifier in a coordinated manner to ensure the training effect and diagnostic performance of the diagnostic model. In addition, key components such as gears and bearings of rotating machinery are easily disturbed by excitation factors such as noise under complex working conditions. The vibration signals generated by them are usually nonlinear, non-stationary, and non-periodic. It is difficult to extract key overall and local fault features from the directly acquired vibration signals. When the overall and local fault features of the vibration signals are not obvious, they are input into the generator and discriminator as training sets. The generated data is of low quality, feature redundancy is complex, and the training of the generator and discriminator is difficult. At the same time, the generated data is combined with the real training data and input into the classifier. The classification accuracy is low, and it is impossible to train a diagnostic model with high diagnostic accuracy. Therefore, it is a technical problem that needs to be solved urgently to realize the extraction of key overall and local fault features of vibration signals of rotating machinery under complex working conditions, generate high-quality, information-rich, and feature-clear vibration signals, and diagnose rotating machinery faults based on the vibration signals.
发明内容Summary of the invention
本发明提供基于改进EEMD和生成对抗网络的旋转机械故障诊断方法,通过改进EEMD方法及直接合并策略形成合并分量,将合并分量转换为灰度图像构建训练数据集,结合数据集及生成对抗网络扩充故障诊断训练集,提高诊断模型的准确率及泛化能力。The present invention provides a rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network. A merged component is formed by improving the EEMD method and a direct merging strategy, and the merged component is converted into a grayscale image to construct a training data set. The data set and the generative adversarial network are combined to expand the fault diagnosis training set, thereby improving the accuracy and generalization ability of the diagnosis model.
方法包括:Methods include:
S101:获取旋转机械设备预设部件在预设工况下的振动信号,对获取的每个振动信号进行标签设置;S101: Acquire vibration signals of preset components of rotating mechanical equipment under preset working conditions, and set labels for each acquired vibration signal;
S102:通过改进EEMD算法将所述振动信号分解为多个本征模态分量,根据合并策略将本征模态分量进行合并,形成携带标签的一维振动信号;S102: decomposing the vibration signal into multiple intrinsic mode components by improving the EEMD algorithm, and merging the intrinsic mode components according to a merging strategy to form a one-dimensional vibration signal carrying a label;
S103:将一维振动信号通过灰度处理转化为二维图像,并构建单通道二维图像训练数据集;S103: converting the one-dimensional vibration signal into a two-dimensional image through grayscale processing, and constructing a single-channel two-dimensional image training data set;
S104:基于单通道二维图像训练数据集,结合卷积神经网络、注意力机制、生成器、鉴别器,分别建立第一旋转机械故障诊断模型和第二旋转机械故障诊断模型,并设定目标函数;S104: Based on the single-channel two-dimensional image training data set, a convolutional neural network, an attention mechanism, a generator, and a discriminator are combined to respectively establish a first rotating machinery fault diagnosis model and a second rotating machinery fault diagnosis model, and set an objective function;
S105:将单通道二维图像训练数据集输入到生成对抗网络中,基于鉴别器损失及生成器损失训练生成对抗网络,并扩充故障数据,利用真实训练数据及生成训练数据训练分类器实现故障诊断。S105: Input the single-channel two-dimensional image training data set into the generative adversarial network, train the generative adversarial network based on the discriminator loss and the generator loss, expand the fault data, and use the real training data and the generated training data to train the classifier to realize fault diagnosis.
进一步需要说明的是,步骤S102还包括:It should be further explained that step S102 also includes:
基于改进EEMD算法,向振动信号并入随机幅值高斯噪声,形成添加噪声的信号,基于信号极值分布特性判断并入噪声后信号极值分布特性是否满足改善极值分布特性判断条件;Based on the improved EEMD algorithm, the vibration signal Incorporate random amplitude Gaussian noise to form a noise-added signal , based on the signal extreme value distribution characteristics, the signal after incorporating noise is judged Whether the extreme value distribution characteristics meet the judgment conditions for improving the extreme value distribution characteristics;
若满足,则将信号配置为携带标签的一维振动信号;If satisfied, the signal A one-dimensional vibration signal configured to carry a tag;
若不满足,则将振动信号重新加入随机幅值高斯噪声直至信号极值分布特性,满足改善极值分布特性判断条件。If not satisfied, the vibration signal Re-add random amplitude Gaussian noise until the signal The extreme value distribution characteristics meet the judgment conditions for improving the extreme value distribution characteristics.
进一步需要说明的是,步骤S102中还通过公式(1)对满足判断条件的信号进行量化处理,公式(1)为:It should be further explained that in step S102, the signal satisfying the judgment condition is further evaluated by formula (1). After quantization, formula (1) is:
(1); (1);
其中,P表示极值分布特性评价值,、分别表示信号极值点x轴方向、y轴方向均匀分布及波动特性评价值,、分别表示平衡系数;Where P represents Extreme value distribution characteristic evaluation value, , They represent the uniform distribution of the signal extreme point in the x -axis direction and the y -axis direction and the fluctuation characteristic evaluation value, respectively. , They represent the balance coefficients respectively;
L 1和L 2分别表示x轴极大值序列和极小值序列的长度,和分别表示在x轴上的实际极值序列和理想极值序列,表示在x轴上的实际极大值序列,表示在x轴上的实际极小值序列,、分别表示x轴上的理想极大值、极小值序列,和分别表示的极大值和极小值,表示方差运算符,、分别表示极大值、极小值的数量,T表示的总长度,表示极值平均方差,表示极值点密度; L1 and L2 represent the length of the maximum and minimum value sequences on the x- axis , respectively. and Respectively The actual extreme value sequence and the ideal extreme value sequence on the x -axis, represents the actual maximum value sequence on the x- axis, represents the actual minimum sequence on the x- axis, , Respectively represent the ideal maximum and minimum sequences on the x- axis, and Respectively The maximum and minimum values of represents the variance operator, , Respectively represent the number of maximum and minimum values, T represents The total length of represents the extreme value mean variance, represents the density of extreme points;
公式(1)中,、用于评价添加噪声后的信号极值点x轴方向、y轴方向均匀分布及波动特性,、根据实际需要调整平衡系数形成极值分布特性评价值P。In formula (1), , Used to evaluate the signal after adding noise The extreme points are evenly distributed in the x -axis and y -axis directions and have fluctuation characteristics. , The balance coefficient is adjusted according to actual needs to form the extreme value distribution characteristic evaluation value P.
进一步需要说明的是,步骤S102还包括:It should be further explained that step S102 also includes:
当信号的极值分布特性评价值满足公式(2)的条件时,对信号进行分解;When the signal Evaluation value of extreme value distribution characteristics When the conditions of formula (2) are met, the signal To decompose;
(2); (2);
p表示所有满足改善极值分布特性判断条件的极值分布特性值P的序列,i表示添加噪声后信号个数; p represents the sequence of all extreme value distribution characteristic values P that meet the judgment condition of improving extreme value distribution characteristics, i represents the number of signals after adding noise;
将信号分解为若干IMF分量,通过相关系数、谱峭度对IMF分量进行量化,并设定量化阈值,筛选出相关系数、谱峭度量化值高于阈值的IMF分量,对所述IMF分量进行合并,形成合并后的振动信号。The signal Decompose into several IMF components, quantize the IMF components by correlation coefficient and spectral kurtosis, set a quantization threshold, screen out IMF components whose correlation coefficient and spectral kurtosis quantization values are higher than the threshold, merge the IMF components to form a merged vibration signal.
进一步需要说明的是,S104还包括:It should be further explained that S104 also includes:
基于公式(5)向生成器、鉴别器中嵌入Wasserstein距离,Wasserstein距离用于衡量生成数据和真实数据之间的差异;Based on formula (5), Wasserstein distance is embedded into the generator and discriminator. Wasserstein distance is used to measure the difference between generated data and real data.
(5); (5);
其中,表示和组合得到联合分布集合,表示所有可能联合分布中的其中一种分布,表示联合分布中采样的真实样本x和生成样本y之间的距离期望值,表示所有可能联合分布中样本距离期望值取到的下界。in, express and The combination is the joint distribution set, represents one of all possible joint distributions, represents the expected value of the distance between the real sample x and the generated sample y sampled in the joint distribution, It represents the lower bound of the expected value of the sample distance in all possible joint distributions.
进一步需要说明的是,S104中,基于固定生成器,训练鉴别器,配置第一旋转机械故障诊断模型为:It should be further explained that, in S104, based on the fixed generator, the discriminator is trained and the first rotating machinery fault diagnosis model is configured as:
; ;
基于固定鉴别器,训练生成器,配置第二旋转机械故障诊断模型为:Based on the fixed discriminator, the generator is trained and the second rotating machinery fault diagnosis model is configured as:
; ;
设定的目标函数为:The objective function is set as:
(6); (6);
其中,表示固定生成器时最大化鉴别器数值,表示固定鉴别器时最小化生成器数值,、分别表示数据服从真实数据分布、生成数据分布,、分别表示数据服从真实数据分布、生成数据分布的期望值,表示数据服从真实数据分布时鉴别器输出分布的期望值,表示生成数据的鉴别器输出值,表示生成数据的鉴别器输出分布期望值。in, represents the maximum discriminator value when the generator is fixed, represents the minimum value of the generator when the discriminator is fixed, , They respectively indicate that the data obeys the real data distribution and the generated data distribution. , They respectively represent the expected value of the data obeying the real data distribution and the generated data distribution, represents the expected value of the discriminator output distribution when the data follows the true data distribution, represents the discriminator output value of the generated data, represents the expected value of the discriminator output distribution for generated data.
进一步需要说明的是,步骤S104还包括:基于真实样本及生成的样本图像数据对分类器进行训练,通过多个卷积层、批量归一化层、池化层提取数据特征,使用全连接层将所有特征聚集起来,利用Softmax激活函数输出所归属类别标签概率,从而对类别标签进行划分,并定义样本图像数据与其对应的类别之间的映射关系。It should be further explained that step S104 also includes: training the classifier based on real samples and generated sample image data, extracting data features through multiple convolutional layers, batch normalization layers, and pooling layers, using a fully connected layer to aggregate all features, and using the Softmax activation function to output the probability of the belonging category label, thereby dividing the category label and defining the mapping relationship between the sample image data and its corresponding category.
进一步需要说明的是,步骤S104还包括:设定交叉熵为分类器处理多分类任务的目标函数如公式(8)所示,交叉熵用于衡量模型预测的概率分布与实际标签的概率分布之间的差异;It should be further explained that step S104 also includes: setting cross entropy as the objective function of the classifier for processing multi-classification tasks as shown in formula (8), and cross entropy is used to measure the difference between the probability distribution predicted by the model and the probability distribution of the actual label;
(8); (8);
其中,N是振动信号的数量,K是类别数量,i、k分别是振动信号、类别索引,是第i个振动信号的实际标签,是第i个振动信号属于第k类的概率,表示概率取对数,L表示交叉熵损失值;Where N is the number of vibration signals, K is the number of categories, i and k are vibration signal and category indexes respectively. is the actual label of the i -th vibration signal, is the probability that the i -th vibration signal belongs to the k-th class, represents the logarithm of probability, L represents the cross entropy loss value;
鉴别器还基于公式(6)建立如公式(9)的鉴别器损失函数,The discriminator also establishes the discriminator loss function as shown in formula (9) based on formula (6):
(9); (9);
其中,、分别表示数据服从真实数据分布、生成数据分布的期望值,表示输入数据后鉴别器输出,表示输入真实数据后鉴别器输出期望值,表示输入生成数据后鉴别器输出,表示输入生成数据后鉴别器输出期望值,表示鉴别器损失函数值;in, , They respectively represent the expected value of the data obeying the real data distribution and the generated data distribution, represents the discriminator output after input data, It indicates the expected value of the discriminator output after inputting real data. represents the output of the discriminator after inputting generated data, represents the expected value of the discriminator output after inputting the generated data, represents the discriminator loss function value;
生成器还基于公式(6)建立如公式(10)的生成器损失函数,The generator also establishes the generator loss function as shown in formula (10) based on formula (6):
(10); (10);
其中,分别表示数据服从生成数据分布的期望值,表示输入生成数据后鉴别器输出,表示输入生成数据后鉴别器输出期望值,表示生成器损失函数值。in, They represent the expected value of the data obeying the generated data distribution, represents the output of the discriminator after inputting generated data, represents the expected value of the discriminator output after inputting the generated data, Represents the generator loss function value.
进一步需要说明的是,设定如公式(11)的分类器损失函数,用于评估模型的预测分布和真实分布之间的距离,训练并调整分类器参数;It should be further explained that the classifier loss function as shown in formula (11) is set to evaluate the distance between the predicted distribution and the true distribution of the model, and to train and adjust the classifier parameters;
(11); (11);
其中,Loss表示交叉熵损失,y表示实际标签数据,表示模型预测标签数据。Among them, Loss represents the cross entropy loss, y represents the actual label data, Represents the model prediction label data.
本发明还提供基于改进EEMD和生成对抗网络的旋转机械故障诊断系统,系统包括:信号获取标注模块、合并处理模块、灰度处理模块、训练集构建模块以及训练故障诊断模块;The present invention also provides a rotating machinery fault diagnosis system based on improved EEMD and generative adversarial network, the system comprising: a signal acquisition and annotation module, a merging processing module, a grayscale processing module, a training set construction module and a training fault diagnosis module;
信号获取标注模块用于获取旋转机械预设部件在预设工况下的振动信号,对获取的所述振动信号进行标签标注;The signal acquisition and labeling module is used to acquire the vibration signal of the preset component of the rotating machinery under the preset working condition, and label the acquired vibration signal;
合并处理模块用于通过改进EEMD算法将所述振动信号分解为多个本征模态分量,根据合并策略将本征模态分量进行合并,形成携带标签的一维振动信号;The merging processing module is used to decompose the vibration signal into multiple intrinsic mode components by improving the EEMD algorithm, and merge the intrinsic mode components according to the merging strategy to form a one-dimensional vibration signal carrying a label;
灰度处理模块用于将一维振动信号通过灰度处理转化为二维图像,并构建单通道二维图像训练数据集;The grayscale processing module is used to convert the one-dimensional vibration signal into a two-dimensional image through grayscale processing and construct a single-channel two-dimensional image training data set;
训练集构建模块基于卷积神经网络、注意力机制构建生成器、鉴别器,建立旋转机械故障诊断模型,并设定目标函数;The training set construction module builds generators and discriminators based on convolutional neural networks and attention mechanisms, establishes a rotating machinery fault diagnosis model, and sets the objective function;
训练故障诊断模块用于将单通道二维图像训练数据集输入到生成对抗网络中,基于鉴别器损失及生成器损失训练生成对抗网络,并扩充故障数据,利用真实训练数据及生成训练数据训练分类器实现故障诊断。The training fault diagnosis module is used to input the single-channel two-dimensional image training data set into the generative adversarial network, train the generative adversarial network based on the discriminator loss and the generator loss, expand the fault data, and use the real training data and the generated training data to train the classifier to realize fault diagnosis.
从以上技术方案可以看出,本发明具有以下优点:It can be seen from the above technical solutions that the present invention has the following advantages:
本发明提供的基于改进EEMD和生成对抗网络的旋转机械故障诊断方法,通过为每个振动信号设置标签,可以确保后续的振动信号处理和分析具有明确的指向性,从而提高了故障诊断的准确性和效率。The rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network provided by the present invention can ensure that subsequent vibration signal processing and analysis have clear directionality by setting a label for each vibration signal, thereby improving the accuracy and efficiency of fault diagnosis.
本申请涉及的旋转机械故障诊断方法利用改进的EEMD算法,能够更有效地将复杂的振动信号分解为多个本征模态分量,这些分量携带了原始信号中不同频率和尺度的信息。还基于合并策略的应用进一步简化了信号结构,使得一维振动信号更加易于后续处理和分析。The rotating machinery fault diagnosis method involved in the present application utilizes an improved EEMD algorithm to more effectively decompose a complex vibration signal into multiple intrinsic mode components, which carry information of different frequencies and scales in the original signal. The signal structure is further simplified based on the application of a merging strategy, making the one-dimensional vibration signal easier to process and analyze later.
对于本申请的方法来讲将一维振动信号通过灰度处理转化为二维图像,可以充分利用图像处理技术来分析振动信号的特征,从而拓展了故障诊断的方法论。还有助于可视化信号中的异常模式,提高诊断的直观性和便捷性。For the method of this application, the one-dimensional vibration signal is converted into a two-dimensional image through grayscale processing, which can make full use of image processing technology to analyze the characteristics of the vibration signal, thereby expanding the methodology of fault diagnosis. It also helps to visualize abnormal patterns in the signal and improve the intuitiveness and convenience of diagnosis.
本申请结合卷积神经网络、注意力机制、生成器、鉴别器和分类器构建的旋转机械故障诊断模型,能够自动学习和提取振动信号中的深层次特征。而且相较于传统的故障诊断方法,具有更高的识别准确率和泛化能力,可以应对复杂多变的工况和故障模式。还通过生成对抗网络生成额外的故障数据,可以有效解决故障诊断中数据不足的问题。扩充后的数据集不仅提高了模型的训练效果,还增强了模型对未知故障模式的适应能力。The rotating machinery fault diagnosis model constructed by this application combines convolutional neural networks, attention mechanisms, generators, discriminators and classifiers, which can automatically learn and extract deep-level features in vibration signals. Moreover, compared with traditional fault diagnosis methods, it has higher recognition accuracy and generalization ability, and can cope with complex and changeable working conditions and fault modes. It also generates additional fault data through generative adversarial networks, which can effectively solve the problem of insufficient data in fault diagnosis. The expanded data set not only improves the training effect of the model, but also enhances the model's adaptability to unknown fault modes.
对于本方法来讲整个故障诊断过程自动化程度高,减少了人工干预和错误判断的可能性,从而提高了诊断的效率和准确性。For this method, the entire fault diagnosis process has a high degree of automation, which reduces the possibility of human intervention and misjudgment, thereby improving the efficiency and accuracy of diagnosis.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
为了更清楚地说明本发明的技术方案,下面将对描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the description will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
图1为基于改进EEMD和生成对抗网络的旋转机械故障诊断方法流程图;FIG1 is a flow chart of a rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network;
图2为改进EEMD分解和IMF分量合并方法流程图;FIG2 is a flow chart of the improved EEMD decomposition and IMF component merging method;
图3为A、B样本的spEEMD分解图;Figure 3 shows the spEEMD decomposition diagram of samples A and B;
图4为A、B样本保留分量示意图;Figure 4 is a schematic diagram of the retained components of samples A and B;
图5为A、B样本合并分量示意图;Figure 5 is a schematic diagram of the combined components of samples A and B;
图6为旋转机械故障诊断模型整体框架示意图;FIG6 is a schematic diagram of the overall framework of the rotating machinery fault diagnosis model;
图7为生成器结构图;Fig. 7 is a diagram of the generator structure;
图8为鉴别器结构图;Fig. 8 is a diagram of the discriminator structure;
图9为分类器结构图;Fig. 9 is a diagram of the classifier structure;
图10为模型训练过程中生成器、鉴别器、分类器损失曲线图;Figure 10 is a graph of the loss curves of the generator, discriminator, and classifier during model training;
图11为基于改进EEMD和生成对抗网络的旋转机械故障诊断混淆矩阵图;FIG11 is a confusion matrix diagram of rotating machinery fault diagnosis based on improved EEMD and generative adversarial network;
图12为单通道二维图像训练数据集示意图。FIG12 is a schematic diagram of a single-channel two-dimensional image training data set.
具体实施方式DETAILED DESCRIPTION
本申请提供的基于改进EEMD和生成对抗网络的旋转机械故障诊断方法中,为了说明而不是为了限定,提出了诸如特定系统结构、技术之类的具体细节,以便透彻理解本申请实施例。然而,本领域技术人员应当清楚,在没有这些具体细节的其它实施例中也可以实现本申请。In the rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network provided in the present application, specific details such as specific system structures and technologies are proposed for the purpose of explanation rather than limitation, so as to thoroughly understand the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details.
应当理解的是,当在本申请说明书和所附权利要求书中使用时,术语“包括”指示所描述特征、整体、步骤、操作、元素和/或组件的存在,但并不排除一个或多个其它特征、整体、步骤、操作、元素、组件和/或其集合的存在或添加。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”,除非是以其他方式另外特别强调。It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and/or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and/or their collections. The terms "comprising", "including", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.
本申请提及的“一个或多个”是指一个、两个或两个以上,本申请提及的“多个”是指两个或两个以上。在本申请的描述中,除非另有说明,“/”表示或的意思,比如,A/B可以表示A或B。本文中的“和/或”仅仅是一种描述关联对象的关联关系,表示可以存在三种关系,比如,A和/或B,可以表示:单独存在A,同时存在A和B,单独存在B这三种情况。The "one or more" mentioned in this application refers to one, two or more than two, and the "multiple" mentioned in this application refers to two or more. In the description of this application, unless otherwise specified, "/" means or, for example, A/B can mean A or B. The "and/or" in this article is just a way to describe the association relationship of associated objects, indicating that there can be three relationships. For example, A and/or B can mean: A exists alone, A and B exist at the same time, and B exists alone.
在本申请中描述的“一个实施例”或“一些实施例”等语句意味着在本申请的一个或多个实施例中包括该实施例描述的特定特征、结构或特点。由此,在本申请中的不同之处出现的“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等语句不是必然都参考相同的实施例,而是意味着“一个或多个但不是所有的实施例”,除非是以其他方式另外特别强调。The phrases such as "one embodiment" or "some embodiments" described in the present application mean that the specific features, structures or characteristics described in the embodiment are included in one or more embodiments of the present application. Therefore, the phrases such as "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments" etc. that appear in different places in the present application do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.
在本申请实施例中,“当……时”、“在……的情况下”、“若”以及“如果”等描述均指在某种客观情况下设备(如,下文所述的终端设备或者接入网设备)会做出相应的处理,并非是限定时间,且也不要求设备(如,下文所述的终端设备或者接入网设备)在实现时一定要有判断的动作,也不意味着存在其它限定。In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that under certain objective circumstances, the device (such as the terminal device or access network device described below) will make corresponding processing, which does not limit the time, and does not require the device (such as the terminal device or access network device described below) to have a judgment action when implementing it, nor does it mean that there are other limitations.
在本申请实施例中,“预先定义”可以是协议定义。其中,“预先定义”可以通过在设备(例如,包括终端设备和网络设备)中预先保存相应的旋转机械齿轮、轴承等关键部件的运行参数、代码、表格或其他可用于指示相关信息的方式来实现,本申请对于其具体的实现方式不做限定。In the embodiment of the present application, "pre-definition" may be a protocol definition. The "pre-definition" may be implemented by pre-saving the operating parameters, codes, tables or other methods that can be used to indicate relevant information of the corresponding key components such as rotating mechanical gears and bearings in the device (for example, including terminal devices and network devices). The present application does not limit the specific implementation method.
对于本申请提供的旋转机械故障诊断方法来讲,利用改进EEMD方法及分量合并策略获取特征丰富、信噪比高的合并分量,基于合并分量制作诊断模型训练集用于训练,降低诊断模型训练难度。本申请在旋转机械齿轮、轴承等关键部件在数据不平衡的情况下,通过注意力机制挖掘数据中的多类关键局部特征,保证生成振动信号存在关键诊断信息。For the rotating machinery fault diagnosis method provided in this application, the improved EEMD method and component merging strategy are used to obtain the merged components with rich features and high signal-to-noise ratio, and the diagnostic model training set is made based on the merged components for training, thereby reducing the difficulty of diagnostic model training. In the case of data imbalance of key components such as rotating machinery gears and bearings, this application mines multiple types of key local features in the data through the attention mechanism to ensure that the generated vibration signal contains key diagnostic information.
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
请参阅图1所示是一具体实施例中基于改进EEMD和生成对抗网络的旋转机械故障诊断方法的流程图,方法包括:Please refer to FIG. 1 which is a flowchart of a rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network in a specific embodiment. The method includes:
S101:获取旋转机械设备预设部件在预设工况下的振动信号,对获取的每个振动信号进行标签设置。S101: Acquire vibration signals of preset components of rotating mechanical equipment under preset working conditions, and set labels for each acquired vibration signal.
在一个示例性实施例中,旋转机械设备可以包括:汽轮机、燃气轮机、离心式压缩机、水轮机、汽轮机、柴油机、电动机、水泵以及如离心通风机、轴流通风机等。In an exemplary embodiment, the rotating mechanical equipment may include: a steam turbine, a gas turbine, a centrifugal compressor, a water turbine, a steam turbine, a diesel engine, an electric motor, a water pump, and centrifugal fans, axial flow fans, etc.
其中,旋转机械设备的预设部件可以涉及旋转机械齿轮、轴承等关键零部件。旋转机械设备会应用到一些复杂工况,这里的复杂工况包括:变转速工况,当旋转机械设备的转速发生变化时,设备的振动信号也会受到影响,可能导致信号故障特征微弱,难以提取。还有强噪声干扰环境,这容易导致设备在运行过程中可能会受到来自外部环境的强噪声干扰,如其他机械设备的运行噪声、空气流动噪声等,这些噪声会淹没设备的振动信号,导致信噪比降低,故障特征相对微弱。还涉及高温、高湿、高粘度、高腐蚀等环境:这些恶劣的环境条件会对旋转机械设备的运行产生影响,如导致设备部件磨损加剧、润滑不良等,进而影响设备的振动信号和性能。还有一些旋转机械设备的内部零部件在运行过程中可能会发生磨损、碰撞等相互作用,这些相互作用会产生复杂的振动信号,使得信号中的有效成分被大量冗余干扰覆盖,增加故障诊断的难度。Among them, the preset components of rotating machinery and equipment may involve key components such as rotating machinery gears and bearings. Rotating machinery and equipment will be applied to some complex working conditions, including: variable speed working conditions. When the speed of the rotating machinery and equipment changes, the vibration signal of the equipment will also be affected, which may cause the signal fault characteristics to be weak and difficult to extract. There is also a strong noise interference environment, which may easily cause the equipment to be subject to strong noise interference from the external environment during operation, such as the operating noise of other mechanical equipment, air flow noise, etc. These noises will drown out the vibration signal of the equipment, resulting in a decrease in the signal-to-noise ratio and relatively weak fault characteristics. It also involves high temperature, high humidity, high viscosity, high corrosion and other environments: these harsh environmental conditions will affect the operation of rotating machinery and equipment, such as causing increased wear of equipment parts and poor lubrication, thereby affecting the vibration signal and performance of the equipment. There are also some internal parts of rotating machinery and equipment that may interact with each other such as wear and collision during operation. These interactions will produce complex vibration signals, causing the effective components in the signal to be covered by a large amount of redundant interference, increasing the difficulty of fault diagnosis.
在复杂工况下多振动信号表现为信号特征不明显、信噪比低,本实施例考虑旋转机械中滚动轴承在不同类别、不同故障程度下的振动信号,对每种振动信号进行标签标注并整理形成初始数据集。Under complex working conditions, multiple vibration signals exhibit unclear signal characteristics and low signal-to-noise ratio. This embodiment considers the vibration signals of rolling bearings in rotating machinery under different categories and different fault degrees, labels each vibration signal and organizes it to form an initial data set.
为了能够说明本申请实施例的实现方式,本实施例使用美国凯斯西储大学(CWRU)开放的驱动端深沟球轴承SKF6205的数据集,其通过电动机驱动端轴承座上方的加速度传感器采集振动加速度信号,轴承分为健康(Normal)、外圈故障(ORF)、滚珠故障(BF)和内圈故障(IRF)四种类别,每种故障均有三种断层直径(0.1778mm、0.3556mm、0.5334mm)。健康及故障轴承分别装入测试电机中在四种转速条件下(1730rpm、1750rpm、1772rpm、1797rpm)记录加速度信号数据,采样频率为12000HZ。In order to illustrate the implementation of the embodiment of the present application, this embodiment uses the data set of the drive end deep groove ball bearing SKF6205 opened by Case Western Reserve University (CWRU) in the United States. The acceleration sensor above the bearing seat at the drive end of the motor collects the vibration acceleration signal. The bearings are divided into four categories: healthy (Normal), outer ring fault (ORF), ball fault (BF) and inner ring fault (IRF). Each fault has three fault diameters (0.1778mm, 0.3556mm, 0.5334mm). The healthy and faulty bearings are respectively installed in the test motor to record the acceleration signal data under four speed conditions (1730rpm, 1750rpm, 1772rpm, 1797rpm), and the sampling frequency is 12000HZ.
本实施例选取电机转速条件1730rpm的振动信号作为实验数据,选取故障直径0.1778mm下的内圈、滚动体、外圈故障,选取故障直径0.3556mm下的内圈、滚动体、外圈故障,同时添加健康类别数据,共计七类标签数据,以4096为数据长度基准截取每种类别300个样本,本实施例数据集如表1所示。In this embodiment, the vibration signal of the motor speed condition of 1730rpm is selected as the experimental data, the inner ring, rolling element, and outer ring faults with a fault diameter of 0.1778mm are selected, and the inner ring, rolling element, and outer ring faults with a fault diameter of 0.3556mm are selected. At the same time, healthy category data is added, totaling seven categories of label data, and 300 samples of each category are intercepted based on 4096 as the data length. The data set of this embodiment is shown in Table 1.
表1:滚动轴承振动信号数据集Table 1: Rolling bearing vibration signal dataset
S102:通过改进EEMD算法将所述振动信号分解为多个本征模态分量,根据合并策略将本征模态分量进行合并,形成携带标签的一维振动信号。S102: Decomposing the vibration signal into multiple intrinsic mode components by improving the EEMD algorithm, and merging the intrinsic mode components according to a merging strategy to form a one-dimensional vibration signal carrying a label.
本实施例涉及的集成经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)是经验模态分解(Empirical Mode Decomposition,EMD)的一种改进和扩展方法。The ensemble empirical mode decomposition (EEMD) involved in this embodiment is an improved and extended method of the empirical mode decomposition (EMD).
其中,EMD是一种处理非线性和非平稳信号的自适应方法,通过将信号分解成若干本征模态函数(Intrinsic Mode Functions,IMF)和一个剩余项来分析信号的不同频率成分,EEMD是在EMD方法基础上向输入的原始信号加入高斯噪声改善模态混叠问题的方法。而向原始信号中加入高斯噪声就涉及到噪声幅值A N和集成次数N这两个关键参数,噪声幅值和集成次数对EEMD分解结果的质量和性能有重要影响。面对复杂工况下旋转机械非线性、非平稳振动信号,基于经验及实验法则确定EEMD中噪声幅值和集成次数,存在计算成本高、可重复性差及分解结果偏差大等弊端。因此,本实施例将原始信号添加的特定幅值高斯噪声替换为随机幅值的高斯噪声,改进EEMD方法只需考虑集成次数一个参数,减少参数选择对分解效果的影响。Among them, EMD is an adaptive method for processing nonlinear and non-stationary signals. It analyzes the different frequency components of the signal by decomposing the signal into several intrinsic mode functions (IMF) and a residual term. EEMD is a method of adding Gaussian noise to the input original signal based on the EMD method to improve the modal aliasing problem. Adding Gaussian noise to the original signal involves two key parameters, the noise amplitude AN and the number of integrations N. The noise amplitude and the number of integrations have an important impact on the quality and performance of the EEMD decomposition results. In the face of nonlinear and non-stationary vibration signals of rotating machinery under complex working conditions, the noise amplitude and the number of integrations in EEMD are determined based on experience and experimental rules, which has the disadvantages of high computational cost, poor repeatability and large deviation of decomposition results. Therefore, in this embodiment, the specific amplitude Gaussian noise added to the original signal is replaced by Gaussian noise of random amplitude. The improved EEMD method only needs to consider one parameter, the number of integrations, to reduce the influence of parameter selection on the decomposition effect.
对于本实施例来讲,改进EEMD方法中特定幅值的高斯噪声替换为随机幅值的高斯噪声,向振动信号加入随机幅值高斯噪声,形成添加噪声的信号,基于信号极值分布特性评价,通过公式(1)对进行量化,并判断加入噪声后信号极值分布特性是否得到改善,若有改善则继续向下进行,若无改善则初始信号重新加入随机幅值高斯噪声直至极值分布特性得到改善。For this embodiment, the Gaussian noise of the specific amplitude in the improved EEMD method is replaced by Gaussian noise of random amplitude, and the vibration signal is Add random amplitude Gaussian noise to form a signal with added noise , based on the evaluation of the signal extreme value distribution characteristics , through formula (1) Quantify and determine whether the signal extreme value distribution characteristics are improved after adding noise. If so, Continue downward, if there is no improvement, the initial signal Re-add random amplitude Gaussian noise until The extreme value distribution characteristics are improved.
(1)。 (1).
需要说明的是,P表示极值分布特性评价值,、分别表示信号极值点x轴方向、y轴方向均匀分布及波动特性评价值,、表示平衡系数,和的具体数值可以根据旋转机械故障诊断实际使用环境进行设置及调整,这里不做限定;L 1和L 2分别表示x轴极大值序列和极小值序列的长度,和分别表示在x轴上的实际极值序列和理想极值序列,表示在x轴上的实际极大值序列,表示在x轴上的实际极小值序列,、分别表示x轴上的理想极大值、极小值序列,和分别表示的极大值和极小值,表示方差运算符,、分别表示极大值、极小值的数量,T表示的总长度,表示极值平均方差,表示极值点密度。It should be noted that P represents Extreme value distribution characteristic evaluation value, , They represent the uniform distribution of the signal extreme point in the x -axis direction and the y -axis direction and the fluctuation characteristic evaluation value, respectively. , represents the balance coefficient, and The specific value of can be set and adjusted according to the actual use environment of rotating machinery fault diagnosis, and is not limited here; L1 and L2 represent the length of the maximum value sequence and the minimum value sequence of the x- axis respectively, and Respectively The actual extreme value sequence and the ideal extreme value sequence on the x -axis, represents the actual maximum value sequence on the x- axis, represents the actual minimum sequence on the x- axis, , Respectively represent the ideal maximum and minimum sequences on the x- axis, and Respectively The maximum and minimum values of represents the variance operator, , Respectively represent the number of maximum and minimum values, T represents The total length of represents the extreme value mean variance, Represents the density of extreme points.
本实施例的公式(1)中,、用于评价添加噪声后的信号极值点x轴方向、y轴方向均匀分布及波动特性,、根据实际需要调整平衡系数形成极值分布特性评价值P。当的极值分布特性评价值P超过之前P的均值时,则该进入后续分解,判断条件表示为公式(2)。In the formula (1) of this embodiment, , Used to evaluate the signal after adding noise The extreme points are evenly distributed in the x -axis and y -axis directions and have fluctuation characteristics. , According to actual needs, adjust the balance coefficient to form the extreme value distribution characteristic evaluation value P. When the extreme value distribution characteristic evaluation value P exceeds the mean of the previous P , then the Entering the subsequent decomposition, the judgment condition is expressed as formula (2).
(2); (2);
其中,表示第个添加噪声后信号的极值分布特性评价值,p表示所有满足改善极值分布特性判断条件的极值分布特性值P的序列,其初始值为原始信号的极值分布特性值。in, Indicates The extreme value distribution characteristic evaluation value of the signal after adding noise, p represents the sequence of all extreme value distribution characteristic values P that meet the judgment conditions of improving the extreme value distribution characteristic, and its initial value is the original signal The extreme value distribution characteristic value of .
本实施例中,可以将获取的每个振动信号定义样本,并基于每个振动信号设置的标签对应设置到所对应的样本上。这样就将振动信号作为旋转机械故障诊断分析的样本。下述样本即为步骤S101获取的旋转机械设备预设部件在预设工况下的振动信号。In this embodiment, each vibration signal obtained can be defined as a sample, and the label set based on each vibration signal can be set to the corresponding sample. In this way, the vibration signal is used as a sample for rotating machinery fault diagnosis and analysis. The following sample is the vibration signal of the preset component of the rotating machinery equipment under the preset working condition obtained in step S101.
具体来讲,每个样本通过改进EEMD方法分解为若干IMF分量,通过相关系数、谱峭度对IMF分量进行量化,设定量化阈值,将相关系数、谱峭度量化值高于阈值的IMF分量挑选出来,此时挑选出的IMF分量故障特征明显、信息丰富,因此将挑选出的IMF分量合并,形成整体特征明显的振动信号。Specifically, each sample is decomposed into several IMF components through the improved EEMD method. The IMF components are quantified by correlation coefficient and spectral kurtosis. A quantization threshold is set, and the IMF components with correlation coefficient and spectral kurtosis quantization values higher than the threshold are selected. At this time, the selected IMF components have obvious fault characteristics and rich information. Therefore, the selected IMF components are merged to form a vibration signal with obvious overall characteristics.
本实施例的相关系数是衡量两个信号之间的线性相关性,其中,高的正相关系数表示信号在同一方向变化,而高的负相关系数表示信号在相反方向变化,通过公式(3)计算每个IMF分量和原始信号的相关系数。The correlation coefficient of this embodiment is used to measure the linear correlation between two signals, wherein a high positive correlation coefficient indicates that the signals change in the same direction, while a high negative correlation coefficient indicates that the signals change in opposite directions. The correlation coefficient of each IMF component and the original signal is calculated by formula (3).
(3); (3);
式中,表示IMF分量与初始信号的相关系数,n表示信号长度,x、y分别表示IMF分量、初始信号,xi、yi分别表示IMF分量第i个数据点、初始信号第i个数据点,、分别IMF分量数据均值、初始信号数据均值。In the formula, represents the correlation coefficient between the IMF component and the initial signal, n represents the signal length, x and y represent the IMF component and the initial signal respectively, xi and yi represent the i -th data point of the IMF component and the i -th data point of the initial signal respectively, , They are the mean of IMF component data and the mean of initial signal data respectively.
本实施例的相关系数在故障诊断中通常用于评估信号之间的关联性,衡量两个信号之间的线性关系,取值范围在-1到1之间,其中,1表示两个信号完全正相关,即一个信号增大时,另一个信号也随之增大。-1表示两个信号完全负相关,即一个信号增大时,另一个信号随之减小。0表示两个信号无线性相关性。根据经验所得,当相关系数的绝对值大于0.5时,表明信号之间有较强的关联性,因此将相关系数阈值设定为0.5。The correlation coefficient of this embodiment is usually used to evaluate the correlation between signals in fault diagnosis, and to measure the linear relationship between two signals. The value range is between -1 and 1, where 1 means that the two signals are completely positively correlated, that is, when one signal increases, the other signal also increases. -1 means that the two signals are completely negatively correlated, that is, when one signal increases, the other signal decreases. 0 means that the two signals have no linear correlation. According to experience, when the absolute value of the correlation coefficient is greater than 0.5, it indicates that there is a strong correlation between the signals, so the correlation coefficient threshold is set to Set to 0.5.
本实施例的谱峭度定义是通过计算频谱的四阶中心矩与二阶中心矩的比值衡量信号频谱的尖锐程度,式(4)为计算公式。The spectral kurtosis of this embodiment is defined as measuring the sharpness of the signal spectrum by calculating the ratio of the fourth-order central moment to the second-order central moment of the spectrum. Formula (4) is the calculation formula.
(4); (4);
式中,表示IMF分量的谱峭度值,表示IMF分量的频谱,N表示频谱频率成分的数量,表示IMF分量频谱在第i个频率的值,表示IMF分量频谱的均值。In the formula, represents the spectral kurtosis value of the IMF component, represents the spectrum of the IMF component, N represents the number of spectral frequency components, represents the value of the IMF component spectrum at the i -th frequency, Represents the mean of the IMF component spectrum.
谱峭度用于衡量信号频谱的尖锐程度,较高的谱峭度通常表明信号中存在异常或故障特征,根据经验所得,谱峭度大于3时,表明信号中存在尖锐、突出的频谱成分,常与冲击、故障特征相关,谱峭度小于或等于3时,表明信号频谱平滑,无明显尖锐成分,故障特征不明显。因此,将谱峭度阈值设定为3。Spectral kurtosis is used to measure the sharpness of the signal spectrum. A higher spectral kurtosis usually indicates that there are abnormal or fault features in the signal. According to experience, when the spectral kurtosis is greater than 3, it indicates that there are sharp and prominent spectral components in the signal, which are often related to shock and fault features. When the spectral kurtosis is less than or equal to 3, it indicates that the signal spectrum is smooth, there is no obvious sharp component, and the fault feature is not obvious. Therefore, the spectral kurtosis threshold is set to Set to 3.
本实施例通过改进EEMD方法对每个样本进行分解得到若干IMF分量,对每个样本的IMF分量通过相关系数、谱峭度进行指标量化,并根据相关系数阈值及谱峭度阈值选择后续进行合并的IMF分量,通过改进EEMD分解和IMF分量合并方法如图2所示。In this embodiment, each sample is decomposed into several IMF components by improving the EEMD method, and the IMF components of each sample are quantified by the correlation coefficient and the spectral kurtosis. The IMF components to be subsequently merged are selected according to the correlation coefficient threshold and the spectral kurtosis threshold. The improved EEMD decomposition and IMF component merging method is shown in FIG2 .
本实施例在凯斯西储大学数据集350个训练样本上进行验证,设置改进EEMD方法集成次数为30,这里、对应设置为2、5,部分样本改进EEMD分解效果如图3所示,通过改进EEMD方法将原始信号分解为多个IMF分量,通过相关系数及谱峭度对每个IMF分量进行量化,计算每个IMF分量同初始信号的相关系数及每个IMF分量的谱峭度,如表2所示。其中,图3中的各个附图的纵坐标均为振幅。图3中的(a1)为A样本IMF1分解图;图3中的(b1)为A样本IMF2分解图;图3中的(c1)为A样本IMF3分解图;图3中的(d1)为A样本IMF4分解图;图3中的(e1)为A样本IMF5分解图;图3中的(a2)为B样本IMF1分解图;图3中的(b2)为B样本IMF2分解图;图3中的(c2)为B样本IMF3分解图;图3中的(d2)为B样本IMF4分解图;图3中的(e2)为B样本IMF5分解图。若相关系数、谱峭度均达到指标阈值,则保留该IMF分量,若相关系数、谱峭度其中一个指标未达到指标阈值,则丢弃该IMF分量,最后将所有保留的IMF分量合并起来形成整体特征明显的信号。This example is verified on 350 training samples of the Case Western Reserve University dataset, and the number of integrations of the improved EEMD method is set to 30. , The corresponding settings are 2 and 5, and the improved EEMD decomposition effect of some samples is shown in Figure 3. The original signal is decomposed into multiple IMF components by the improved EEMD method, and each IMF component is quantified by the correlation coefficient and spectral kurtosis. The correlation coefficient of each IMF component with the initial signal and the spectral kurtosis of each IMF component are calculated, as shown in Table 2. Among them, the vertical coordinates of each figure in Figure 3 are amplitudes. (a1) in Figure 3 is the decomposition diagram of IMF1 of sample A; (b1) in Figure 3 is the decomposition diagram of IMF2 of sample A; (c1) in Figure 3 is the decomposition diagram of IMF3 of sample A; (d1) in Figure 3 is the decomposition diagram of IMF4 of sample A; (e1) in Figure 3 is the decomposition diagram of IMF5 of sample A; (a2) in Figure 3 is the decomposition diagram of IMF1 of sample B; (b2) in Figure 3 is the decomposition diagram of IMF2 of sample B; (c2) in Figure 3 is the decomposition diagram of IMF3 of sample B; (d2) in Figure 3 is the decomposition diagram of IMF4 of sample B; (e2) in Figure 3 is the decomposition diagram of IMF5 of sample B. If the correlation coefficient and spectral kurtosis both reach the index threshold, the IMF component is retained. If one of the correlation coefficient and spectral kurtosis does not reach the index threshold, the IMF component is discarded. Finally, all the retained IMF components are merged to form a signal with obvious overall characteristics.
根据表2数据,A样本IMF2、IMF3为保留的分量,B样本IMF2、IMF3为保留的分量,如图4所示。According to the data in Table 2, IMF2 and IMF3 of sample A are the retained components, and IMF2 and IMF3 of sample B are the retained components, as shown in Figure 4.
其中,图4中的(a1)为A样本IMF2保留的分量示意图,图4中的(b1)为A样本IMF3保留的分量示意图。图4中的(a2)为B样本IMF2保留的分量示意图,图4中的(b2)为B样本IMF3保留的分量示意图。Among them, (a1) in Figure 4 is a schematic diagram of the components retained by IMF2 of sample A, and (b1) in Figure 4 is a schematic diagram of the components retained by IMF3 of sample A. (a2) in Figure 4 is a schematic diagram of the components retained by IMF2 of sample B, and (b2) in Figure 4 is a schematic diagram of the components retained by IMF3 of sample B.
将A、B样本中保留的分量分别合并起来形成特征明显的信号用作后续模型训练,如图5所示。其中,图5中的(a)为A样本合并分量示意图。图5中的(b)B样本合并分量示意图。The components retained in samples A and B are merged to form a signal with obvious characteristics for subsequent model training, as shown in Figure 5. Figure 5 (a) is a schematic diagram of the merged components of sample A. Figure 5 (b) is a schematic diagram of the merged components of sample B.
表2: A、B样本每个IMF分量指标量化Table 2: Quantification of each IMF component indicator for samples A and B
S103:将一维振动信号通过灰度处理转化为二维图像,并构建单通道二维图像训练数据集。S103: Convert the one-dimensional vibration signal into a two-dimensional image through grayscale processing, and construct a single-channel two-dimensional image training data set.
本实施例,通过改进EEMD将原始信号进行分解,将分解形成的IMF分量通过相关系数及谱峭度进行量化,根据指标阈值选择保留下来的分量,把保留下来的分量形成合并分量,用于后续模型训练。所有初始样本经过处理均形成整体特征明显的合并分量,对所有样本合并分量进行灰度处理形成单通道二维图像样本训练集,并根据初始样本携带的标签信息对图像数据集进行标注,注明所对应的故障类型。In this embodiment, the original signal is decomposed by improving EEMD, the IMF components formed by the decomposition are quantified by correlation coefficient and spectral kurtosis, the retained components are selected according to the index threshold, and the retained components are formed into a merged component for subsequent model training. All initial samples are processed to form a merged component with obvious overall characteristics, and all sample merged components are gray-processed to form a single-channel two-dimensional image sample training set, and the image data set is annotated according to the label information carried by the initial sample to indicate the corresponding fault type.
示例性的讲,本实施例选取凯斯西储数据集七种标签类别,每种类别选取300个样本,对300个样本进行改进EEMD分解、合并,对合并分量进行灰度处理形成单通道二维图像,标注所对应故障类型,构建单通道二维图像训练数据集,如图12所示。Exemplarily speaking, this embodiment selects seven label categories of the Case Western Reserve data set, selects 300 samples for each category, performs improved EEMD decomposition and merging on the 300 samples, performs grayscale processing on the merged components to form a single-channel two-dimensional image, labels the corresponding fault types, and constructs a single-channel two-dimensional image training data set, as shown in Figure 12.
S104:基于单通道二维图像训练数据集,结合卷积神经网络、注意力机制、生成器、鉴别器,分别建立第一旋转机械故障诊断模型和第二旋转机械故障诊断模型,并设定目标函数。S104: Based on a single-channel two-dimensional image training data set, a convolutional neural network, an attention mechanism, a generator, and a discriminator are combined to respectively establish a first rotating machinery fault diagnosis model and a second rotating machinery fault diagnosis model, and set an objective function.
在一个示例性实施例中,卷积神经网络(Convolutional Neural Network,CNN)是一种专门用于处理具有类似网格结构的数据的深度学习模型。本实施例是对单通道二维图像训练数据集中的二维图像进行处理。In an exemplary embodiment, a convolutional neural network (CNN) is a deep learning model specifically used to process data with a grid-like structure. This embodiment processes a two-dimensional image in a single-channel two-dimensional image training data set.
其中,CNN通过使用卷积层、池化层和全连接层来提取特征和进行分类。主要用于对二维图像的图像识别、物体检测、语义分割等任务。CNN的关键组件涉及卷积层、激活函数、池化层、全连接层、归一化层,输入图像通过多个卷积层和池化层,逐步提取特征,提取的特征通过全连接层,进行分类或回归任务。使用反向传播算法训练模型,调整卷积核和全连接层的权重,从而进行模型训练。Among them, CNN uses convolutional layers, pooling layers, and fully connected layers to extract features and perform classification. It is mainly used for image recognition, object detection, semantic segmentation, and other tasks of two-dimensional images. The key components of CNN involve convolutional layers, activation functions, pooling layers, fully connected layers, and normalization layers. The input image passes through multiple convolutional layers and pooling layers to gradually extract features. The extracted features pass through the fully connected layer for classification or regression tasks. The model is trained using the back propagation algorithm to adjust the weights of the convolution kernel and the fully connected layer to perform model training.
本实施例中的注意力机制(Self-Attention)是深度学习模型中一种增强特定部分信息的技术,使模型能够选择性地关注输入的某些部分。最初应用于自然语言处理(NLP),特别是机器翻译任务,现在也被广泛应用于图像处理等领域。The attention mechanism (Self-Attention) in this embodiment is a technology in deep learning models that enhances specific parts of information, enabling the model to selectively focus on certain parts of the input. It was originally applied to natural language processing (NLP), especially machine translation tasks, and is now also widely used in fields such as image processing.
其中,多头注意力机制通过并行计算多个注意力机制,捕捉不同子空间中的信息,然后将结果进行拼接和变换,以获取更丰富的特征表示,通过计算注意力权重增强输入的关键部分信息,从而聚焦数据局部特征。Among them, the multi-head attention mechanism captures information in different subspaces by calculating multiple attention mechanisms in parallel, and then splices and transforms the results to obtain richer feature representations. It enhances the key part of the input information by calculating the attention weights, thereby focusing on the local features of the data.
根据本申请的实施例,基于卷积神经网络以及多头注意力机制构建生成器、鉴别器、分类器,旋转机械故障诊断模型整体框架如图6所示,具体为:生成噪声及真实数据对应标签输入到生成器中,通过生成器嵌入层、转置卷积层、注意力机制层生成图像扩充不平衡数据集,生成的图像和真实图像一起输入鉴别器中,通过增强鉴别器辨别真假的能力训练鉴别器。再基于固定鉴别器训练生成器,最终目的生成整体特征、局部特征明显的高质量图像,将生成的高质量图像及真实图像整理成分类器训练数据集,通过卷积神经网络及softmax函数实现故障诊断。According to the embodiments of the present application, a generator, a discriminator, and a classifier are constructed based on a convolutional neural network and a multi-head attention mechanism. The overall framework of the rotating machinery fault diagnosis model is shown in Figure 6, which is specifically as follows: noise and labels corresponding to real data are generated and input into the generator, and an image expansion unbalanced data set is generated through the generator embedding layer, transposed convolution layer, and attention mechanism layer. The generated image and the real image are input into the discriminator together, and the discriminator is trained by enhancing the discriminator's ability to distinguish true from false. The generator is then trained based on a fixed discriminator, and the ultimate goal is to generate high-quality images with obvious overall and local features, and the generated high-quality images and real images are organized into a classifier training data set, and fault diagnosis is achieved through a convolutional neural network and a softmax function.
作为一个示例,本申请中的生成器是基于噪声及真实数据对应标签生成图像,扩充不平衡类别的数据,生成器结构如图7所示。As an example, the generator in this application generates images based on noise and corresponding labels of real data, expands the data of unbalanced categories, and the generator structure is shown in Figure 7.
具体为:噪声向量及标签向量输入到Embedding嵌入层中转换为低维稠密向量,通过Flatten打平层展平,展平后的向量输入到FC全连接层形成的特征图,经ReLU激活函数后输入到通道数为128、卷积核尺寸为5、步长为2的转置卷积层Conv1,并通过BN批量归一化和ReLU激活函数进行处理,接着输入到通道数为64、卷积核尺寸为5、步长为2的转置卷积层Conv2中,并通过BN批量归一化和ReLU激活函数进行处理,再输入到通道数为32、卷积核尺寸为5、步长为2的转置卷积层Conv3中,通过BN批量归一化和ReLU激活函数处理后输入到Multi-head Attention多头注意力模块中增强局部特征,最后输入到通道数为1、卷积核尺寸为3、步长为2的转置卷积层Conv4中,经过Tanh激活函数处理后输出一个单通道灰度图像。Specifically, the noise vector and label vector are input into the Embedding layer to be converted into a low-dimensional dense vector, flattened by the Flatten layer, and the flattened vector is input into the FC fully connected layer to form The feature map is input into the transposed convolution layer Conv1 with 128 channels, 5 convolution kernel size and 2 stride after ReLU activation function, and then input into the transposed convolution layer Conv2 with 64 channels, 5 convolution kernel size and 2 stride, and processed by BN batch normalization and ReLU activation function, and then input into the transposed convolution layer Conv3 with 32 channels, 5 convolution kernel size and 2 stride, and then input into the Multi-head Attention module to enhance local features after BN batch normalization and ReLU activation function processing, and finally input into the transposed convolution layer Conv4 with 1 channel, 3 convolution kernel size and 2 stride, and output a single channel after Tanh activation function processing. Grayscale image.
本申请中的鉴别器是基于卷积神经网络构建,其结构如图8所示。The discriminator in this application is built based on a convolutional neural network, and its structure is shown in FIG8 .
具体为:真实样本图像数据及生成样本图像数据以的形式输入到通道数为32、卷积核大小为5、步长为2的二维卷积层Conv1,经BN批量归一化和Relu激活函数处理,通过步长为2的Max pooling1二维最大池化层池化,接着输入到通道数为64、卷积核大小为3、步长为2的二维卷积层Conv2,经BN批量归一化和Relu激活函数处理,通过步长为2的Maxpooling2二维最大池化层池化,再输入到通道数为128、卷积核大小为3、步长为2的二维卷积层Conv3中,经BN批量归一化和Relu激活函数处理,通过步长为2的Max pooling3二维最大池化层池化,最后输入到通道数为256、卷积核大小为3、步长为2的二维卷积层Conv4中,经BN批量归一化、Relu激活函数及步长为2的Max pooling4二维最大池化层,通过Flatten打平层展平后输入到输出为1的FC全连接层中,FC全连接层的输出即真实样本相关数值。Specifically: real sample image data and generated sample image data , and then input into the two-dimensional convolution layer Conv1 with a channel number of 32, a convolution kernel size of 5, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling1 two-dimensional maximum pooling layer with a stride of 2, and then input into the two-dimensional convolution layer Conv2 with a channel number of 64, a convolution kernel size of 3, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling2 two-dimensional maximum pooling layer with a stride of 2, and then input into the two-dimensional convolution layer Conv3 with a channel number of 128, a convolution kernel size of 3, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling3 two-dimensional maximum pooling layer with a stride of 2, and finally input into the two-dimensional convolution layer Conv4 with a channel number of 256, a convolution kernel size of 3, and a stride of 2. After BN batch normalization, Relu activation function and Max pooling2 with a stride of 2 The pooling4 two-dimensional maximum pooling layer is flattened by the Flatten layer and then input into the FC fully connected layer whose output is 1. The output of the FC fully connected layer is the relevant value of the real sample.
本实施例的分类器是基于卷积神经网络构建,通过足够的生成样本、真实样本数据训练,实现故障智能诊断,结构如图9所示。The classifier of this embodiment is constructed based on a convolutional neural network, and is trained with sufficient generated samples and real sample data to achieve intelligent fault diagnosis. The structure is shown in FIG9 .
具体为:真实样本图像数据及生成样本图像数据以的形式输入到通道数为32、卷积核大小为3、步长为2的二维卷积层Conv1,经BN批量归一化和Relu激活函数处理,通过步长为2的Max pooling1二维最大池化层池化,接着输入到通道数为64、卷积核大小为3、步长为2的二维卷积层Conv2,经BN批量归一化和Relu激活函数处理,通过步长为2的Max pooling2二维最大池化层池化,再输入到通道数为128、卷积核大小为3、步长为2的二维卷积层Conv3中,经BN批量归一化和Relu激活函数处理,通过步长为2的Max pooling3二维最大池化层池化,最后输入到通道数为256、卷积核大小为3、步长为2的二维卷积层Conv4中,经BN批量归一化、Relu激活函数及步长为2的Max pooling4二维最大池化层,通过Flatten打平层展平后,输入到输出为256的FC1全连接层中,经Relu激活函数处理后输入到输出为128的FC2全连接层中,输出经Relu激活函数处理后输入到输出为7的全连接层FC3中,FC3全连接层的输出通过Softmax函数处理,输出数据归属类别的概率,实现故障诊断。Specifically: real sample image data and generated sample image data , and then input into the two-dimensional convolution layer Conv1 with a channel number of 32, a convolution kernel size of 3, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling1 two-dimensional maximum pooling layer with a stride of 2, and then input into the two-dimensional convolution layer Conv2 with a channel number of 64, a convolution kernel size of 3, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling2 two-dimensional maximum pooling layer with a stride of 2, and then input into the two-dimensional convolution layer Conv3 with a channel number of 128, a convolution kernel size of 3, and a stride of 2. After BN batch normalization and Relu activation function processing, it is pooled through the Max pooling3 two-dimensional maximum pooling layer with a stride of 2, and finally input into the two-dimensional convolution layer Conv4 with a channel number of 256, a convolution kernel size of 3, and a stride of 2. After BN batch normalization, Relu activation function and Max pooling2 with a stride of 2 The pooling4 two-dimensional maximum pooling layer is flattened by the Flatten layer and then input into the FC1 fully connected layer with an output of 256. After being processed by the Relu activation function, it is input into the FC2 fully connected layer with an output of 128. The output is processed by the Relu activation function and then input into the FC3 fully connected layer with an output of 7. The output of the FC3 fully connected layer is processed by the Softmax function to output the probability of the data belonging to a category to achieve fault diagnosis.
本实施例基于单通道灰度图像数据训练旋转机械故障诊断模型,其中,生成器、鉴别器中嵌入Wasserstein距离,见公式(5),Wasserstein距离用于衡量生成数据和真实数据之间的差异。In this embodiment, a rotating machinery fault diagnosis model is trained based on single-channel grayscale image data, wherein the generator and the discriminator are embedded with Wasserstein distance, as shown in formula (5). The Wasserstein distance is used to measure the difference between the generated data and the real data.
(5); (5);
这里的表示和组合得到联合分布集合,表示所有可能联合分布中的其中一种分布,表示联合分布中采样的真实样本x和生成样本y之间的距离期望值,表示所有可能联合分布中样本距离期望值取到的下界。Here express and The combination is the joint distribution set, represents one of all possible joint distributions, represents the expected value of the distance between the real sample x and the generated sample y sampled in the joint distribution, It represents the lower bound of the expected value of the sample distance in all possible joint distributions.
进一步需要说明的是,本实施例生成器、鉴别器整体目标函数见公式(6),公式(6)从两个角度进行分析。It should be further explained that the overall objective function of the generator and discriminator of this embodiment is shown in formula (6). Formula (6) is analyzed from two perspectives.
一方面,固定生成器对鉴别器进行训练,得到第一旋转机械故障诊断模型为:On the one hand, the fixed generator is used to train the discriminator, and the first rotating machinery fault diagnosis model is obtained as follows:
。 .
另一方面,固定鉴别器对生成器训练,得到第二旋转机械故障诊断模型为:On the other hand, the fixed discriminator is used to train the generator, and the second rotating machinery fault diagnosis model is obtained as follows:
。 .
设定的目标函数为:The objective function is set as:
(6)。 (6).
其中,表示固定生成器时最大化鉴别器数值,表示固定鉴别器时最小化生成器数值,分别表示数据服从真实数据分布、生成数据分布,即x为真实数据、z为生成数据,、分别表示数据服从真实数据分布、生成数据分布的期望值,表示数据服从真实数据分布时鉴别器输出分布的期望值,表示生成数据的鉴别器输出值,表示生成数据的鉴别器输出分布期望值。in, represents the maximum discriminator value when the generator is fixed, represents the minimum value of the generator when the discriminator is fixed, They respectively indicate that the data obeys the real data distribution and the generated data distribution, that is, x is the real data and z is the generated data. , They respectively represent the expected value of the data obeying the real data distribution and the generated data distribution, represents the expected value of the discriminator output distribution when the data follows the true data distribution, represents the discriminator output value of the generated data, represents the expected value of the discriminator output distribution for generated data.
本实施例基于充足的真实样本及生成的样本图像数据对分类器进行训练,通过多个卷积层、批量归一化层、池化层提取数据特征,使用全连接层将所有特征聚集起来,This embodiment trains the classifier based on sufficient real samples and generated sample image data, extracts data features through multiple convolutional layers, batch normalization layers, and pooling layers, and uses a fully connected layer to aggregate all features.
利用Softmax激活函数输出所归属类别标签概率,从而对类别标签进行划分,并定义样本图像数据与其对应的类别之间的映射关系。Softmax激活函数用于多类别分类任务中,见公式(7),其将一组数值转换为表示概率分布的向量,任意实数向量转换为范围在(0,1)之间的数值,且所有输出的总和为1。The Softmax activation function is used to output the probability of the class label to which it belongs, thereby dividing the class labels and defining the mapping relationship between the sample image data and its corresponding class. The Softmax activation function is used in multi-class classification tasks, as shown in formula (7). It converts a set of numerical values into a vector representing the probability distribution. Any real number vector is converted into a numerical value in the range of (0,1), and the sum of all outputs is 1.
本实施例在分类问题中,Softmax函数通常用于将神经网络的输出转换为类别概率分布。In the classification problem of this embodiment, the Softmax function is usually used to convert the output of the neural network into a category probability distribution.
(7); (7);
其中,i表示元素编号,表示向量中的第i个元素,e为自然对数的底,表示向量第i个元素的指数值,C表示类别数量,c表示向量中元素索引,表示向量第c个元素的指数值,表示正确类别对应输出节点的概率值。Where i represents the element number, represents the i -th element in the vector, e is the base of the natural logarithm, Represents the i -th element of a vector The index value of C represents the number of categories, c represents the element index in the vector, Represents the cth element of a vector The index value of Indicates the probability value of the output node corresponding to the correct category.
本实施例设定交叉熵为分类器处理多分类任务的目标函数,交叉熵用于衡量模型预测的概率分布与实际标签的概率分布之间的差异,见公式(8):In this embodiment, cross entropy is set as the objective function of the classifier for processing multi-classification tasks. Cross entropy is used to measure the difference between the probability distribution predicted by the model and the probability distribution of the actual label, as shown in formula (8):
(8); (8);
其中,N是样本数量,K是类别数量,i、k分别是样本、类别索引,是第i个样本的实际标签,是第i个样本属于第k类的概率,通过公式(7)计算,表示概率取对数,L表示交叉熵损失值。Among them, N is the number of samples, K is the number of categories, i and k are sample and category indexes respectively. is the actual label of the i -th sample, is the probability that the i -th sample belongs to the k-th class, calculated by formula (7), represents the logarithm of probability, and L represents the cross entropy loss value.
S105:将单通道二维图像训练数据集输入到生成对抗网络中,基于鉴别器损失及生成器损失训练生成对抗网络,并扩充故障数据,利用真实训练数据及生成训练数据训练分类器实现故障诊断。S105: Input the single-channel two-dimensional image training data set into the generative adversarial network, train the generative adversarial network based on the discriminator loss and the generator loss, expand the fault data, and use the real training data and the generated training data to train the classifier to realize fault diagnosis.
本实施例中,将单通道二维图像训练数据集依次输入到生成器和鉴别器中,生成器和鉴别器通过各自损失函数交替训练,达到训练要求后,生成器可生成特征明显的高质量单通道二维图像数据,基于生成数据及真实数据,通过最小化交叉熵损失训练分类器,分类器训练完成后,旋转机械故障诊断模型可实现高准确率故障诊断。In this embodiment, a single-channel two-dimensional image training data set is input into the generator and the discriminator in turn. The generator and the discriminator are alternately trained through their respective loss functions. After meeting the training requirements, the generator can generate high-quality single-channel two-dimensional image data with obvious features. Based on the generated data and the real data, the classifier is trained by minimizing the cross entropy loss. After the classifier training is completed, the rotating machinery fault diagnosis model can achieve high-accuracy fault diagnosis.
需要说明的是,本实施例生成对抗网络中嵌入Wasserstein距离,使鉴别器还基于公式(6)建立如公式(9)的鉴别器损失函数,这里的鉴别器损失函数值为一个实数值,用来衡量输入数据的质量,即估计生成样本、真实样本之间Wasserstein距离。It should be noted that the Wasserstein distance is embedded in the generative adversarial network of this embodiment, so that the discriminator also establishes a discriminator loss function such as formula (9) based on formula (6). The discriminator loss function value here is a real value used to measure the quality of the input data, that is, to estimate the Wasserstein distance between the generated sample and the real sample.
(9); (9);
其中,分别表示数据服从真实数据分布、生成数据分布的期望值,表示输入数据后鉴别器输出,表示输入真实数据后鉴别器输出期望值,表示输入生成数据后鉴别器输出,表示输入生成数据后鉴别器输出期望值,表示鉴别器损失函数值。in, They respectively represent the expected value of the data obeying the real data distribution and the generated data distribution, represents the discriminator output after input data, It indicates the expected value of the discriminator output after inputting real data. represents the discriminator output after inputting generated data, represents the expected value of the discriminator output after inputting the generated data, Represents the discriminator loss function value.
本实施例中的生成器还基于公式(6)建立如公式(10)的生成器损失函数。考虑到生成器目标生成足够逼真的数据,使鉴别器对这些生成样本的评分尽可能高。生成器损失函数值表示鉴别器对生成数据的评分,通过最小化生成器损失函数值,即最大化鉴别器对生成数据的评分,使生成器生成特征明显、高质量的数据。The generator in this embodiment also establishes a generator loss function such as formula (10) based on formula (6). Considering that the generator aims to generate sufficiently realistic data, the discriminator scores these generated samples as high as possible. The generator loss function value represents the score of the discriminator on the generated data. By minimizing the generator loss function value, that is, maximizing the score of the discriminator on the generated data, the generator generates data with obvious features and high quality.
(10); (10);
其中,分别表示数据服从生成数据分布的期望值,表示输入生成数据后鉴别器输出,表示输入生成数据后鉴别器输出期望值,表示生成器损失函数值。in, They represent the expected value of the data obeying the generated data distribution, represents the discriminator output after inputting generated data, represents the expected value of the discriminator output after inputting the generated data, Represents the generator loss function value.
本实施例设定交叉熵损失为分类器损失函数,见公式(11)用于评估模型的预测分布和真实分布之间的距离,训练并调整分类器参数。In this embodiment, the cross entropy loss is set as the classifier loss function, as shown in Formula (11), which is used to evaluate the distance between the predicted distribution and the true distribution of the model, and train and adjust the classifier parameters.
(11); (11);
其中,Loss表示交叉熵损失,y表示实际标签数据,表示模型预测标签数据。Among them, Loss represents the cross entropy loss, y represents the actual label data, Represents the model prediction label data.
本实施例基于鉴别器损失函数、生成器损失函数、交叉熵损失函数训练旋转机械故障诊断模型,通过最大化鉴别器损失函数及最小化生成器损失函数形成对抗训练,通过最小化交叉熵损失训练分类器,鉴别器、生成器互相交替训练,分类器在真实数据、生成数据基础上训练,鉴别器损失用于更新鉴别器参数,生成器损失、分类器损失联合更新生成器参数,分类器损失更新分类器参数,当对抗训练达到平衡且分类损失值达到要求时,实现旋转机械在不平衡数据集下高准确率故障诊断。This embodiment trains a rotating machinery fault diagnosis model based on the discriminator loss function, the generator loss function, and the cross-entropy loss function. Adversarial training is formed by maximizing the discriminator loss function and minimizing the generator loss function. The classifier is trained by minimizing the cross-entropy loss. The discriminator and the generator are trained alternately. The classifier is trained based on real data and generated data. The discriminator loss is used to update the discriminator parameters. The generator loss and the classifier loss are jointly used to update the generator parameters. The classifier loss updates the classifier parameters. When the adversarial training reaches a balance and the classification loss value meets the requirements, high-accuracy fault diagnosis of rotating machinery is achieved under unbalanced data sets.
在上述实施例的基础上,为了进一步提高上述实施例提供旋转机械故障诊断方法的可靠性,为一种可实施的方式,在一实施例中,在凯斯西储大学数据集上进行训练,选择1730rpm下不同标签类别、不同故障程度数据共七种类别,考虑到不平衡数据下故障诊断模型,首先在七种类别数据中均选择50个样本共350个样本,在改进EEMD和合并策略形成整体特征明显数据的基础上,输入到生成器、鉴别器、分类器中,通过生成器损失、鉴别器损失、分类器损失训练并更新参数,获取高准确率旋转机械故障诊断模型。On the basis of the above embodiments, in order to further improve the reliability of the rotating machinery fault diagnosis method provided by the above embodiments, an implementable method is provided. In one embodiment, training is performed on the Case Western Reserve University data set, and seven categories of data with different label categories and different fault degrees at 1730rpm are selected. Considering the fault diagnosis model under unbalanced data, first, 50 samples are selected from each of the seven categories of data, totaling 350 samples. On the basis of improving EEMD and merging strategies to form overall feature-obvious data, the samples are input into the generator, discriminator, and classifier. The generator loss, discriminator loss, and classifier loss are used to train and update the parameters to obtain a high-accuracy rotating machinery fault diagnosis model.
本实施例中,旋转机械故障诊断模型训练过程中生成器、鉴别器、分类器损失如图10所示,随着迭代次数增加,分类器损失趋于最小并处于稳定,生成器、鉴别器损失趋于收敛并达到平衡,模型训练稳定,训练难度降低。本实施例的混淆矩阵如图11所示,标签1-7对应七种故障类别,当预测标签、真实标签同为同一类别时即分类正确,标签1-6均分类正确,准确率达100%,标签7存在误分类,少部分数据预测结果为标签6,故障诊断准确率达98.93%,所提方法具有较好的稳定性、鲁棒性,能够实现不平衡数据集下旋转机械故障识别、诊断。In this embodiment, the generator, discriminator, and classifier losses during the training of the rotating machinery fault diagnosis model are shown in Figure 10. As the number of iterations increases, the classifier loss tends to be minimum and stable, the generator and discriminator losses tend to converge and reach a balance, the model training is stable, and the training difficulty is reduced. The confusion matrix of this embodiment is shown in Figure 11. Labels 1-7 correspond to seven fault categories. When the predicted label and the true label are the same category, the classification is correct. Labels 1-6 are all correctly classified with an accuracy of 100%. Label 7 is misclassified, and a small part of the data is predicted to be label 6. The fault diagnosis accuracy rate is 98.93%. The proposed method has good stability and robustness, and can realize the identification and diagnosis of rotating machinery faults under unbalanced data sets.
基于上述方法本发明利用改进EEMD方法及分量合并策略获取特征丰富、信噪比高的合并分量,基于合并分量制作诊断模型训练集用于训练,降低诊断模型训练难度。本申请还在旋转机械齿轮、轴承等关键部件在数据不平衡的情况下,通过注意力机制挖掘数据中的多类关键局部特征,保证生成振动信号存在关键诊断信息。本申请的方法结合改进EEMD和生成对抗网络进行数据不平衡情况下故障诊断,旋转机械齿轮、轴承等关键部件在变转速、变负载等复杂工况下所训练的诊断模型具有较好的稳定性、鲁棒性。Based on the above method, the present invention uses an improved EEMD method and a component merging strategy to obtain a merged component with rich features and a high signal-to-noise ratio, and produces a diagnostic model training set based on the merged components for training, thereby reducing the difficulty of diagnostic model training. The present application also mines multiple types of key local features in the data through an attention mechanism for key components such as rotating machinery gears and bearings in the case of data imbalance, ensuring that the generated vibration signal contains key diagnostic information. The method of the present application combines improved EEMD and generative adversarial networks to perform fault diagnosis in the case of data imbalance, and the diagnostic models trained for key components such as rotating machinery gears and bearings under complex working conditions such as variable speeds and variable loads have good stability and robustness.
应理解,上述实施例中各步骤的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本发明实施例的实施过程构成任何限定。It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
以下是本公开实施例提供的基于改进EEMD和生成对抗网络的旋转机械故障诊断系统的实施例,该系统与上述各实施例的基于改进EEMD和生成对抗网络的旋转机械故障诊断方法属于同一个发明构思,在基于改进EEMD和生成对抗网络的旋转机械故障诊断系统的实施例中未详尽描述的细节内容,可以参考上述基于改进EEMD和生成对抗网络的旋转机械故障诊断方法的实施例。The following is an embodiment of a rotating machinery fault diagnosis system based on improved EEMD and generative adversarial networks provided in an embodiment of the present disclosure. This system and the rotating machinery fault diagnosis method based on improved EEMD and generative adversarial networks in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the rotating machinery fault diagnosis system based on improved EEMD and generative adversarial networks, please refer to the above-mentioned embodiment of the rotating machinery fault diagnosis method based on improved EEMD and generative adversarial networks.
系统包括:信号获取标注模块、合并处理模块、灰度处理模块、训练集构建模块以及训练故障诊断模块。The system includes: a signal acquisition and annotation module, a merging processing module, a grayscale processing module, a training set construction module and a training fault diagnosis module.
信号获取标注模块用于获取旋转机械预设部件在预设工况下的振动信号,对获取的所述振动信号进行标签标注。The signal acquisition and labeling module is used to acquire the vibration signal of the preset component of the rotating machinery under the preset working condition, and label the acquired vibration signal.
合并处理模块用于通过改进EEMD算法将所述振动信号分解为多个本征模态分量,根据合并策略将本征模态分量进行合并,形成携带标签的一维振动信号。The merging processing module is used to decompose the vibration signal into multiple intrinsic mode components by improving the EEMD algorithm, and merge the intrinsic mode components according to the merging strategy to form a one-dimensional vibration signal carrying a label.
灰度处理模块用于将一维振动信号通过灰度处理转化为二维图像,并构建单通道二维图像训练数据集。The grayscale processing module is used to convert the one-dimensional vibration signal into a two-dimensional image through grayscale processing and construct a single-channel two-dimensional image training data set.
训练集构建模块基于单通道二维图像训练数据集,结合卷积神经网络、注意力机制、生成器、鉴别器,分别建立第一旋转机械故障诊断模型和第二旋转机械故障诊断模型,并设定目标函数。The training set construction module is based on a single-channel two-dimensional image training dataset, combined with a convolutional neural network, an attention mechanism, a generator, and a discriminator, to establish the first rotating machinery fault diagnosis model and the second rotating machinery fault diagnosis model, and set the objective function.
训练故障诊断模块用于将单通道二维图像训练数据集输入到生成对抗网络中,基于鉴别器损失及生成器损失训练生成对抗网络,并扩充故障数据,利用真实训练数据及生成训练数据训练分类器实现故障诊断。The training fault diagnosis module is used to input the single-channel two-dimensional image training data set into the generative adversarial network, train the generative adversarial network based on the discriminator loss and the generator loss, expand the fault data, and use the real training data and the generated training data to train the classifier to realize fault diagnosis.
本实施例通过为每个振动信号设置标签,可以确保后续的振动信号处理和分析具有明确的指向性,从而提高了故障诊断的准确性和效率。This embodiment sets a label for each vibration signal, thereby ensuring that subsequent vibration signal processing and analysis have clear directionality, thereby improving the accuracy and efficiency of fault diagnosis.
本申请涉及的旋转机械故障诊断方法利用改进的EEMD算法,能够更有效地将复杂的振动信号分解为多个本征模态分量,这些分量携带了原始信号中不同频率和尺度的信息。还基于合并策略的应用进一步简化了信号结构,使得一维振动信号更加易于后续处理和分析。The rotating machinery fault diagnosis method involved in the present application utilizes an improved EEMD algorithm to more effectively decompose a complex vibration signal into multiple intrinsic mode components, which carry information of different frequencies and scales in the original signal. The signal structure is further simplified based on the application of a merging strategy, making the one-dimensional vibration signal easier to process and analyze later.
对于本申请的方法来讲将一维振动信号通过灰度处理转化为二维图像,可以充分利用图像处理技术来分析振动信号的特征,从而拓展了故障诊断的方法论。还有助于可视化信号中的异常模式,提高诊断的直观性和便捷性。For the method of this application, the one-dimensional vibration signal is converted into a two-dimensional image through grayscale processing, which can make full use of image processing technology to analyze the characteristics of the vibration signal, thereby expanding the methodology of fault diagnosis. It also helps to visualize abnormal patterns in the signal and improve the intuitiveness and convenience of diagnosis.
本申请结合卷积神经网络、注意力机制、生成器、鉴别器和分类器构建的旋转机械故障诊断模型,能够自动学习和提取振动信号中的深层次特征。而且相较于传统的故障诊断方法,具有更高的识别准确率和泛化能力,可以应对复杂多变的工况和故障模式。还通过生成对抗网络生成额外的故障数据,可以有效解决故障诊断中数据不足的问题。扩充后的数据集不仅提高了模型的训练效果,还增强了模型对未知故障模式的适应能力。The rotating machinery fault diagnosis model constructed by this application combines convolutional neural networks, attention mechanisms, generators, discriminators and classifiers, which can automatically learn and extract deep-level features in vibration signals. Moreover, compared with traditional fault diagnosis methods, it has higher recognition accuracy and generalization ability, and can cope with complex and changeable working conditions and fault modes. It also generates additional fault data through generative adversarial networks, which can effectively solve the problem of insufficient data in fault diagnosis. The expanded data set not only improves the training effect of the model, but also enhances the model's adaptability to unknown fault modes.
对于本方法来讲整个故障诊断过程自动化程度高,减少了人工干预和错误判断的可能性,从而提高了诊断的效率和准确性。For this method, the entire fault diagnosis process has a high degree of automation, which reduces the possibility of human intervention and misjudgment, thereby improving the efficiency and accuracy of diagnosis.
本申请提供的基于改进EEMD和生成对抗网络的旋转机械故障诊断系统是结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、计算机软件或者二者的结合来实现,为了清楚地说明硬件和软件的可互换性,在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本发明的范围。The rotating machinery fault diagnosis system based on improved EEMD and generative adversarial network provided by the present application is a unit and algorithm step of each example described in combination with the embodiments disclosed herein, and can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
所属技术领域的技术人员能够理解,本申请提供的基于改进EEMD和生成对抗网络的旋转机械故障诊断方法的各个方面可以实现为系统、方法或程序产品。因此,本公开的各个方面可以具体实现为以下形式,即:完全的硬件实施方式、完全的软件实施方式(包括固件、微代码等),或硬件和软件方面结合的实施方式,这里可以统称为“电路”、“模块”或“系统”。Those skilled in the art will appreciate that various aspects of the rotating machinery fault diagnosis method based on improved EEMD and generative adversarial network provided in the present application can be implemented as a system, method or program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, which can be collectively referred to as "circuit", "module" or "system" here.
对所公开的实施例的上述说明,使本领域专业技术人员能够实现或使用本发明。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的,本文中所定义的一般原理可以在不脱离本发明的精神或范围的情况下,在其它实施例中实现。因此,本发明将不会被限制于本文所示的这些实施例,而是要符合与本文所公开的原理和新颖特点相一致的最宽的范围。The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims (3)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410865137.7A CN118395364B (en) | 2024-07-01 | 2024-07-01 | Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410865137.7A CN118395364B (en) | 2024-07-01 | 2024-07-01 | Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN118395364A CN118395364A (en) | 2024-07-26 |
CN118395364B true CN118395364B (en) | 2024-09-13 |
Family
ID=91985380
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410865137.7A Active CN118395364B (en) | 2024-07-01 | 2024-07-01 | Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN118395364B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114386452A (en) * | 2021-12-06 | 2022-04-22 | 西安交通大学 | Fault detection method of sun gear of nuclear power circulating water pump |
CN117200202A (en) * | 2023-09-07 | 2023-12-08 | 国网上海市电力公司 | A two-stage photovoltaic output identification method, equipment and storage medium |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102254103B (en) * | 2011-07-27 | 2013-08-28 | 西安交通大学 | Method for determining screening time in self-adaptive ensemble empirical mode decomposition (EEMD) |
CN110617966A (en) * | 2019-09-23 | 2019-12-27 | 江南大学 | Bearing fault diagnosis method based on semi-supervised generation countermeasure network |
CN111458142B (en) * | 2020-04-02 | 2022-08-23 | 苏州新传品智能科技有限公司 | Sliding bearing fault diagnosis method based on generation of countermeasure network and convolutional neural network |
CN114563186B (en) * | 2022-01-21 | 2023-01-06 | 苏州大学 | Rolling bearing unbalance fault diagnosis method and system |
CN116754230A (en) * | 2023-04-25 | 2023-09-15 | 山东省计算中心(国家超级计算济南中心) | Bearing abnormality detection and fault diagnosis method based on deep convolution generation countermeasure network |
CN117290771B (en) * | 2023-09-22 | 2025-03-18 | 南京航空航天大学 | Rotating machinery fault diagnosis method based on improved auxiliary classification generative adversarial network |
CN118094342B (en) * | 2024-04-23 | 2024-07-26 | 山东大学 | Unsupervised domain adaptive fault diagnosis method and system for rotating machinery |
-
2024
- 2024-07-01 CN CN202410865137.7A patent/CN118395364B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114386452A (en) * | 2021-12-06 | 2022-04-22 | 西安交通大学 | Fault detection method of sun gear of nuclear power circulating water pump |
CN117200202A (en) * | 2023-09-07 | 2023-12-08 | 国网上海市电力公司 | A two-stage photovoltaic output identification method, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN118395364A (en) | 2024-07-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Fault diagnosis of rotating machinery based on combination of deep belief network and one-dimensional convolutional neural network | |
CN113505655B (en) | Intelligent bearing fault diagnosis method for digital twin system | |
Meng et al. | Data segmentation and augmentation methods based on raw data using deep neural networks approach for rotating machinery fault diagnosis | |
Gao et al. | Bearing fault diagnosis based on adaptive convolutional neural network with Nesterov momentum | |
CN113486931B (en) | Rolling bearing enhanced diagnosis method based on PDA-WGANGP | |
CN111275024A (en) | Fault diagnosis method for rolling bearing of variable-working-condition fan driven by limited data containing noise label | |
CN115753103A (en) | Fault diagnosis method and system based on standard self-learning data enhancement | |
CN114169377A (en) | G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment | |
Chen et al. | Fast robust capsule network with dynamic pruning and multiscale mutual information maximization for compound-fault diagnosis | |
CN110297479A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit based on the fusion of convolutional neural networks information | |
Ugli et al. | Automatic optimization of one-dimensional CNN architecture for fault diagnosis of a hydraulic piston pump using genetic algorithm | |
CN118094342B (en) | Unsupervised domain adaptive fault diagnosis method and system for rotating machinery | |
CN115409052A (en) | Fault diagnosis method and system for wind generating set bearing under data imbalance | |
Liu et al. | Multiscale residual antinoise network via interpretable dynamic recalibration mechanism for rolling bearing fault diagnosis with few samples | |
CN114491823A (en) | Train bearing fault diagnosis method based on improved generation countermeasure network | |
CN117992863A (en) | Rotary machine fault diagnosis method based on interpretable stationary wavelet packet convolution network | |
CN118296426A (en) | CWDC-GAN-based small sample fault diagnosis and identification method and system | |
CN118687854A (en) | Rolling bearing fault diagnosis method in noisy environment based on multimodal information fusion | |
Lou et al. | Machinery fault diagnostic method based on numerical simulation driving partial transfer learning | |
Lu et al. | An imbalanced sample intelligent fault diagnosis method using data enhancement and improved broad learning system | |
CN118395364B (en) | Rotary machine fault diagnosis method and system based on improved EEMD and generation countermeasure network | |
She et al. | Multiscale permutation entropy gray image coding method and its application in bearing fault diagnosis | |
CN111783941B (en) | Mechanical equipment diagnosis classification method based on probability confidence convolutional neural network | |
CN111721534A (en) | A method and system for online assessment of rolling bearing health status | |
He et al. | Rolling bearing fault diagnosis based on meta-learning with few-shot samples |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |