CN116659865A - Bearing fault detection method and system based on definite learning and convolutional neural network - Google Patents
Bearing fault detection method and system based on definite learning and convolutional neural network Download PDFInfo
- Publication number
- CN116659865A CN116659865A CN202310647151.5A CN202310647151A CN116659865A CN 116659865 A CN116659865 A CN 116659865A CN 202310647151 A CN202310647151 A CN 202310647151A CN 116659865 A CN116659865 A CN 116659865A
- Authority
- CN
- China
- Prior art keywords
- vibration signal
- bearing
- neural network
- bearing vibration
- dynamic information
- 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.)
- Pending
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 61
- 238000001514 detection method Methods 0.000 title claims abstract description 48
- 238000013528 artificial neural network Methods 0.000 claims abstract description 43
- 230000007246 mechanism Effects 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000000034 method Methods 0.000 claims description 47
- 230000006870 function Effects 0.000 claims description 28
- 238000012549 training Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 abstract description 12
- 238000003745 diagnosis Methods 0.000 abstract description 9
- 230000008859 change Effects 0.000 abstract description 2
- 238000000605 extraction Methods 0.000 description 13
- 238000005096 rolling process Methods 0.000 description 12
- 210000002569 neuron Anatomy 0.000 description 5
- 230000005284 excitation Effects 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 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
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
- 238000007619 statistical method 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
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- 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
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Acoustics & Sound (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The application belongs to the field of bearing fault diagnosis, and provides a bearing fault detection method and system based on a definite learning and convolution neural network, comprising the steps of obtaining a one-dimensional bearing vibration signal of a bearing fault, and expanding the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using a high-gain observer; carrying out maximum value standardization processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal; modeling internal dynamic information of the bearing vibration signal based on a determined learning mechanism, and generating a dynamic information graph of the bearing vibration signal according to the internal dynamic information of the bearing vibration signal; based on a dynamic information graph of the bearing vibration signal, fault detection is carried out in a pre-trained convolutional neural network model, and a bearing fault detection result is obtained. According to the application, the dynamic information of the bearing vibration signal is visualized into a dynamic information diagram for the first time from the perspective of a dynamic system, and the occurrence of faults is detected through the change of the dynamic information.
Description
Technical Field
The application belongs to the technical field of bearing fault diagnosis, and particularly relates to a bearing fault detection method and system based on a definite learning and convolution neural network.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Fault diagnosis by monitoring the operating state of a machine, identifying the type of fault that has occurred and the time of occurrence is critical to improving the reliability and safety of the machine system. Bearings are critical components of rotating equipment and have a high failure rate (e.g., the failure caused by the bearings accounts for about 41% of the total failure of the motor). The detection of bearing faults is critical, which not only can destroy the reliability and safety of equipment, but can even cause serious economic loss and casualties.
Feature extraction is a key step in fault diagnosis methods. Common artificial feature extraction methods include statistical analysis, short-time Fourier transform, wavelet transform, empirical mode decomposition, and the like. However, the effectiveness of artificial feature extraction is highly dependent on a priori knowledge. In recent years, intelligent fault diagnosis based on deep learning can perform automatic feature extraction for end-to-end fault detection, and Convolutional Neural Networks (CNNs) are particularly suitable for deep feature extraction. Currently, there are some methods that can perform fault diagnosis on feature learning of an image by CNN by converting a vibration time-series signal into a vibration image. Existing studies indicate that when a system fails, there is often a change from healthy dynamics to failure dynamics. At present, a CNN fault diagnosis method for converting kinetic information into an image from the perspective of a power system does not exist.
Disclosure of Invention
In order to solve the problems, the application provides a bearing fault detection method and a system based on a definite learning and convolution neural network.
According to some embodiments, a first aspect of the present application provides a method for detecting a bearing fault based on a deterministic learning and convolutional neural network, which adopts the following technical scheme:
the bearing fault detection method based on the determined learning and convolution neural network comprises the following steps:
acquiring a one-dimensional bearing vibration signal of a bearing fault, and expanding the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using a high-gain observer;
carrying out maximum value standardization processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
modeling internal dynamic information of the bearing vibration signal based on a determined learning mechanism, and generating a dynamic information graph of the bearing vibration signal according to the internal dynamic information of the bearing vibration signal;
based on a dynamic information graph of the bearing vibration signal, fault detection is carried out in a pre-trained convolutional neural network model, and a bearing fault detection result is obtained.
Further, the maximum value standardization processing is performed on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal, which specifically comprises the following steps:
carrying out maximum value standardization processing on each dimension in the two-dimensional bearing vibration signal;
the normalization function is:
wherein ,xi Representing a one-dimensional time series of the ith sample,representing the maximum value of all samples of the vibration data set in the j-dimension,/->Representing the normalized sample.
Further, the modeling of the internal dynamics information of the bearing vibration signal based on the determined learning mechanism generates a dynamics information map of the bearing vibration signal according to the internal dynamics information of the bearing vibration signal, specifically:
modeling the bearing vibration signal by adopting a definite learning mechanism to obtain a dynamic information model of the bearing vibration signal;
gridding the track area of the normalized bearing vibration signal to obtain gridded dynamic information input;
based on the gridding dynamic information input, a dynamic information graph of the bearing vibration signal is obtained by utilizing a dynamic information model of the bearing signal.
Further, the dynamic information model of the bearing vibration signal specifically comprises the following steps:
wherein ,representing status trace +.>Neighborhood of->Indicating that RBF neural network weight is converged and then is positioned at Tk a To Tk b The average weights over the time interval, i=1, 2, t is the sampling time,the weights S (z) representing the neural network to be estimated represent regression vectors of radial basis function composition.
Further, the dynamic information graph of the bearing vibration signal is an m×m matrix, specifically:
wherein ,zi (1),…,z i (m), i=1, 2 means the interval [ -1,1 []Evenly dividing the data points into m-1 parts of corresponding data points;
the matrix is converted into a gray scale map, i.e., a dynamic information map for each bearing vibration signal is formed.
Further, the training process of the convolutional neural network model specifically comprises the following steps:
aiming at the generated vibration signal dynamic information graph, the training set and the testing set are divided according to the proportion of 7:3;
inputting a training set bearing vibration signal dynamic information graph into a convolutional neural network to detect bearing faults;
the loss function adopts cross entropy loss, and a gradient descent algorithm is selected to optimize and update network parameters, so that a trained convolutional neural network model is obtained;
and inputting the dynamic information graph of the vibration signal of the bearing by using the test set into a trained convolutional neural network model to obtain a network fault detection result.
Further, the structure of the convolutional neural network model comprises a convolutional layer, a pooling layer, a leveling layer, a full-connection layer and a full-connection layer which are sequentially connected;
and (3) using ELU activation functions and batch normalization processing between the convolution layer and the pooling layer, flattening and inputting the finally extracted feature map into a full-connection layer, and taking the output of the full-connection layer as the output of the whole convolution neural network.
According to some embodiments, a second aspect of the present application provides a bearing fault detection system based on a deterministic learning and convolutional neural network, which adopts the following technical scheme:
a bearing failure detection system based on a deterministic learning and convolutional neural network, comprising:
the data acquisition module is configured to acquire a one-dimensional bearing vibration signal of a bearing fault, and expand the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using the high-gain observer;
the standardized processing module is configured to perform maximum value standardized processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
a dynamics information acquisition module configured to model internal dynamics information of the bearing vibration signal based on the determined learning mechanism, and generate a dynamics information map of the bearing vibration signal from the internal dynamics information of the bearing vibration signal;
the bearing fault detection module is configured to perform fault detection by utilizing a pre-trained convolutional neural network model based on a dynamic information graph of the bearing vibration signal to obtain a bearing fault detection result.
According to some embodiments, a third aspect of the present application provides a computer-readable storage medium.
A computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps in a method of bearing failure detection based on a deterministic learning and convolutional neural network as described in the first aspect above.
According to some embodiments, a fourth aspect of the application provides a computer device.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in a method of bearing failure detection based on a deterministic learning and convolutional neural network as described in the first aspect above when the program is executed.
Compared with the prior art, the application has the beneficial effects that:
the application provides a definite learning convolutional neural network framework for bearing fault detection by combining the accurate modeling of definite learning on the inherent dynamics of data and the automatic feature extraction capability of the convolutional neural network, and carries out bearing fault detection on the basis. Specifically, first, the intrinsic kinetic information of the original vibration time series is locally and accurately modeled by determining a learning mechanism. Secondly, the dynamic information learned by the determined learning mechanism is visualized as a dynamic information graph. And finally, extracting features of the dynamic information graph by using a convolutional neural network, and effectively classifying and detecting different bearing fault types based on the extracted distinguishing features. The application provides a high-efficiency and accurate method for bearing fault detection, and can effectively improve the reliability and safety of a rotary mechanical system.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application.
FIG. 1 is a flow chart of a method for detecting bearing faults based on a deterministic learning and convolutional neural network in an embodiment of the present application;
FIG. 2 is a training flow diagram of a dynamic radial basis function neural network model in an embodiment of the application;
fig. 3 is a block diagram of a convolutional neural network in an embodiment of the present application.
Detailed Description
The application will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Embodiments of the application and features of the embodiments may be combined with each other without conflict.
Example 1
As shown in fig. 1, the present embodiment provides a method for detecting bearing faults based on a deterministic learning and convolutional neural network, and the present embodiment is exemplified by the application of the method to a server, and it can be understood that the method can also be applied to a terminal, and can also be applied to a system and a terminal, and implemented through interaction between the terminal and the server. The server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and can also be a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, security services CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like. The terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein. In this embodiment, the method includes the steps of:
acquiring a one-dimensional bearing vibration signal of a bearing fault, and expanding the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using a high-gain observer;
carrying out maximum value standardization processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
modeling internal dynamic information of the bearing vibration signal based on a determined learning mechanism, and generating a dynamic information graph of the bearing vibration signal according to the internal dynamic information of the bearing vibration signal;
based on a dynamic information graph of the bearing vibration signal, fault detection is carried out in a pre-trained convolutional neural network model, and a bearing fault detection result is obtained.
The embodiment discloses a bearing fault detection method based on combination of deterministic learning and a convolutional neural network, and provides a deterministic learning convolutional neural network framework for bearing fault detection by combining the accurate modeling capability of the inherent dynamics of data by a deterministic learning mechanism and the automatic feature extraction capability of the convolutional neural network, and the bearing fault detection is carried out on the basis. Specifically, the inherent dynamics of the original vibration time series are first learned, i.e. modeled, by determining a learning mechanism. The dynamic information learned by the determined learning mechanism is visualized as a dynamic information graph, and then the dynamic information graph is input into the convolutional neural network designed by the method for automatic feature extraction. And effectively classifying and detecting the fault types of different bearings by using the extracted distinguishing features. The application provides a high-efficiency and accurate method for bearing fault detection, and can effectively improve the reliability and safety of a rotary mechanical system.
The dynamic information extraction method used in the embodiment comes from determining a learning mechanism, and the further fault diagnosis method adopts a convolutional neural network.
Deterministic learning theory provides an effective method for coping with challenges in knowledge acquisition, expression and utilization in dynamic systems. Under the condition of meeting the continuous excitation (PE), a deterministic learning mechanism provides a solution for the local accurate identification or modeling of a nonlinear power system. The modeling dynamics may be expressed and stored by a constant RBF neural network. And such stored knowledge can be effectively used for tasks in dynamic environments such as: dynamic pattern recognition, intelligent control, and the like.
(1) A Radial Basis Function (RBF) neural network is adopted;
(2) The periodic (or regression) track meets the condition of partial continuous excitation;
(3) Realizing local accurate modeling of nonlinear system dynamics along a track;
(4) Stored in a constant neural network.
As a deep learning neural network, convolutional neural networks are specifically designed to process structured array data such as images. The superior feature extraction capabilities of convolutional neural networks make them a major choice for many vision applications. CNN consists of a convolutional layer, a pooling layer, and a fully connected layer.
The convolution layer is a two-dimensional convolution neural network, the convolution layer repeatedly acts on the whole receptive field by using a plurality of convolution kernels, and all input data are traversed to realize feature extraction. The pooling layer, i.e. the sampling layer, is typically used to reduce the spatial resolution of the feature map and to preserve the most important feature information to prevent overfitting during training.
The full-connection layer is an output layer of the CNN, each node is connected with all nodes of the upper layer and is responsible for connecting all neurons output by the feature map of the upper layer to each node, and finally, the output of one interconnected neuron is obtained. The full connection layer is used as a classifier.
The bearing rolling system is a dynamic system which runs stably, and vibration signals of the system can be acquired through the sensor. When the rolling body, the inner ring or the outer ring of the bearing are damaged due to overload, friction, corrosion and other reasons, the vibration signal of the stabilizing system is changed. The vibration signal generated by both the normal bearing rolling system and the system after failure can be referred to as a time series signal. The method of the application effectively saves and distinguishes the dynamic information of the rolling system of the bearing when the nature of fault detection is carried out, and carries out rapid identification and classification aiming at the fault type to be tested, and finally obtains the fault diagnosis result. A flow chart of the method of this embodiment is shown in fig. 1.
The main technical problems to be solved in the embodiment are as follows: how to mine hidden features in the rolling bearing signals from the angle of dynamic feature extraction, and complete fault detection tasks by combining the dynamic feature information with a convolutional neural network. Specifically, the method of the embodiment firstly converts a bearing vibration time sequence signal with one dimension into a two-dimensional vibration time sequence signal through a high-gain observer, then utilizes the inherent dynamics information of a determined learning mechanism modeling time sequence signal to visualize the obtained dynamics information into a dynamics information graph, finally utilizes a convolutional neural network to conduct depth feature extraction, and utilizes the extracted features to judge whether the rolling bearing is in a normal state or a certain fault state.
The specific technical scheme of the embodiment is realized by the following steps:
(1) Collecting fault vibration signals: the method comprises the steps of acquiring one-dimensional vibration signals of a rolling bearing system in a time domain through an acceleration vibration sensor, determining the frequency of the signals acquired by the sensor, and then cutting the acquired vibration signals in equal length to construct a fault data set.
The method comprises the steps of collecting rolling bearing vibration signals of different fault types through a sensor, recording bearing fault labels, cutting the collected vibration signals to be equal in length, and constructing a data set of the rolling bearing vibration signals by taking each sample as one sample.
The sampling frequency is 12kHz, other sampling frequencies can be adopted, and the modeling accuracy of a learning mechanism is determined to be higher as the sampling time is smaller.
(2) The one-dimensional samples are spread into two-dimensional data samples using a high gain observer.
And observing the internal state of the bottom layer of the vibration system by using a high-gain observer (HGO) from the discrete one-dimensional vibration data acquired by the digital sensor to obtain two-dimensional data.
The purpose of using the high-gain observer is to observe the hidden internal dynamic state of the one-dimensional vibration signal, and take the observed signal as the second state of the bearing vibration system, different high-gain parameters influence the observed result, so that the accuracy of modeling dynamics information is further influenced, and proper parameters need to be selected.
(3) Data standardization processing: the data is normalized and each dimension of the sample is maximum normalized.
And (3) carrying out maximum value standardization processing on each dimension of the data in the step (2). The normalization function is:
wherein ,xi Representing a one-dimensional time sequence of ith samples, x max Representing the maximum value of all samples in the j dimension for the vibration data set,representing the normalized sample.
(4) Establishing a dynamic RBF neural network model of a rolling bearing vibration signal: the inherent dynamics of the rolling bearing vibration signal are modeled using a deterministic learning mechanism. The learning mechanism is determined to model the bearing vibration time sequence signal, and the dynamic characteristic of the vibration signal changing along with time can be extracted.
The training flow diagram of the RBF neural network is shown in fig. 2. Wherein x is 1 ,x 2 Respectively representing two states of the bearing vibration signal.Is the estimated state of the RBF neural network, i.e., the output of the neural network.
And (3) learning and training vibration signals under normal conditions and different faults of the bearing by adopting a radial basis function (RBF NN), wherein a Gaussian function is selected as the radial basis function. The training adopts a learning method according to a determined learning mechanism to realize that the weight of the dynamic RBF neural network converges to the optimal value; there are two cases of weight convergence of RBF neural networks: one is that the neuron of RBF neural network along the vibration signal track after expanding the dimension meets the condition of continuous excitation, its weight converges to the optimal value; the other is that the neurons of the RBF neural network far away from the track are not excited and the parameters are not updated, and the initial weight is still almost zero.
The RBF neural network approximation of the internal dynamics of each bearing vibration time series signal sample is an approximation of the internal dynamics along the vicinity of the vibration signal track, and the dynamics away from the track are not approximated. And saving the average value of each weight value in a period of time after the weight value of the dynamic RBF neural network is converged to generate a constant RBF neural network, so as to provide required knowledge/model for generating the dynamic information graph.
The bearing vibration signal generated by the rolling system can be expressed as a nonlinear power system described by a normal differential equation;
wherein x= [ x ] 1 ,x 2 ] T ∈R 2 Is the state of the nonlinear power system, p represents the parameter vector of the nonlinear power system, F (x; p) = [ F ] 1 (x；p),f 2 (x；p)]Is a non-formThe unknown dynamics of the linear motor system.
φ(X)＝[x(t 0 ),x(t 0 +T),…,x(t 0 +(N-1)T)]Is a sampling time sequence obtained from the nonlinear dynamics system (2) through a sampling period T, namely a real vibration signal obtained by a sensor. Selecting a sampling time of suitable size, the sampled data can be represented by the following euler sampling model:
x i (k+1)＝x i (k)+Tf i (x(k)；p) (3)
wherein x (k) = [ x ] 1 (k),x 2 (k)]∈R 2 Is the system state of the Euler model, f i (x (k); p) is the internal dynamics of the sampled data, modeled by the RBF neural network, the network parameters are accurate and converging, and T is the sampling time.
The RBF neural network is constructed as follows:
where W is the weight vector of the network, Z is the network input vector, i.e. the state of the sampling model (3), and M is the number of neurons. S (Z) = [ S ] 1 (||Z-ξ 1 ||),…,s M (||Z-ξ M ||)]w is the regression vector of the RBF,is a Gaussian radial basis function, eta is the width of the receiving domain, and zeta i =1, …, M is the neuron center of the RBF neural network.
The dynamic neural network identifier is constructed as follows:
wherein ,is two state vectors of the identifier, which correspond to two states of the vibration signal respectively. 0<α 1 ,α 2 <1 represents the identifier gain, < >>Weight vector representing vibration signal state 1RBF neural network to be estimated, < ->And a weight vector to be estimated of the RBF neural network representing the vibration signal state 2, wherein T is sampling time.
The weight updating strategy of the RBF neural network updates the weight according to the weight updating rate designed by the Lyapunov stability, and is described as follows:
wherein, gamma is the learning gain parameter, e 1 ,e 2 Respectively representing tracking errors of the two states.
According to a deterministic learning mechanism, which is capable of locally accurately approximating/modeling an unknown dynamic function f, a regression vector S (x (k)) with regression trajectory input satisfies a partially continuous excitation condition i (x (k); p) and stored in a constant RBF neural network
wherein ,i=1, 2 represents the weighted value of RBF neural network after convergence at Tk a To Tk b Average weight in time interval, E i Representing the approximation error.
The established constant RBF neural network model is the dynamic information of the vibration signal. I.e. in the above formulaAs the dynamics information of the vibration signal, the dynamics information map of the present embodiment is essentially a graphical representation of modeling dynamics.
The learning and training process described in the step (4) is a knowledge acquisition process, and the learned knowledge is stored in a constant RBF neural network weight vector mode. The weight vector of each group of neural networks represents the intrinsic dynamics of one sample, and is used as an expression for generating a dynamic information graph corresponding to the vibration signal sample.
(5) Generating a kinetic information graph: after maximum normalization and after learning the intrinsic dynamics of the vibration data using deterministic learning, a constant RBF network is obtained that captures the intrinsic dynamics along the state trajectory, namely: representing status trace +.>Is a neighborhood of (c). Since maximum normalization is performed, we have +.>I.e. the neighborhood of the state trace is limited to [ -1,1]×[-1,1]In a rectangular area therebetween. Thus, even though the track area of each sample +.>Different, we can be in +>Comparing the modeled kinetic information of each sample, i.e. +.> Track area of each sample->Different, mainly reflected in the different vibration signal amplitude ranges of different fault types collected by the sensor.
At a given resolution, the kinetic information may form an m x m matrix,
wherein ,zi (1),…,z i (m), i=1, 2 means the interval [ -1,1 []Evenly dividing into m-1 parts of corresponding data points. The matrix is further converted into a gray scale map, i.e. a kinetic information map for each sample is formed. The higher the resolution required, the larger the value of m can be chosen.
The process of generating the kinetic information map in step (5) is a process of determining knowledge reuse learned by learning mechanisms. The generated dynamic information graph is a numerical matrix with the size of m multiplied by m, the value of m cannot be too small, the smaller m can lead to fewer data points for generating the dynamic information graph and cannot achieve the effect of dynamic information reproduction, the larger the value of m is, the better the larger the value of m is, the higher the resolution of the generated dynamic information graph is, the requirement on calculation power is increased, and because a constant RBF neural network for generating dynamic information is learned from discrete vibration signal samples through a definite learning mechanism, the size suggestion of the value of m is not more than the data length of a single sample, and the significance of the data length of a sample on the improvement of fault detection precision is not great.
The specific process of generating the dynamic diagram is as follows:
determining learning modeling unknown dynamics: after modeling the intrinsic dynamics information of the vibration data using deterministic learning, a constant RBF neural network along the state trajectory is obtained,
i.e. Representing status trace +.>Is a neighborhood of (c).
Gridding input: since the maximum value normalization is performed, there areI.e. the neighborhood of the state trace is limited to [ -1,1]×[-1,1]In a rectangular area therebetween. Thus, even the trace area of each sampleDifferent, can be +.>Comparing the modeled kinetic information of each sample, i.e.Two-dimensional interval [ -1,1]×[-1,1]Evenly divided into (m-1) x (m-1) parts, corresponding to m x m data points, these data points will be used as input data for generating the kinetic information map.
Gridding output: putting the gridded m×m data points into the built modelThe output matrix is the dynamic information graph.
(6) Building a convolutional neural network: in the convolutional neural network model provided by the application, the convolutional layer and the pooling layer form a sub-module, and an ELU activation function and a batch normalization layer are used between the convolutional layer and the pooling layer to improve training efficiency and enhance generalization performance of the network. This sub-module is repeated four times, followed by a flattening layer, and then two fully connected layers are used to complete the classification task.
And constructing a convolutional neural network by using a PyTorch framework, wherein the network structure is a convolutional layer- > pooling layer- > flattening layer- > full-connection layer. And (3) using ELU activation functions and batch normalization processing between the convolution layer and the pooling layer, flattening the finally extracted feature map and inputting the feature map into a full-connection layer, wherein the output of the full-connection layer is used as the output of the whole convolution neural network, and a network structure diagram is shown in figure 3.
The size of the convolution kernel in each convolution layer is 9×9,5×5,3×3, and the size of the pooling layer kernel is 2×2.
To achieve a feature map output of a specified size requires the use of a fill operation, i.e. filling 0 at the feature map edges, the pooling layer maintains feature invariance, downsamples the feature map, expands the receptive field and prevents overfitting to some extent. The selection of the ELU activation function introduces nonlinearities into the network.
(7) Training convolutional neural networks:
dividing the data set: for the generated vibration signal dynamics information graph, the training set and the testing set are divided according to the proportion of 7:3.
Firstly, random seed setting of a training program is carried out, and the result of each training is ensured to be reproducible. Selecting proper batch size, inputting a dynamic information graph of a vibration signal of a training set bearing into a convolutional neural network, adopting cross entropy loss for a loss function, selecting a gradient descent algorithm to optimize and update network parameters, setting the dynamic learning rate to be 0.001 initially, using a learning rate scheduler ReduceLROnPlateau, gradually reducing the learning rate along with the training, better converging to obtain a model with higher performance until the loss function converges, selecting the network parameters with the minimum cross entropy loss in the training process, and storing to obtain the trained convolutional neural network.
The learning rate is one of the most important super parameters affecting the performance of the model, the learning scheduler is used for conditionally reducing the learning rate in the training process and improving the performance of the model, after each batch of training is completed, the learning scheduler detects whether the performance of the model is improved, if the performance of the model is not improved all the time when the set batch is reached, the learning rate is reduced, and after a cold period, the process is repeated until the last batch of training is completed.
(8) Classification of faulty bearing signals:
inputting a dynamic information graph of a bearing vibration signal to be tested into a trained convolutional neural network to obtain a network fault detection result, and comparing the obtained result with a fault type label to know whether the vibration signal is normal or in a certain fault state.
Example two
The embodiment provides a bearing fault detection system based on a definite learning and convolution neural network, which comprises the following components:
the data acquisition module is configured to acquire a one-dimensional bearing vibration signal of a bearing fault, and expand the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using the high-gain observer;
the standardized processing module is configured to perform maximum value standardized processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
the dynamic information acquisition module is configured to acquire dynamic information of the normalized bearing vibration signal by utilizing a pre-trained dynamic radial basis function neural network model based on a determined learning mechanism, and generate a dynamic information graph of the bearing vibration signal;
the bearing fault detection module is configured to detect faults in a pre-trained convolutional neural network model based on a dynamic information graph of the bearing vibration signal to obtain a bearing fault detection result.
The above modules are the same as examples and application scenarios implemented by the corresponding steps, but are not limited to what is disclosed in the first embodiment. It should be noted that the modules described above may be implemented as part of a system in a computer system, such as a set of computer-executable instructions.
The foregoing embodiments are directed to various embodiments, and details of one embodiment may be found in the related description of another embodiment.
The proposed system may be implemented in other ways. For example, the system embodiments described above are merely illustrative, such as the division of the modules described above, are merely a logical function division, and may be implemented in other manners, such as multiple modules may be combined or integrated into another system, or some features may be omitted, or not performed.
Example III
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps in the bearing failure detection method based on determining learning and convolutional neural networks as described in the above embodiment one.
Example IV
The present embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the program to implement the steps in the method for detecting bearing failure based on determining learning and convolutional neural network according to the above embodiment.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of a hardware embodiment, a 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, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), or the like.
While the foregoing description of the embodiments of the present application has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the application, but rather, it is intended to cover all modifications or variations within the scope of the application as defined by the claims of the present application.
Claims (10)
1. The bearing fault detection method based on the determined learning and convolution neural network is characterized by comprising the following steps of:
acquiring a one-dimensional bearing vibration signal of a bearing fault, and expanding the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using a high-gain observer;
carrying out maximum value standardization processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
modeling internal dynamic information of the bearing vibration signal based on a determined learning mechanism, and generating a dynamic information graph of the bearing vibration signal according to the internal dynamic information of the bearing vibration signal;
based on a dynamic information graph of the bearing vibration signal, fault detection is carried out in a pre-trained convolutional neural network model, and a bearing fault detection result is obtained.
2. The method for detecting bearing faults based on the deterministic learning and convolutional neural network according to claim 1, wherein the maximum value normalization processing is performed on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal, specifically:
carrying out maximum value standardization processing on each dimension in the two-dimensional bearing vibration signal;
the normalization function is:
wherein ,xi Representing a one-dimensional time series of the ith sample,representing the maximum value of all samples of the vibration data set in the j-dimension,/->Representing the normalized sample.
3. The method for detecting bearing faults based on a deterministic learning and convolutional neural network according to claim 1, wherein the method for generating a dynamic information map of a bearing vibration signal according to internal dynamic information of the bearing vibration signal by modeling the internal dynamic information of the bearing vibration signal based on a deterministic learning mechanism is specifically as follows:
modeling the bearing vibration signal by adopting a definite learning mechanism to obtain a dynamic information model of the bearing vibration signal;
gridding the track area of the normalized bearing vibration signal to obtain gridded dynamic information input;
based on the gridding dynamic information input, a dynamic information graph of the bearing vibration signal is obtained by utilizing a dynamic information model of the bearing signal.
4. The method for detecting bearing faults based on a deterministic learning and convolution neural network as claimed in claim 3, wherein the dynamic information model of the bearing vibration signal is specifically:
wherein ,representing status trace +.>Neighborhood of->Indicating that RBF neural network weight is converged and then is positioned at Tk a To Tk b Average weights in time interval, i=1, 2, t is sampling time, +.>Representing the neural network weights to be estimated, S (z) represents the regression vector of radial basis function components.
5. The method for detecting bearing faults based on a deterministic learning and convolution neural network as claimed in claim 3, wherein the dynamic information graph of the bearing vibration signal is an m x m matrix, specifically:
wherein ,zi (1),…,z i (m), i=1, 2 means the interval [ -1,1 []Evenly dividing the data points into m-1 parts of corresponding data points;
the matrix is converted into a gray scale map, i.e., a dynamic information map for each bearing vibration signal is formed.
6. The bearing fault detection method based on the deterministic learning and convolutional neural network as set forth in claim 1, wherein the training process of the convolutional neural network model is specifically as follows:
aiming at the generated vibration signal dynamic information graph, the training set and the testing set are divided according to the proportion of 7:3;
inputting a training set bearing vibration signal dynamic information graph into a convolutional neural network to detect bearing faults;
the loss function adopts cross entropy loss, and a gradient descent algorithm is selected to optimize and update network parameters, so that a trained convolutional neural network model is obtained;
and inputting the dynamic information graph of the vibration signal of the bearing by using the test set into a trained convolutional neural network model to obtain a network fault detection result.
7. The method for detecting bearing faults based on a deterministic learning and convolutional neural network as claimed in claim 6, wherein the structure of the convolutional neural network model comprises a convolutional layer, a pooling layer, a flattening layer, a fully connected layer and a fully connected layer which are connected in sequence;
and (3) using ELU activation functions and batch normalization processing between the convolution layer and the pooling layer, flattening and inputting the finally extracted feature map into a full-connection layer, and taking the output of the full-connection layer as the output of the whole convolution neural network.
8. A bearing failure detection system based on a deterministic learning and convolutional neural network, comprising:
the data acquisition module is configured to acquire a one-dimensional bearing vibration signal of a bearing fault, and expand the one-dimensional bearing vibration signal into a two-dimensional bearing vibration signal by using the high-gain observer;
the standardized processing module is configured to perform maximum value standardized processing on the two-dimensional bearing vibration signal to obtain a normalized bearing vibration signal;
a dynamics information acquisition module configured to model internal dynamics information of the bearing vibration signal based on the determined learning mechanism, and generate a dynamics information map of the bearing vibration signal from the internal dynamics information of the bearing vibration signal;
the bearing fault detection module is configured to perform fault detection by utilizing a pre-trained convolutional neural network model based on a dynamic information graph of the bearing vibration signal to obtain a bearing fault detection result.
9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements the steps in the method for bearing failure detection based on a deterministic learning and convolutional neural network according to any one of claims 1-7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps in the method for bearing failure detection based on a deterministic learning and convolutional neural network as in any one of claims 1-7 when the program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310647151.5A CN116659865A (en) | 2023-05-31 | 2023-05-31 | Bearing fault detection method and system based on definite learning and convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310647151.5A CN116659865A (en) | 2023-05-31 | 2023-05-31 | Bearing fault detection method and system based on definite learning and convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116659865A true CN116659865A (en) | 2023-08-29 |
Family
ID=87709299
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310647151.5A Pending CN116659865A (en) | 2023-05-31 | 2023-05-31 | Bearing fault detection method and system based on definite learning and convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116659865A (en) |
-
2023
- 2023-05-31 CN CN202310647151.5A patent/CN116659865A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kanarachos et al. | Detecting anomalies in time series data via a deep learning algorithm combining wavelets, neural networks and Hilbert transform | |
| Cheng et al. | Machine health monitoring using adaptive kernel spectral clustering and deep long short-term memory recurrent neural networks | |
| CN108334936B (en) | Fault prediction method based on migration convolutional neural network | |
| Xia et al. | Multi-stage fault diagnosis framework for rolling bearing based on OHF Elman AdaBoost-Bagging algorithm | |
| CN112673381B (en) | Method and related device for identifying confrontation sample | |
| Wang et al. | A combination of residual and long–short-term memory networks for bearing fault diagnosis based on time-series model analysis | |
| CN114357594B (en) | Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU | |
| CN112633339A (en) | Bearing fault intelligent diagnosis method, bearing fault intelligent diagnosis system, computer equipment and medium | |
| CN113095370A (en) | Image recognition method and device, electronic equipment and storage medium | |
| CN112131907A (en) | Method and device for training classification model | |
| CN112527604A (en) | Deep learning-based operation and maintenance detection method and system, electronic equipment and medium | |
| CN112765890A (en) | Dynamic domain adaptive network-based multi-working-condition rotating machine residual life prediction method | |
| Islam et al. | Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal | |
| CN116453438B (en) | Display screen parameter detection method, device, equipment and storage medium | |
| CN114091554A (en) | Training set processing method and device | |
| CN111160526B (en) | Online testing method and device for deep learning system based on MAPE-D annular structure | |
| JP2019105871A (en) | Abnormality candidate extraction program, abnormality candidate extraction method and abnormality candidate extraction apparatus | |
| CN117036843A (en) | Target detection model training method, target detection method and device | |
| Rahadian et al. | Image encoding selection based on Pearson correlation coefficient for time series anomaly detection | |
| US20200012941A1 (en) | Method and system for generation of hybrid learning techniques | |
| Liu et al. | Deep learning approach for sensor data prediction and sensor fault diagnosis in wind turbine blade | |
| Castro-Cabrera et al. | Adaptive classification using incremental learning for seismic-volcanic signals with concept drift | |
| Chopra et al. | Classification of faults in damadics benchmark process control system using self organizing maps | |
| CN116659865A (en) | Bearing fault detection method and system based on definite learning and convolutional neural network | |
| US20210049462A1 (en) | Computer system and model training method |
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 |