CN116007937A - Intelligent fault diagnosis method and device for mechanical equipment transmission part - Google Patents
Intelligent fault diagnosis method and device for mechanical equipment transmission part Download PDFInfo
- Publication number
- CN116007937A CN116007937A CN202211575101.2A CN202211575101A CN116007937A CN 116007937 A CN116007937 A CN 116007937A CN 202211575101 A CN202211575101 A CN 202211575101A CN 116007937 A CN116007937 A CN 116007937A
- Authority
- CN
- China
- Prior art keywords
- self
- mechanical equipment
- layer
- network
- fault diagnosis
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 74
- 230000005540 biological transmission Effects 0.000 title claims abstract description 66
- 238000003745 diagnosis Methods 0.000 title claims abstract description 64
- 238000012549 training Methods 0.000 claims abstract description 59
- 238000012545 processing Methods 0.000 claims abstract description 12
- 230000007246 mechanism Effects 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 24
- 239000013598 vector Substances 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 14
- 238000003860 storage Methods 0.000 claims description 10
- 238000010276 construction Methods 0.000 claims description 8
- 230000036541 health Effects 0.000 claims description 8
- 238000013519 translation Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000004931 aggregating effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 abstract description 5
- 239000000284 extract Substances 0.000 abstract description 4
- 230000015556 catabolic process Effects 0.000 abstract description 3
- 238000006731 degradation reaction Methods 0.000 abstract description 3
- 238000012423 maintenance Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 11
- 238000012360 testing method Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 5
- 239000000306 component Substances 0.000 description 5
- 239000002131 composite material Substances 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000005096 rolling process Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000003862 health status Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000009533 lab test Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
本发明涉及机械设备智能运维技术领域,尤其是指一种机械设备传动部件智能故障诊断方法及装置。本发明采集不同工况下的振动信号作为训练样本,避免了由于工况改变导致模型性能下降的问题,通过建立自监督预训练网络,充分利用了容易获取的无标签样本训练网络,使网络能够提取更有效的特征,减少对标签样本的依赖;另外,通过基于自注意力机制的编码器模型和解码器模型,提取更全面的全局特征,同时抑制冗余特征,增强有效特征,无需通过预处理对输入数据进行预增强,提高了诊断效率。
The invention relates to the technical field of intelligent operation and maintenance of mechanical equipment, in particular to a method and device for intelligent fault diagnosis of transmission parts of mechanical equipment. The present invention collects vibration signals under different working conditions as training samples, which avoids the problem of model performance degradation due to changes in working conditions. Extract more effective features and reduce the dependence on label samples; in addition, through the encoder model and decoder model based on the self-attention mechanism, more comprehensive global features are extracted, while redundant features are suppressed, and effective features are enhanced without pre-processing. The processing pre-enhances the input data, improving the diagnostic efficiency.
Description
技术领域technical field
本发明涉及机械设备智能运维技术领域,尤其是指一种机械设备传动部件智能故障诊断方法及装置。The invention relates to the technical field of intelligent operation and maintenance of mechanical equipment, in particular to a method and device for intelligent fault diagnosis of transmission parts of mechanical equipment.
背景技术Background technique
机械设备是工业生产中的最主要载体,对于保证产品质量、提高生产效率和创造经济效益发挥着重要的作用。机械设备传动部件是机械设备的核心部件,发挥着承担载荷和动力传递等关键作用。然而,机械设备传动部件通常运行在恶劣的工作环境下,并且需要长时间不间断地服役,一旦发生故障就会产生严重的后果。对机械设备传动部件及时的监测和诊断能够合理规划设备维护时间,增加经济效益,避免灾难性的后果。因此,开展机械设备传动部件的智能诊断方法研究对于提高其稳定性和可靠性具有重大的意义。Mechanical equipment is the most important carrier in industrial production, and plays an important role in ensuring product quality, improving production efficiency and creating economic benefits. Mechanical equipment transmission components are the core components of mechanical equipment, which play key roles in load bearing and power transmission. However, the transmission parts of mechanical equipment usually operate in harsh working environments and need to serve for a long time without interruption. Once a failure occurs, serious consequences will occur. Timely monitoring and diagnosis of mechanical equipment transmission components can reasonably plan equipment maintenance time, increase economic benefits, and avoid catastrophic consequences. Therefore, it is of great significance to carry out research on intelligent diagnosis methods for mechanical equipment transmission components to improve their stability and reliability.
近年来,随着工业大数据和人工智能技术的发展,深度学习被广泛应用于机械设备传动部件的智能故障诊断方法中。通过构建深度神经网络,深度置信网络,循环神经网络和卷积神经网络等多层深度结构,智能故障诊断模型能够对信号与其对应故障类型之间的复杂函数关系进行近似映射,从而得到准确的诊断结果。相比传统方法,基于深度学习的智能故障诊断方法具有不需要数据预处理,自动提取特征和能够处理大数据等优点,因此得到越来越广泛的研究。In recent years, with the development of industrial big data and artificial intelligence technology, deep learning has been widely used in intelligent fault diagnosis methods for mechanical equipment transmission components. By constructing multi-layer deep structures such as deep neural network, deep belief network, recurrent neural network and convolutional neural network, the intelligent fault diagnosis model can approximate the complex functional relationship between the signal and its corresponding fault type, so as to obtain accurate diagnosis result. Compared with traditional methods, intelligent fault diagnosis methods based on deep learning have the advantages of not requiring data preprocessing, automatically extracting features, and being able to handle large data, so they have been more and more widely studied.
尽管基于深度学习的智能诊断方法具有上述的优点,但其自身存在的一些不足阻碍着它在工程上的实际应用,一方面,基于深度学习的智能诊断方法在训练网络模型的过程中需要用到大量的标签样本,然而在实际工程中这是不现实的,因为判断故障的类型需要专家知识和工程经验,标注这些样本需要耗费大量的人力物力,面对新的诊断任务或新的数据集时需从头训练网络模型;另一方面,现有的方法大多由于提取高层次特征的能力不足,选择通过预处理的手段对特征预增强,这一步会降低整个诊断过程的效率,并且当其处理原始振动数据时诊断性能会下降。其次,现有的深度学习模型通常从局部提取特征,但无法判断其是冗余特征还是故障类型相关的有效特征。最后,现有方法通常设置训练数据和测试数据在同一工况下,当工况改变时模型的性能会不可避免下降,降低了方法的实用性。这些不足将导致现有的基于自监督学习的智能故障诊断方法学习到的特征不够全面,诊断模型的泛化能力和实用性较差。Although the intelligent diagnosis method based on deep learning has the above advantages, its own shortcomings hinder its practical application in engineering. On the one hand, the intelligent diagnosis method based on deep learning needs to be used in the process of training the network model A large number of labeled samples, however, is unrealistic in actual engineering, because judging the type of fault requires expert knowledge and engineering experience, labeling these samples requires a lot of manpower and material resources, when faced with new diagnostic tasks or new data sets The network model needs to be trained from scratch; on the other hand, most of the existing methods choose to pre-enhance the features by means of preprocessing due to their insufficient ability to extract high-level features. Diagnostic performance degrades when vibrating data. Second, existing deep learning models usually extract features from local parts, but cannot judge whether they are redundant features or effective features related to fault types. Finally, existing methods usually set the training data and test data under the same working conditions, and the performance of the model will inevitably decline when the working conditions change, which reduces the practicability of the method. These deficiencies will lead to the fact that the features learned by the existing intelligent fault diagnosis method based on self-supervised learning are not comprehensive enough, and the generalization ability and practicability of the diagnosis model are poor.
发明内容Contents of the invention
为此,本发明所要解决的技术问题在于克服现有技术中学习到的特征不够全面、泛化性能差,且高度依赖标签样本的问题。Therefore, the technical problem to be solved by the present invention is to overcome the problems that the learned features in the prior art are not comprehensive enough, have poor generalization performance, and are highly dependent on labeled samples.
为解决上述技术问题,本发明提供了一种机械设备传动部件智能故障诊断方法,包括:In order to solve the above technical problems, the present invention provides a method for intelligent fault diagnosis of transmission parts of mechanical equipment, including:
采集机械设备传动部件在不同工况下的振动信号,对所述振动信号按预设数据点长度进行多次截取,得到多个样本数据,按预设比例对所述多个样本数据进行划分,对其中一部分样本数据按故障类别进行标定,作为有标签数据集,将另一部分作为无标签数据集;Collect vibration signals of mechanical equipment transmission parts under different working conditions, intercept the vibration signals multiple times according to the length of preset data points, obtain multiple sample data, and divide the multiple sample data according to preset ratios, Part of the sample data is calibrated according to the fault category as a labeled data set, and the other part is used as an unlabeled data set;
通过构建随机掩码模块、基于自注意力机制的编码器模型和解码器模型建立自监督预训练网络,利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码和重构,所述重构包括编码过程和解码过程,计算重构前后信号间的重构损失,并以最小化重构损失作为目标函数对所述自监督预训练网络进行优化,更新自监督预训练网络的网络参数;Establish a self-supervised pre-training network by constructing a random mask module, an encoder model and a decoder model based on a self-attention mechanism, and use the self-supervised pre-training network to mask and reconstruct samples in the unlabeled data set , the reconstruction includes an encoding process and a decoding process, calculates the reconstruction loss between signals before and after reconstruction, and optimizes the self-supervised pre-training network with the minimum reconstruction loss as the objective function, and updates the self-supervised pre-training network network parameters;
通过构建所述编码器模型和分类器模型建立微调网络,并将优化好的自监督预训练网络中编码器模型对应的参数迁移至微调网络的编码器模型中,利用所述微调网络对所述有标签数据集中的样本进行分类,计算分类损失,并以最小化分类损失作为目标函数对所述微调网络进行优化,更新微调网络的网络参数,得到机械设备传动部件智能故障诊断模型;Establish a fine-tuning network by constructing the encoder model and classifier model, and migrate the parameters corresponding to the encoder model in the optimized self-supervised pre-training network to the encoder model of the fine-tuning network, and use the fine-tuning network to Classify the samples in the labeled data set, calculate the classification loss, and optimize the fine-tuning network with minimizing the classification loss as the objective function, update the network parameters of the fine-tuning network, and obtain an intelligent fault diagnosis model for mechanical equipment transmission parts;
将待测试的振动信号输入所述机械设备传动部件智能故障诊断模型中,得到机械设备传动部件的故障类别。The vibration signal to be tested is input into the intelligent fault diagnosis model of the transmission part of the mechanical equipment to obtain the fault category of the transmission part of the mechanical equipment.
优选地,所述利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码包括:Preferably, using the self-supervised pre-training network to mask the samples in the unlabeled data set includes:
对输入样本[X1;X2;...;Xn]进行随机掩码操作,得到掩码输出矩阵Xm=[X1;X2;...;Xn][c1;c2;...;cn]T,其中,n为所述预设数据点长度,x1至xn表示输入矩阵的n个行向量,c1至cn表示随机生成的掩码向量,其中一半的值为1,其余为0,T表示转置;Perform a random mask operation on the input samples [X 1 ; X 2 ; ...; X n ] to obtain the masked output matrix X m = [X 1 ; X 2 ; ...; X n ][c 1 ; c 2 ;...; c n ] T , wherein, n is the length of the preset data point, x 1 to x n represent n row vectors of the input matrix, c 1 to c n represent randomly generated mask vectors, Half of them have a value of 1, and the rest are 0, T means transpose;
将所述掩码输出矩阵中为0的行向量进行移除,得到最终的掩码输出矩阵XM=Xm-X0。The row vectors of 0 in the mask output matrix are removed to obtain the final mask output matrix X M =X m −X 0 .
优选地,所述重构的编码过程包括:Preferably, the reconstructed encoding process includes:
将所述最终的掩码输出矩阵加上一行用于聚合分类信息的类别令牌向量,并叠加用于指示位置信息的位置编码,得到第一编码特征其中,表示类别令牌向量,表示合并矩阵,表示位置编码;Adding a row of category token vectors for aggregating classification information to the final mask output matrix, and superimposing position encoding for indicating position information to obtain the first encoding feature in, represents a category token vector, represents the merged matrix, Indicates the location code;
将所述第一编码特征经层标准化处理,得到第二编码特征其中,γ为缩放参数,β为平移参数,标准化值第一编码特征在最后一维的均值方差Ndim为第一编码特征最后一维的维数;Standardize the first coding feature to obtain the second coding feature Among them, γ is the scaling parameter, β is the translation parameter, and the normalized value The mean of the first encoded feature in the last dimension variance N dim is the dimension of the last dimension of the first encoding feature;
将所述第二编码特征通过多级堆叠的多头自注意力层和多层感知器层进行处理,得到目标编码特征其中,第d级多头自注意力层和第d级多层感知器层的处理过程为:The second encoding feature is processed through a multi-stage stacked multi-head self-attention layer and a multi-layer perceptron layer to obtain the target encoding feature Among them, the processing process of the d-level multi-head self-attention layer and the d-level multi-layer perceptron layer is:
计算第d-1级多层感知器层输出的自注意力值将其与第d-1级多层感知器层的输出残差连接,得到第d级多头自注意力层的输出其中,depth为多头自注意力层和多层感知器层堆叠的层级数,当d=1时,为第二编码特征的自注意力值,为所述第一编码特征;Calculate the self-attention value of the output of the d-1th multilayer perceptron layer Compare it with the output of the d-1th multilayer perceptron layer Residual connection to get the output of the d-th level multi-head self-attention layer Among them, depth is the number of layers stacked by the multi-head self-attention layer and the multi-layer perceptron layer. When d=1, is the self-attention value of the second encoding feature, is the first encoded feature;
将第d级多头自注意力层的输出和第一线性层权重的乘积与第一线性层偏差b1的和,经过GeLU激活函数处理后,与第二线性层权重相乘再与第二线性层偏差b2相加,最后与第d级多头自注意力层的输出残差连接,得到d级多层感知器层的输出 The output of the d-th level multi-head self-attention layer and the first linear layer weights The sum of the product of and the first linear layer bias b 1 , after GeLU activation function processing, and the second linear layer weight Multiply and then add to the second linear layer bias b 2 , and finally with the output of the d-th multi-head self-attention layer Residual connections to get the output of the d-level multilayer perceptron layer
优选地,所述重构的解码过程包括:Preferably, the reconstructed decoding process includes:
将所述目标编码特征恢复在随机掩码模块被移除的向量yd=y+X0,并通过第三线性层对其进行降维,得到第一解码特征yd′=yd*wd+bd,其中,为第三线性层权重矩阵,bd为第三线性层偏差;Restore the target encoding feature to the vector y d =y+X 0 removed in the random mask module, and perform dimensionality reduction on it through the third linear layer to obtain the first decoding feature y d ′=y d *w d +b d , where, is the weight matrix of the third linear layer, b d is the deviation of the third linear layer;
将所述第一解码特征经过所述多级堆叠的多头自注意力层和多层感知器层进行处理,并通过第四线性层恢复维数,得到重构特征。The first decoding feature is processed through the multi-level stacked multi-head self-attention layer and multi-layer perceptron layer, and the dimension is restored through the fourth linear layer to obtain the reconstructed feature.
优选地,所述重构损失为:Preferably, the reconstruction loss is:
其中,nB代表批尺寸,代表所述重构特征,xi代表所述掩码输出矩阵。where n B represents the batch size, represents the reconstructed feature, and xi represents the mask output matrix.
优选地,所述目标函数的优化方法为反向传播和梯度下降法。Preferably, the optimization method of the objective function is backpropagation and gradient descent method.
优选地,所述分类器模型为Softmax分类器,通过Softmax分类器输出每种故障属于不同类别的概率值,对应的故障概率值的计算公式为:Preferably, the classifier model is a Softmax classifier, and the Softmax classifier outputs the probability values that each fault belongs to a different category, and the calculation formula of the corresponding fault probability value is:
其中,xi′代表第i个输入特征,yi代表对应类别的输出概率值,ns表示故障类别总量。Among them, x i ′ represents the i-th input feature, y i represents the output probability value of the corresponding category, and n s represents the total number of fault categories.
优选地,所述分类损失为:Preferably, the classification loss is:
其中,nb表示批尺寸,ns′代表健康状态的种类总数,1{·}代表一个函数在yj=k时返回1,否则返回0,yj代表样本的真实标签,代表所有健康状态的预测概率。Among them, n b represents the batch size, n s ′ represents the total number of types in the healthy state, 1{ } represents a function that returns 1 when y j =k, otherwise it returns 0, y j represents the real label of the sample, Represents the predicted probabilities for all health states.
本发明还提供了一种机械设备传动部件智能故障诊断装置,包括:The present invention also provides an intelligent fault diagnosis device for transmission parts of mechanical equipment, including:
数据集构建模块,用于采集机械设备传动部件在不同工况下的振动信号,对所述振动信号按预设数据点长度进行多次截取,得到多个样本数据,按预设比例对所述多个样本数据进行划分,对其中一部分样本数据按故障类别进行标定,作为有标签数据集,将另一部分作为无标签数据集;The data set construction module is used to collect the vibration signals of mechanical equipment transmission parts under different working conditions, intercept the vibration signals multiple times according to the preset data point length, obtain multiple sample data, and compare the vibration signals according to the preset ratio Divide multiple sample data, and calibrate part of the sample data according to the fault category as a labeled data set, and use the other part as an unlabeled data set;
自监督预训练网络构建模块,用于通过构建随机掩码模块、基于自注意力机制的编码器模型和解码器模型建立自监督预训练网络,利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码和重构,所述重构包括编码过程和解码过程,计算重构前后信号间的重构损失,并以最小化重构损失作为目标函数对所述自监督预训练网络进行优化,更新自监督预训练网络的网络参数;The self-supervised pre-training network building block is used to establish a self-supervised pre-training network by constructing a random mask module, an encoder model and a decoder model based on a self-attention mechanism, and utilizes the self-supervised pre-training network to the unlabeled The samples in the data set are masked and reconstructed, the reconstruction includes the encoding process and the decoding process, the reconstruction loss between the signals before and after the reconstruction is calculated, and the self-supervised pre-training is performed with the minimum reconstruction loss as the objective function Optimize the network and update the network parameters of the self-supervised pre-training network;
预测模型构建模块,用于通过构建所述编码器模型和分类器模型建立微调网络,并将优化好的自监督预训练网络中编码器模型对应的参数迁移至微调网络的编码器模型中,利用所述微调网络对所述有标签数据集中的样本进行分类,计算分类损失,并以最小化分类损失作为目标函数对所述微调网络进行优化,更新微调网络的网络参数,得到机械设备传动部件智能故障诊断模型;The predictive model building module is used to establish the fine-tuning network by constructing the encoder model and the classifier model, and migrate the parameters corresponding to the encoder model in the optimized self-supervised pre-training network to the encoder model of the fine-tuning network, using The fine-tuning network classifies the samples in the labeled data set, calculates the classification loss, and optimizes the fine-tuning network with minimizing the classification loss as the objective function, updates the network parameters of the fine-tuning network, and obtains the intelligence of the transmission parts of mechanical equipment. Fault diagnosis model;
故障诊断模块,用于将待测试的振动信号输入所述机械设备传动部件智能故障诊断模型中,得到机械设备传动部件的故障类别。The fault diagnosis module is used to input the vibration signal to be tested into the intelligent fault diagnosis model of the transmission part of the mechanical equipment to obtain the fault category of the transmission part of the mechanical equipment.
本发明还提供了一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序,所述计算机程序被处理器执行时实现上述一种机械设备传动部件智能故障诊断方法的步骤。The present invention also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned intelligent fault diagnosis method for transmission parts of mechanical equipment are realized.
本发明的上述技术方案相比现有技术具有以下优点:The above technical solution of the present invention has the following advantages compared with the prior art:
本发明所述的机械设备传动部件智能故障诊断方法,采集不同工况下的振动信号作为训练样本,避免了由于工况改变导致模型性能下降的问题,通过建立自监督预训练网络,充分利用了容易获取的无标签样本训练网络,使网络能够提取更有效的特征,减少对标签样本的依赖;另外,通过基于自注意力机制的编码器模型和解码器模型,提取更全面的全局特征,同时抑制冗余特征,增强有效特征,无需通过预处理对输入数据进行预增强,提高了诊断效率。使用本发明可完成机械设备传动部件智能故障诊断模型在新工况下的诊断任务且拥有良好的泛化性能,适用于现场实时识别机械设备传动部件的健康状态,为基于深度自监督学习的智能诊断方法提供一个可靠、便利的工具,具有重要的领域意义与广阔的应用前景。The intelligent fault diagnosis method of mechanical equipment transmission parts according to the present invention collects vibration signals under different working conditions as training samples, avoids the problem of model performance degradation due to changes in working conditions, and makes full use of the self-supervised pre-training network by establishing a The easy-to-obtain unlabeled sample training network enables the network to extract more effective features and reduce dependence on labeled samples; in addition, through the encoder model and decoder model based on the self-attention mechanism, more comprehensive global features are extracted, and at the same time Suppress redundant features, enhance effective features, without pre-enhancing input data through preprocessing, and improve diagnostic efficiency. Using the present invention can complete the diagnostic task of the intelligent fault diagnosis model of mechanical equipment transmission parts under new working conditions and has good generalization performance. It is suitable for real-time identification of the health status of mechanical equipment transmission parts. The diagnostic method provides a reliable and convenient tool, which has important field significance and broad application prospects.
附图说明Description of drawings
为了使本发明的内容更容易被清楚的理解,下面根据本发明的具体实施例并结合附图,对本发明作进一步详细的说明,其中:In order to make the content of the present invention more easily understood, the present invention will be described in further detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:
图1是本发明一种机械设备传动部件智能故障诊断方法的实现流程图;Fig. 1 is the realization flowchart of a kind of intelligent fault diagnosis method of mechanical equipment transmission part of the present invention;
图2是本发明自监督预训练网络示意图;Fig. 2 is a schematic diagram of the self-supervised pre-training network of the present invention;
图3是本发明微调网络示意图;Fig. 3 is a schematic diagram of fine-tuning network of the present invention;
图4是本发明自监督预训练网络及微调网络中的编码器模型示意图;Fig. 4 is a schematic diagram of the encoder model in the self-supervised pre-training network and the fine-tuning network of the present invention;
图5为本发明实施例提供的机械设备传动部件智能故障诊断装置的结构框图。Fig. 5 is a structural block diagram of an intelligent fault diagnosis device for transmission parts of mechanical equipment provided by an embodiment of the present invention.
具体实施方式Detailed ways
本发明的核心是提供一种机械设备传动部件智能故障诊断方法、装置及计算机存储介质,学习到的特征更加全面,提高了诊断模型的泛化能力和实用性。The core of the present invention is to provide an intelligent fault diagnosis method, device and computer storage medium for transmission parts of mechanical equipment. The learned features are more comprehensive, and the generalization ability and practicability of the diagnosis model are improved.
为了使本技术领域的人员更好地理解本发明方案,下面结合附图和具体实施方式对本发明作进一步的详细说明。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。In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
请参考图1,图1为本发明所提供的机械设备传动部件智能故障诊断方法的实现流程图;具体操作步骤如下:Please refer to Fig. 1, Fig. 1 is the realization flowchart of the intelligent fault diagnosis method of mechanical equipment transmission parts provided by the present invention; the specific operation steps are as follows:
S101:采集机械设备传动部件在不同工况下的振动信号,对所述振动信号按预设数据点长度进行多次截取,得到多个样本数据,按预设比例对所述多个样本数据进行划分,对其中一部分样本数据按故障类别进行标定,作为有标签数据集,将另一部分作为无标签数据集;S101: Collect the vibration signals of the transmission parts of the mechanical equipment under different working conditions, intercept the vibration signals multiple times according to the preset data point length, obtain multiple sample data, and perform the multiple sample data according to the preset ratio Divide some of the sample data according to the fault category, as a labeled data set, and use the other part as an unlabeled data set;
使用传感器采集机械设备传动部件在多种工况下的振动信号,对振动信号按一定的数据点长度截取获得大量的样本集X,选定一种工况下的样本作为测试数据集其他工况下的样本为训练数据集,对训练数据集中已知故障类型的样本根据故障类别进行标定,设定类别标签Y,划分训练数据集为无标签数据集以及有标签数据集其中nT,nU,nL分别代表测试数据集,无标签数据集和有标签数据集中的样本数目。其中所述数据集均来源于实验室实验平台。Use sensors to collect vibration signals of mechanical equipment transmission parts under various working conditions, intercept the vibration signals according to a certain data point length to obtain a large number of sample sets X, and select samples under one working condition as the test data set The samples under other working conditions are training data sets, and the samples of known fault types in the training data set are calibrated according to the fault category, and the category label Y is set, and the training data set is divided into unlabeled data sets and a labeled dataset Among them, n T , n U , and n L represent the number of samples in the test data set, unlabeled data set and labeled data set, respectively. The data sets mentioned above are all derived from the laboratory experiment platform.
如图2所示:as shown in picture 2:
S102:通过构建随机掩码模块、基于自注意力机制的编码器模型和解码器模型建立自监督预训练网络,利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码和重构,所述重构包括编码过程和解码过程,计算重构前后信号间的重构损失,并以最小化重构损失作为目标函数对所述自监督预训练网络进行优化,更新自监督预训练网络的网络参数;S102: Establish a self-supervised pre-training network by constructing a random mask module, an encoder model and a decoder model based on a self-attention mechanism, and use the self-supervised pre-training network to mask and sum samples in the unlabeled data set Reconstruction, the reconstruction includes an encoding process and a decoding process, calculates the reconstruction loss between the signals before and after the reconstruction, and optimizes the self-supervised pre-training network with the minimum reconstruction loss as the objective function, and updates the self-supervised pre-training network Network parameters for training the network;
通过所述随机掩码模块对输入的无标签信号进行掩码并剔除,通过所述编码器模型从保留的原始信号中提取表征,通过所述解码器模型根据提取到的表征重构信号,所述编码器模型和解码器模型均基于自注意力机制,可以有效避免模型提取到的特征不够全面和缺少针对性的问题。The input unlabeled signal is masked and eliminated through the random mask module, the representation is extracted from the retained original signal through the encoder model, and the signal is reconstructed according to the extracted representation through the decoder model, so Both the encoder model and the decoder model are based on the self-attention mechanism, which can effectively avoid the problem that the features extracted by the model are not comprehensive enough and lack of pertinence.
所述编码器模型包括多头自注意力层、层标准化层、多层感知器层和残差连接。The encoder model includes multi-head self-attention layers, layer normalization layers, multi-layer perceptron layers and residual connections.
所述解码器模型包括线性层、多头自注意力层、层标准化层、多层感知器层和残差连接。The decoder model includes linear layers, multi-head self-attention layers, layer normalization layers, multi-layer perceptron layers and residual connections.
如图3所示:As shown in Figure 3:
S103:通过构建所述编码器模型和分类器模型建立微调网络,并将优化好的自监督预训练网络中编码器模型对应的参数迁移至微调网络的编码器模型中,利用所述微调网络对所述有标签数据集中的样本进行分类,计算分类损失,并以最小化分类损失作为目标函数对所述微调网络进行优化,更新微调网络的网络参数,得到机械设备传动部件智能故障诊断模型;S103: Establish a fine-tuning network by constructing the encoder model and classifier model, and migrate the parameters corresponding to the encoder model in the optimized self-supervised pre-training network to the encoder model of the fine-tuning network, and use the fine-tuning network to Classify the samples in the labeled data set, calculate the classification loss, optimize the fine-tuning network with minimizing the classification loss as the objective function, update the network parameters of the fine-tuning network, and obtain an intelligent fault diagnosis model of mechanical equipment transmission parts;
所述微调网络的建立包括编码器模型和分类器模型,其中编码器模型与预训练网络中的编码器模型一致,重复之处不再赘述,通过迁移预训练网络中编码器模型的对应参数,继承其特征提取的能力,然后利用少量带标签数据训练网络适配故障诊断任务。The establishment of the fine-tuning network includes an encoder model and a classifier model, wherein the encoder model is consistent with the encoder model in the pre-training network, and the repetitions will not be repeated. By migrating the corresponding parameters of the encoder model in the pre-training network, Inherit its feature extraction ability, and then use a small amount of labeled data to train the network to adapt to the fault diagnosis task.
所述目标函数的优化方法为反向传播和梯度下降法。The optimization method of the objective function is backpropagation and gradient descent method.
所述分类器模型为Softmax分类器,通过Softmax分类器输出每种故障属于不同类别的概率值,对应的故障概率值的计算公式为:Described classifier model is Softmax classifier, outputs the probability value that each kind of fault belongs to different category by Softmax classifier, and the computing formula of corresponding fault probability value is:
其中,xi′代表第i个输入特征,yi代表对应类别的输出概率值,ns表示故障类别总量。Among them, x i ′ represents the i-th input feature, y i represents the output probability value of the corresponding category, and n s represents the total number of fault categories.
所述分类损失为:The classification loss is:
其中,nb表示批尺寸,ns′代表健康状态的种类总数,1{·}代表一个函数在yj=k时返回1,否则返回0,yj代表样本的真实标签,代表所有健康状态的预测概率。Among them, n b represents the batch size, n s ′ represents the total number of types in the healthy state, 1{ } represents a function that returns 1 when y j =k, otherwise it returns 0, y j represents the real label of the sample, Represents the predicted probabilities for all health states.
S104:将待测试的振动信号输入所述机械设备传动部件智能故障诊断模型中,得到机械设备传动部件的故障类别。S104: Input the vibration signal to be tested into the intelligent fault diagnosis model of the transmission part of the mechanical equipment to obtain the fault category of the transmission part of the mechanical equipment.
本发明所述的机械设备传动部件智能故障诊断方法,采集不同工况下的振动信号作为训练样本,避免了由于工况改变导致模型性能下降的问题,通过建立自监督预训练网络,充分利用了容易获取的无标签样本训练网络,使网络能够提取更有效的特征,减少对标签样本的依赖;另外,通过基于自注意力机制的编码器模型和解码器模型,提取更全面的全局特征,同时抑制冗余特征,增强有效特征,无需通过预处理对输入数据进行预增强,提高了诊断效率。使用本发明可完成机械设备传动部件智能故障诊断模型在新工况下的诊断任务且拥有良好的泛化性能,适用于现场实时识别机械设备传动部件的健康状态,为基于深度自监督学习的智能诊断方法提供一个可靠、便利的工具,具有重要的领域意义与广阔的应用前景。The intelligent fault diagnosis method of mechanical equipment transmission parts according to the present invention collects vibration signals under different working conditions as training samples, avoids the problem of model performance degradation due to changes in working conditions, and makes full use of the self-supervised pre-training network by establishing a The easy-to-obtain unlabeled sample training network enables the network to extract more effective features and reduce dependence on labeled samples; in addition, through the encoder model and decoder model based on the self-attention mechanism, more comprehensive global features are extracted, and at the same time Suppress redundant features, enhance effective features, without pre-enhancing input data through preprocessing, and improve diagnostic efficiency. Using the present invention can complete the diagnostic task of the intelligent fault diagnosis model of mechanical equipment transmission parts under new working conditions and has good generalization performance. It is suitable for real-time identification of the health status of mechanical equipment transmission parts. The diagnostic method provides a reliable and convenient tool, which has important field significance and broad application prospects.
基于以上实施例,本实施例对步骤S102进行进一步详细说明:Based on the above embodiments, this embodiment further describes step S102 in detail:
所述利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码包括:Said using said self-supervised pre-training network to mask samples in said unlabeled data set includes:
对输入样本[X1;X2;...;Xn]进行随机掩码操作,得到掩码输出矩阵Xm=[X1;X2;...;Xn][c1;c2;...;cn]T,其中,n为所述预设数据点长度,x1至xn表示输入矩阵的n个行向量,c1至cn表示随机生成的掩码向量,其中一半的值为1,其余为0,T表示转置;Perform a random mask operation on the input samples [X 1 ; X 2 ; ...; X n ] to obtain the masked output matrix X m = [X 1 ; X 2 ; ...; X n ][c 1 ; c 2 ;...; c n ] T , wherein, n is the length of the preset data point, x 1 to x n represent n row vectors of the input matrix, c 1 to c n represent randomly generated mask vectors, Half of them have a value of 1, and the rest are 0, T means transpose;
将所述掩码输出矩阵中置为0的行向量转化为可训练的参数并暂时移除,得到最终的掩码输出矩阵XM=Xm-X0。The row vectors set to 0 in the mask output matrix are converted into trainable parameters and temporarily removed to obtain the final mask output matrix X M =X m −X 0 .
最终输出的掩码矩阵将输入到编码器模型进行编码表征。The mask matrix of the final output The input is fed to the encoder model for encoding representation.
利用所述自监督预训练网络对所述无标签数据集中的样本进行重构具体包括:Using the self-supervised pre-training network to reconstruct the samples in the unlabeled data set specifically includes:
如图4所示,所述重构的编码过程包括:As shown in Figure 4, the encoding process of the reconstruction includes:
将所述最终的掩码输出矩阵加上一行用于聚合分类信息的类别令牌向量,并叠加用于指示位置信息的位置编码,得到第一编码特征其中,表示类别令牌向量,表示合并矩阵,表示位置编码;Adding a row of category token vectors for aggregating classification information to the final mask output matrix, and superimposing position encoding for indicating position information to obtain the first encoding feature in, represents a category token vector, represents the merged matrix, Indicates the location code;
将所述第一编码特征经层标准化处理,得到第二编码特征X,数学表述为:The first coding feature is subjected to layer standardization processing to obtain the second coding feature X, and the mathematical expression is:
其中,μ为输入在最后一维的均值,σ2为方差,为标准化的值,ε是平滑因子,防止方差为0时,输出为无穷大,在最终批次标准化输出中引入缩放参数γ和平移参数β来进一步提高数值输出稳定性,所述缩放参数γ和平移参数β参数在网络训练时采用反向传播算法进行更新,Ndim为第一编码特征最后一维的维数;Among them, μ is the mean value of the input in the last dimension, σ 2 is the variance, is a standardized value, ε is a smoothing factor, and prevents the output from being infinite when the variance is 0, and introduces a scaling parameter γ and a translation parameter β in the final batch normalized output to further improve the stability of the numerical output. The scaling parameter γ and translation The parameter β is updated using the backpropagation algorithm during network training, and N dim is the dimension of the last dimension of the first encoding feature;
将所述第二编码特征通过多级堆叠的多头自注意力层和多层感知器层进行处理,得到目标编码特征其中,第d级多头自注意力层和第d级多层感知器层的处理过程为:The second encoding feature is processed through a multi-stage stacked multi-head self-attention layer and a multi-layer perceptron layer to obtain the target encoding feature Among them, the processing process of the d-level multi-head self-attention layer and the d-level multi-layer perceptron layer is:
计算第d-1级多层感知器层输出的自注意力值将其与第d-1级多层感知器层的输出残差连接,以防止梯度消失,得到第d级多头自注意力层的输出d=1,2,3,...,depth,其中,depth为多头自注意力层和多层感知器层堆叠的层级数,当d=1时,为第二编码特征的自注意力值,为所述第一编码特征;Calculate the self-attention value of the output of the d-1th multilayer perceptron layer Compare it with the output of the d-1th multilayer perceptron layer Residual connections to prevent vanishing gradients to get the output of the d-th level multi-head self-attention layer d=1,2,3,...,depth, where depth is the number of layers stacked by the multi-head self-attention layer and the multi-layer perceptron layer. When d=1, is the self-attention value of the second encoding feature, is the first encoded feature;
将第d级多头自注意力层的输出和第一线性层权重的乘积与第一线性层偏差b1的和,经过GeLU激活函数处理后,与第二线性层权重相乘再与第二线性层偏差b2相加,最后与第d级多头自注意力层的输出残差连接,得到d级多层感知器层的输出,以实现复杂的非线性映射d=1,2,3,...,depth。The output of the d-th level multi-head self-attention layer and the first linear layer weights The sum of the product of and the first linear layer bias b 1 , after GeLU activation function processing, and the second linear layer weight Multiply and then add to the second linear layer bias b 2 , and finally with the output of the d-th multi-head self-attention layer Residual connections to get the output of d-level multilayer perceptron layers for complex non-linear mappings d=1,2,3,...,depth.
计算多头自注意力值的数学表述为:The mathematical expression for calculating the multi-head self-attention value is:
Q=X·Wq,K=X·Wk,V=X·Wv Q=X·W q , K=X·W k , V=X·W v
headi=Attention(Q·Wi Q,K·Wi K,V·Wi V)head i =Attention(Q·W i Q ,K·W i K ,V·W i V )
yMSA′=Concat(head1,...headi)WO y MSA ′=Concat(head 1 ,...head i )W O
其中,代表查询矩阵,用于后续与键矩阵进行匹配,代表键矩阵,用于被查询矩阵匹配,表示值矩阵,代表从原始输入中提取的信息,和分别为各自的可学习的变换矩阵,为缩放因子,匹配过程计算查询矩阵和键矩阵的乘积、除缩放因子和通过Softmax函数,计算结果为两者的相关性,后续作为值矩阵的权重与值矩阵相乘;通过多组不同的变换矩阵从不同的特征子空间中获取in, Represents the query matrix for subsequent matching with the key matrix, Represents the key matrix, which is used to match the query matrix, Represents a matrix of values, representing the input from the original information extracted from the and are the respective learnable transformation matrices, is the scaling factor, the matching process calculates the product of the query matrix and the key matrix, divides the scaling factor and passes the Softmax function, and the calculation result is the correlation between the two, and the subsequent weight as the value matrix is multiplied by the value matrix; through multiple sets of different transformations matrices are obtained from different eigensubspaces
′'
信息,合并多组自注意力的计算结果,得到多头自注意力值yMSA。information, and combine the calculation results of multiple sets of self-attention to obtain the multi-head self-attention value y MSA .
所述重构的解码过程包括:The reconstructed decoding process includes:
将所述目标编码特征恢复在随机掩码模块被移除的向量yd=y+X0,并通过第三线性层对其进行降维,得到第一解码特征yd′=yd*wd+bd,其中,为第三线性层权重矩阵,bd为第三线性层偏差;Restore the target encoding feature to the vector y d =y+X 0 removed by the random mask module, And through the third linear layer to reduce the dimension, get the first decoding feature y d ′=y d *w d +b d , in, is the weight matrix of the third linear layer, b d is the deviation of the third linear layer;
将所述第一解码特征经过所述多级堆叠的多头自注意力层和多层感知器层进行处理,并通过第四线性层恢复维数,得到重构特征。The first decoding feature is processed through the multi-level stacked multi-head self-attention layer and multi-layer perceptron layer, and the dimension is restored through the fourth linear layer to obtain the reconstructed feature.
所述重构损失为:The reconstruction loss is:
其中,nB代表批尺寸,代表所述重构特征,xi代表所述掩码输出矩阵。where n B represents the batch size, represents the reconstructed feature, and xi represents the mask output matrix.
基于以上实施例,本实施例以一种具体的实验来验证本发明的有效性:Based on above embodiment, present embodiment verifies the validity of the present invention with a kind of specific experiment:
实验所用数据来自于苏州大学轴承故障模拟实验台,轴承故障数据集共有七种健康状态,包括轴承内圈0.2mm故障,滚动体0.2mm故障,外圈0.2mm故障以及复合故障类型(包括内圈与滚动体复合故障,内圈与外圈复合故障,外圈与滚动体复合故障,内圈、外圈与滚动体复合故障)。每种健康状态下的数据均在800rpm转速以及四种不同负载(0KN,0.8KN,1.6KN,2.5KN)下采集。每种工况下每一种健康状态包含200个样本,每个样本含有1024个数据点。选定负载2.5KN下的数据为测试数据集,其余负载下的数据为训练测试集,给定训练测试集中1.6KN负载下的部分样本标签,数量为每种健康状态(5,10,15,20)个,其余均为无标签样本。实验中用到的样本的详细信息如表1所示。The data used in the experiment comes from the bearing fault simulation test bench of Soochow University. There are seven health states in the bearing fault data set, including bearing inner ring 0.2 mm fault, rolling element 0.2 mm fault, outer ring 0.2 mm fault and composite fault types (including inner ring Composite fault with rolling element, composite fault of inner ring and outer ring, composite fault of outer ring and rolling element, composite fault of inner ring, outer ring and rolling element). The data in each health state were collected at 800rpm and four different loads (0KN, 0.8KN, 1.6KN, 2.5KN). Each health state contains 200 samples in each working condition, and each sample contains 1024 data points. The data under the selected load of 2.5KN is the test data set, and the data under the rest of the load is the training test set. Given the partial sample labels under the 1.6KN load of the training test set, the number is each health state (5, 10, 15, 20), and the rest are unlabeled samples. The details of the samples used in the experiment are shown in Table 1.
表1每种健康状态下的样本数量Table 1 Number of samples in each health status
为验证所提出的发明的有效性,本实验案例实施了四个不同少标签样本情况下的诊断任务,其中标签样本的数量分别为每种健康状态下5,10,15和20个,分别占总样本数量的2.5%,5%,7.5%和10%。为验证本发明所提出的一种机械设备传动部件智能故障诊断方法及系统的有效性,其他五种先进的基于自监督学习的智能诊断方法作为比较,包括自注意力编码器(SAE),卷积自编码器(CAE),上下文自编码器(CE),去噪卷积神经网络(DCNN)和去噪自编码器(DAE)。各个网络模型训练时的参数设置为:初始学习率为0.001,批尺寸为128,迭代次数为50。In order to verify the effectiveness of the proposed invention, this experimental case implements four diagnostic tasks under different situations of few labeled samples, where the number of labeled samples is 5, 10, 15 and 20 for each health state, accounting for 2.5%, 5%, 7.5% and 10% of the total sample size. In order to verify the validity of a method and system for intelligent fault diagnosis of mechanical equipment transmission parts proposed by the present invention, other five advanced intelligent diagnosis methods based on self-supervised learning are used as comparisons, including self-attention encoder (SAE), volume Product Autoencoder (CAE), Contextual Autoencoder (CE), Denoising Convolutional Neural Network (DCNN) and Denoising Autoencoder (DAE). The parameter settings for each network model training are: the initial learning rate is 0.001, the batch size is 128, and the number of iterations is 50.
本发明方法与其他自监督学习智能诊断方法相比,所提方法的分类精度在四个不同少标签样本情况下的诊断任务取得了最好的效果,显示出本发明的优越性,比较结果如表2所示。Compared with other self-supervised learning intelligent diagnosis methods, the classification accuracy of the proposed method has achieved the best results in the diagnostic tasks of four different few-label samples, showing the superiority of the present invention. The comparison results are as follows: Table 2 shows.
表2每种方法在各个诊断任务的准确率Table 2 Accuracy of each method in each diagnostic task
本发明针对现有的智能故障诊断方法严重依赖标签样本、诊断网络提取到的特征不够全面和缺少针对性的问题,以机械设备传动部件为研究对象,通过自监督学习方法从无标签信号中提取监督信息,增强网络提取有效表征的性能,利用自注意力机制构建网络模型,提取全面且有针对性的特征,提高模型的泛化性能。Aiming at the problems that the existing intelligent fault diagnosis method relies heavily on label samples, and the features extracted by the diagnosis network are not comprehensive enough and lack of pertinence, the present invention takes the transmission parts of mechanical equipment as the research object, and extracts them from unlabeled signals through a self-supervised learning method. Supervise information, enhance the performance of the network to extract effective representations, use the self-attention mechanism to build a network model, extract comprehensive and targeted features, and improve the generalization performance of the model.
请参考图5,图5为本发明实施例提供的机械设备传动部件智能故障诊断装置的结构框图;具体装置可以包括:Please refer to FIG. 5, which is a structural block diagram of an intelligent fault diagnosis device for transmission parts of mechanical equipment provided by an embodiment of the present invention; the specific device may include:
数据集构建模块100,用于采集机械设备传动部件在不同工况下的振动信号,对所述振动信号按预设数据点长度进行多次截取,得到多个样本数据,按预设比例对所述多个样本数据进行划分,对其中一部分样本数据按故障类别进行标定,作为有标签数据集,将另一部分作为无标签数据集;The data
自监督预训练网络构建模块200,用于通过构建随机掩码模块、基于自注意力机制的编码器模型和解码器模型建立自监督预训练网络,利用所述自监督预训练网络对所述无标签数据集中的样本进行掩码和重构,所述重构包括编码过程和解码过程,计算重构前后信号间的重构损失,并以最小化重构损失作为目标函数对所述自监督预训练网络进行优化,更新自监督预训练网络的网络参数;The self-supervised pre-training
预测模型构建模块300,用于通过构建所述编码器模型和分类器模型建立微调网络,并将优化好的自监督预训练网络中编码器模型对应的参数迁移至微调网络的编码器模型中,利用所述微调网络对所述有标签数据集中的样本进行分类,计算分类损失,并以最小化分类损失作为目标函数对所述微调网络进行优化,更新微调网络的网络参数,得到机械设备传动部件智能故障诊断模型;The prediction
故障诊断模块400,用于将待测试的振动信号输入所述机械设备传动部件智能故障诊断模型中,得到机械设备传动部件的故障类别。The
本实施例的机械设备传动部件智能故障诊断装置用于实现前述的机械设备传动部件智能故障诊断方法,因此机械设备传动部件智能故障诊断装置中的具体实施方式可见前文机械设备传动部件智能故障诊断方法的实施例部分,例如,数据集构建模块100,自监督预训练网络构建模块200,预测模型构建模块300,故障诊断模块400,分别用于实现上述机械设备传动部件智能故障诊断方法中步骤S101,S102,S103,S104,所以,其具体实施方式可以参照相应的各个部分实施例的描述,在此不再赘述。The intelligent fault diagnosis device for transmission parts of mechanical equipment in this embodiment is used to implement the aforementioned intelligent fault diagnosis method for transmission parts of mechanical equipment. Therefore, the specific implementation of the intelligent fault diagnosis device for transmission parts of mechanical equipment can be seen in the intelligent fault diagnosis method for transmission parts of mechanical equipment. In the embodiment part, for example, the data
本发明具体实施例还提供了一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序,所述计算机程序被处理器执行时实现上述一种机械设备传动部件智能故障诊断方法的步骤。A specific embodiment of the present invention also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the above-mentioned method for intelligent fault diagnosis of transmission parts of mechanical equipment is realized A step of.
本领域内的技术人员应明白,本申请的实施例可提供为方法、系统、或计算机程序产品。因此,本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且,本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
本申请是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器,使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。The present application is described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中,使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品,该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上,使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理,从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.
显然,上述实施例仅仅是为清楚地说明所作的举例,并非对实施方式的限定。对于所属领域的普通技术人员来说,在上述说明的基础上还可以做出其它不同形式变化或变动。这里无需也无法对所有的实施方式予以穷举。而由此所引伸出的显而易见的变化或变动仍处于本发明创造的保护范围之中。Apparently, the above-mentioned embodiments are only examples for clear description, and are not intended to limit the implementation. For those of ordinary skill in the art, on the basis of the above description, other changes or changes in various forms can also be made. It is not necessary and impossible to exhaustively list all the implementation manners here. And the obvious changes or changes derived therefrom are still within the scope of protection of the present invention.
Claims (10)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211575101.2A CN116007937B (en) | 2022-12-08 | 2022-12-08 | Intelligent fault diagnosis method and device for transmission parts of mechanical equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211575101.2A CN116007937B (en) | 2022-12-08 | 2022-12-08 | Intelligent fault diagnosis method and device for transmission parts of mechanical equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116007937A true CN116007937A (en) | 2023-04-25 |
CN116007937B CN116007937B (en) | 2024-09-06 |
Family
ID=86020785
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211575101.2A Active CN116007937B (en) | 2022-12-08 | 2022-12-08 | Intelligent fault diagnosis method and device for transmission parts of mechanical equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116007937B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117828337A (en) * | 2024-03-04 | 2024-04-05 | 北京奥康银华科技有限公司 | A digital signal processing method |
CN117874593A (en) * | 2023-12-11 | 2024-04-12 | 华北电力大学 | Fan bearing fault diagnosis model training method, device, electronic equipment and medium |
CN117992862A (en) * | 2024-04-07 | 2024-05-07 | 山东大学 | Two-stage rotating machinery equipment reliability evaluation method and system based on large model |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111562109A (en) * | 2020-05-19 | 2020-08-21 | 重庆邮电大学 | A deep learning state recognition and diagnosis method for mechanical equipment |
CN111916067A (en) * | 2020-07-27 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Training method and device of voice recognition model, electronic equipment and storage medium |
CN112990578A (en) * | 2021-03-16 | 2021-06-18 | 北京航空航天大学 | Industrial low-quality high-noise data modeling and product quality prediction method |
WO2022110730A1 (en) * | 2020-11-27 | 2022-06-02 | 平安科技(深圳)有限公司 | Label-based optimization model training method, apparatus, device, and storage medium |
CN114993677A (en) * | 2022-05-11 | 2022-09-02 | 山东大学 | Rolling bearing fault diagnosis method and system based on unbalanced small sample data |
US20220327035A1 (en) * | 2020-06-03 | 2022-10-13 | Soochow University | Intra-class adaptation fault diagnosis method for bearing under variable working conditions |
CN115410032A (en) * | 2022-07-26 | 2022-11-29 | 北京工业大学 | OCTA image classification structure training method based on self-supervised learning |
-
2022
- 2022-12-08 CN CN202211575101.2A patent/CN116007937B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111562109A (en) * | 2020-05-19 | 2020-08-21 | 重庆邮电大学 | A deep learning state recognition and diagnosis method for mechanical equipment |
US20220327035A1 (en) * | 2020-06-03 | 2022-10-13 | Soochow University | Intra-class adaptation fault diagnosis method for bearing under variable working conditions |
CN111916067A (en) * | 2020-07-27 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Training method and device of voice recognition model, electronic equipment and storage medium |
WO2022110730A1 (en) * | 2020-11-27 | 2022-06-02 | 平安科技(深圳)有限公司 | Label-based optimization model training method, apparatus, device, and storage medium |
CN112990578A (en) * | 2021-03-16 | 2021-06-18 | 北京航空航天大学 | Industrial low-quality high-noise data modeling and product quality prediction method |
CN114993677A (en) * | 2022-05-11 | 2022-09-02 | 山东大学 | Rolling bearing fault diagnosis method and system based on unbalanced small sample data |
CN115410032A (en) * | 2022-07-26 | 2022-11-29 | 北京工业大学 | OCTA image classification structure training method based on self-supervised learning |
Non-Patent Citations (2)
Title |
---|
岳增营 等: "基于语言模型的预训练技术研究综述", 中文信息学报, vol. 35, no. 9, 30 September 2021 (2021-09-30), pages 15 - 29 * |
戴俊 等: "基于生成对抗网络和自动编码器的机械系统异常检测", 仪器仪表学报, vol. 40, no. 9, 30 September 2019 (2019-09-30), pages 16 - 26 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117874593A (en) * | 2023-12-11 | 2024-04-12 | 华北电力大学 | Fan bearing fault diagnosis model training method, device, electronic equipment and medium |
CN117828337A (en) * | 2024-03-04 | 2024-04-05 | 北京奥康银华科技有限公司 | A digital signal processing method |
CN117828337B (en) * | 2024-03-04 | 2024-12-27 | 北京奥康银华科技有限公司 | Digital signal processing method |
CN117992862A (en) * | 2024-04-07 | 2024-05-07 | 山东大学 | Two-stage rotating machinery equipment reliability evaluation method and system based on large model |
CN117992862B (en) * | 2024-04-07 | 2024-07-12 | 山东大学 | Two-stage rotating mechanical equipment reliability assessment method and system based on large model |
Also Published As
Publication number | Publication date |
---|---|
CN116007937B (en) | 2024-09-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109492822B (en) | Temporal and spatial correlation prediction method of air pollutant concentration | |
CN116007937A (en) | Intelligent fault diagnosis method and device for mechanical equipment transmission part | |
CN111783362B (en) | A method and system for determining the remaining service life of an electric gate valve | |
CN109598387A (en) | Forecasting of Stock Prices method and system based on two-way cross-module state attention network model | |
CN113361559B (en) | Multimodal data knowledge information extraction method based on deep-width joint neural network | |
CN111507046B (en) | Method and system for predicting remaining service life of electric gate valve | |
CN113032778B (en) | Semi-supervised network abnormal behavior detection method based on behavior feature coding | |
Dai et al. | Hybrid deep model for human behavior understanding on industrial internet of video things | |
CN106874963A (en) | A kind of Fault Diagnosis Method for Distribution Networks and system based on big data technology | |
CN117744483A (en) | Bearing fault diagnosis method based on fusion of twin information model and measured data | |
CN118152901B (en) | Equipment failure prediction method and system based on data drive | |
CN116304587B (en) | Rolling bearing degradation trend prediction method based on CAE and AGRU | |
CN117909881A (en) | Fault diagnosis method and device for multi-source data fusion pumping unit | |
CN113762967B (en) | Risk information determination methods, model training methods, equipment, program products | |
CN113158134B (en) | Method, device and storage medium for constructing non-invasive load identification model | |
CN117390407A (en) | Fault identification methods, systems, media and equipment for substation equipment | |
CN118467229A (en) | A database alarm intelligent diagnosis method and system based on multimodal knowledge graph fusion and small sample learning | |
CN113051828B (en) | Online prediction method for natural gas water dew point driven by technological parameters | |
CN116596150A (en) | Event Prediction Method Based on Multi-Branch Self-Attention Transformer Hawkes Process Model | |
CN114398500A (en) | An event prediction method based on a graph-augmented pre-training model | |
CN116402352A (en) | Enterprise risk prediction method and device, electronic equipment and medium | |
CN113643141B (en) | Method, device, equipment and storage medium for generating interpretation conclusion report | |
CN115879018A (en) | Cabin equipment state perception and fault diagnosis method based on K-means algorithm and BP neural network | |
CN115563225A (en) | Power grid equipment fault diagnosis method and system based on knowledge graph relation reasoning | |
CN114880538A (en) | Attribute graph community detection method based on self-supervision |
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 |