CN113642465B - Bearing health assessment method based on relational network - Google Patents
Bearing health assessment method based on relational network Download PDFInfo
- Publication number
- CN113642465B CN113642465B CN202110931532.7A CN202110931532A CN113642465B CN 113642465 B CN113642465 B CN 113642465B CN 202110931532 A CN202110931532 A CN 202110931532A CN 113642465 B CN113642465 B CN 113642465B
- Authority
- CN
- China
- Prior art keywords
- bearing
- relationship
- layer
- score
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Signal Processing (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a bearing health assessment method based on a relationship network, and relates to the technical field of bearing health assessment methods. The method comprises the following steps: taking a 1 st sample of the initial running state of a bearing to be evaluated as a support set S, taking m samples of the early normal running state as a query set Q for training, testing a sample y of the current running state, and constructing health labels of the m samples in the query set Q; constructing a relationship network model for bearing health assessment; obtaining a relation network model parameter; calculating a relationship score between the current state sample of the bearing and the support set S by using the trained relationship network model, and then carrying out Savitzky-Golay smooth filtering on the relationship score to obtain a bearing state health score; the method can be used for carrying out health assessment on the bearing under the condition of unknown fault data, and can accurately reflect the health condition of the bearing.
Description
Technical Field
The invention relates to the technical field of signal processing methods, in particular to a bearing health assessment method based on a relational network.
Background
The bearing is one of indispensable parts in mechanical equipment and is also one of parts which are very prone to failure in the mechanical equipment. If the bearing breaks down, the bearing causes economic loss if the bearing breaks down, and threatens personal safety if the bearing breaks down. In order to reduce various hazards caused by bearing faults, the abnormal condition can be timely and accurately detected and the bearing can be replaced by a method for carrying out health assessment on the bearing before the bearing is just slightly in fault and any loss is not caused.
Disclosure of Invention
The invention aims to solve the technical problem of how to provide a bearing health assessment method which is higher in bearing health assessment accuracy and is based on a relational network.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows: a bearing health assessment method based on a relation network is characterized by comprising the following steps:
taking a 1 st sample of the initial running state of a bearing to be evaluated as a support set S, taking m samples of the early normal running state as a query set Q for training, testing a sample y of the current running state, and constructing health labels of the m samples in the query set Q;
constructing a bearing health assessment relation network model;
inputting the frequency spectrum of the query set Q and the frequency spectrum of the support set S into a relational network model, taking the mean square error between a real health label and a prediction label as a loss function value, and training by using a BP back propagation algorithm to obtain relational network model parameters;
and calculating a relationship score between the current running state sample y of the bearing and the support set S by using the trained relationship network model, and then carrying out Savitzky-Golay smooth filtering on the relationship score to obtain a bearing state health score.
Adopt the produced beneficial effect of above-mentioned technical scheme to lie in: the invention takes the frequency spectrum of the vibration signal obtained by fast Fourier transform as the input of the relation network, can reflect the concentrated position and distribution characteristics of energy in the vibration signal, and has strong information capture capability and characterization capability. Therefore, the invention takes the frequency spectrum of the vibration signal as input, is more favorable for representing and extracting the characteristics and can more accurately represent the characteristic information in the vibration signal.
The method applies the relational network to the bearing health assessment problem, and the relational network can carry out health assessment on the unknown running state sample by utilizing the distance or the similarity between the normal running state sample and the unknown running state sample under the condition that only the bearing runs in a normal running state sample, and obtains a good assessment result. The samples of the bearing in the normal running state in the real society are easy to obtain, so that the application prospect of health assessment by utilizing the relation network is greatly improved. And the relation network is different from the traditional fixed measurement mode such as Euclidean measurement, cosine measurement and the like, and the relation network obtains an efficient nonlinear distance measurement through network training, so that the distance or similarity between samples calculated by the relation network is more accurate. Therefore, the distance or the similarity between the normal running state sample of the bearing to be evaluated and the initial running sample of the bearing to be evaluated is learned through the relational network model, and then the trained relational network model is used for calculating the distance or the similarity between the current running state sample of the bearing to be evaluated and the initial running sample of the bearing to be evaluated, so that the health score of the current running state sample of the bearing to be evaluated can be obtained. The nonlinear distance measurement method in the relational network model enables the obtained health score to be more accurate.
And smoothing and denoising the obtained health score by adopting a Savitzky-Golay filtering smoothing method to achieve the effects of removing high-frequency noise points and smoothing a data sequence, so that the final bearing health evaluation result is smoother, and the health condition of the bearing can be reflected more accurately.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
FIG. 1 is a flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a time domain signal diagram of a bearing to be evaluated according to an embodiment of the present invention;
FIG. 3 is a frequency spectrum of a sample of a bearing support set to be evaluated in an embodiment of the present invention;
FIG. 4 is a diagram of a model of a method according to an embodiment of the invention;
FIG. 5 is a schematic structural diagram of a feature extraction module according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a relationship module according to an embodiment of the present invention;
FIG. 7 is a bearing health assessment index graph after smooth filtering based on Savitzky-Golay according to an embodiment of the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
As shown in fig. 1, the invention discloses a bearing health assessment method based on a relationship network, which specifically comprises the following steps:
step 1): constructing a support set S and a query set Q by using bearing data to be evaluated, constructing a health label of a sample in the query set Q, and then performing fast Fourier transform on a vibration signal to obtain a frequency spectrum;
in the embodiment, a group of bearing full-life experimental data in the NASA prediction database is used as bearing data to be evaluated, and 984 groups of data running from a healthy state to a failure state are collected. In this example, 2048 sampling points are selected from each vibration signal segment to obtain 984 samples, each sample is 1 × 2048, as shown in fig. 2, and fig. 2 shows a time domain signal diagram of the bearing to be evaluated. In this example, the 1 st sample of the initial operating state of the bearing to be evaluated is taken as a support set S, 200 samples of the early normal operating state are taken as a query set Q for training, a sample y of the current operating state is used for testing, and a health label "1" is added to 200 samples in the query set Q.
Due to the frequency spectrum signal of the original vibration signal of the bearing, the characteristics of the bearing during operation can be better reflected. The present example uses a fast fourier transform to obtain a sample spectrum of 1 x 1024. As shown in fig. 3, fig. 3 shows a frequency spectrum of a bearing support set sample. Finally, the spectrum of the support set S is obtained: f(s); query set Q spectrum: f (q)1)，F(q2)，F(q3)，...，F(qm) (ii) a Spectrum of current operating state sample y: f (y).
The calculation formula of the fast Fourier transform is as follows:
in the formula: f (k) represents the frequency spectrum of the signal x (N), x (N) represents the bearing vibration signal, and N represents the number of sampling points of the vibration signal; wNIs a twiddle factor.
Step 2): constructing a bearing health assessment relation network model;
fig. 4 shows a model diagram based on a relational network, where the relational network model includes an embedding module and a relational module, where the embedding module extracts the features of the frequency spectrums in the support set and the query set, splices the feature vectors in the support set and the feature vectors in the query set, and inputs the spliced feature vectors to the relational module, and the relational module calculates a relational score between two connected features, where the higher the score is, the more similar the two feature vectors are. In the relational network model, a support set frequency spectrum and a query set frequency spectrum are used as input, a support set sample is a sample in a normal state when a bearing initially operates, a query sample is a sample in an unknown operation state, and a relation score represents the distance or similarity between the support set and the query sample, so that the higher the output relation score is, the healthier the query sample state is.
Fig. 5 shows a schematic structure diagram of the embedded module, and fig. 6 shows a schematic structure diagram of the relationship module. The embedding module includes 4 convolution modules and 3 pooling layers. The relationship module includes two convolution modules, two pooling layers, and two fully-connected layers. Wherein each layer of convolution module is composed of a one-dimensional convolution layer, a batch normalization layer and an activation function
The built embedded module has 7 layers, wherein the 1 st layer, the 3 rd layer, the 5 th layer and the 6 th layer are convolution modules; the 2 nd layer and the 4 th layer are maximum pooling layers; layer 7 is the adaptive max pooling layer.
The constructed relation modules have 6 layers, wherein the 1 st layer and the 3 rd layer are convolution modules; the 2 nd layer and the 4 th layer are maximum pooling layers; the 5 th layer and the 6 th layer are all connecting layers. The number of neurons in the 5 th fully-connected layer is 1 × 8, and the number of neurons in the 6 th fully-connected layer is 1 × 1.
In the relational network model, the convolution kernel size of the 1 st layer convolution of the division feature extraction module is 1 x 10, and the convolution kernel sizes of the other convolutions are all 1 x 3. The adaptive maximum pooling layer size was 1 × 25, and the maximum pooling layer size was 1 × 2.
And the 6 th layer of the full connection layer of the relation module in the activation function is a sigmoid function, and the rest are relu functions.
In some embodiments, the input size of the relationship network is 1 × 1024 and the output size is 1 × 1. The setting of its network parameters may be as shown in table 1.
TABLE 1 network parameters
Step 3): inputting the frequency spectrum in the query set B and the frequency spectrum of the support set into a constructed relational network model, taking the mean square error between the health label and the prediction label as a loss function value, and training by using a BP back propagation algorithm to obtain model parameters;
the loss function calculation formula is as follows:
wherein, the ratio of phi,
representing the parameters of the embedding module f and the relation module g, respectively.
In this example, the training times are set to 200 times, the Batch-Size is set to 32 times, the loss function of the network is the mean square error loss function, an Adam optimizer is adopted, and the learning rate is 0.001. The loss value of the training process of the network is gradually reduced along with the iteration, and the network model reaches a stable state.
In step 3), the training process of the relational network model is as follows:
(1) the frequency spectrum F (S) in the support set S and the frequency spectrum F (Q) in the query set Q are compared1)，F(q2)，F(q3)，...，F(qm) The input embedding module f can obtain the feature representation f (S) of the support set S and the feature representation f (Q) in the query set Qi)。
(2) Sequentially comparing the feature representation f (S) of the support set S with the m feature representations f (Q) in the query set Qi) Splicing together to obtain m characteristic spliced samples Z (f(s), f (q)i))。
(3) Splicing m characteristics Z (f(s), f (q) after splicingi) Input into a relationship module g, and m relationship scores r are generated by the relationship modulei：
ri＝g(Z(f(s),f(qi))
In the formula: r isiRepresenting the relationship score, i.e. distance or similarity, r, between the samples in the support set S and the query set QiIs in the range of 0 to1, g represents a relation module for calculating a relation score between two feature representations, f represents an embedding module for extracting feature representations, and Z represents f(s) and f (q)i) And performing characteristic splicing.
(4) Real health label '1' and prediction label r in query set QiThe mean square error between the two is used as a loss function, and the BP back propagation algorithm is used for training until the model is converged to obtain model parameters.
Step 4): calculating a relationship score between the current running state sample of the bearing and the support set S by using the trained relationship network model, and carrying out Savitzky-Golay smooth filtering on the relationship score to obtain a bearing state health score;
in the step 4), the specific process of calculating the relationship score between the sample of the current running state of the bearing and the support set S by using the trained relationship network model comprises the following steps:
(1) inputting the frequency spectrum F (S) of the support set S and the frequency spectrum F (y) of the current running state sample of the bearing into a trained embedding module f, and obtaining the characteristic representation f (S) of the support set S and the characteristic representation f (y) of the current running state sample;
(2) splicing the feature representation f (S) of the support set S and the feature f (y) of the current operation state together to obtain a sample Z (f (S), f (y)) after feature splicing;
(3) inputting the spliced characteristics Z (f(s), f (y)) into a trained relation module g, and obtaining a relation score r of the characteristic representation f (y) of the current operating state through the relation module gj：
rj＝g(Z(f(s),f(y))
In the formula: r isjRepresenting the score of the relationship, i.e. distance or similarity, r, between the support set S and the current operating state sample yjThe value range of (a) is 0-1, g represents a relationship module and is used for calculating a relationship score between two feature representations, f represents an embedding module and is used for extracting feature representations, and Z represents that f(s) and f (y) are subjected to feature splicing.
In this example, 784 relationship scores are obtained, each score corresponding to a relationship score between the sample y to be measured and the support set S. The closer the relationship score is to "1", the healthier the bearing is, and therefore the purpose of bearing health assessment is achieved.
In the step 4), the method for obtaining the bearing state health score by carrying out Savitzky-Golay smoothing filtering on the relationship score comprises the following steps:
(1) let the health score data obtained by the relational network model be riI ═ m, a., 0, a., m, with an n-th order fit polynomial of
In the formula: b is the coefficient of a polynomial; k is the order of the polynomial; i is a weight of a polynomial; z is a polynomial function of order n
(2) Relationship score riMean square error epsilon with fitted polynomialDIs composed of
(3) To minimize the mean square error, then εDThe derivatives of each coefficient should all be 0, i.e.
The formula (2) is taken into the formula (3) and is unfolded to obtain the product
By solving the above formula, the coefficients b of the polynomial can be obtainedkThus, a fitted n-term polynomial Z (n) is obtained, and the bearing state health score after smoothing can be obtained.
In an embodiment, the window length is set to 199, the order of the polynomial fit is 3, and 784 relationship scores r are obtainedkAnd carrying out Savitzky-Golay smooth filtering to obtain the bearing state health score. As shown in FIG. 7, FIG. 7 shows Savitzky-Golay smooth filtered bearing state health score.
Claims (9)
1. A bearing health assessment method based on a relation network is characterized by comprising the following steps:
taking a 1 st sample of the initial running state of a bearing to be evaluated as a support set S, taking m samples of the early normal running state as a query set Q for training, testing a sample y of the current running state, and constructing health labels of the m samples in the query set Q;
constructing a bearing health assessment relation network model;
inputting the frequency spectrum of the query set Q and the frequency spectrum of the support set S into a relational network model, taking the mean square error between a real health label and a prediction label as a loss function value, and training by using a BP back propagation algorithm to obtain relational network model parameters;
and calculating a relationship score between the current running state sample y of the bearing and the support set S by using the trained relationship network model, and then carrying out Savitzky-Golay smooth filtering on the relationship score to obtain a bearing state health score.
2. The bearing health assessment method based on the relational network according to claim 1, wherein the method for constructing the health label of m samples in the query set Q is as follows:
firstly, acquiring original vibration signal segments of a bearing to be evaluated, wherein each vibration signal segment selects 2N sampling points as a sample; then, taking a first sample of the initial running state of the bearing to be evaluated as a support set S, taking m samples of the early normal running state as a query set Q for training, testing a sample y of the current running state, and marking a health label '1' on the m samples in the query set Q.
3. The relationship network based bearing health assessment method according to claim 1, wherein: obtaining the frequency spectrum of the support set S through a fast Fourier transform formula: f(s); query set Q spectrum: f (q)1)，F(q2)，F(q3)，...，F(qm) (ii) a And spectrum of current operating state sample y: f (y).
4. The relational network based bearing health assessment method according to claim 3, wherein the fast Fourier transform formula is as follows:
in the formula: f (k) represents the frequency spectrum of the signal x (N), x (N) represents the bearing vibration signal, and N represents the number of sampling points of the vibration signal; wNThe frequency spectrum of the vibration signal is obtained after fast Fourier transform, and the frequency spectrum length is N.
5. The relationship network based bearing health assessment method according to claim 1, wherein: the bearing health assessment relationship network model comprises: the system comprises an embedding module and a relation module, wherein the embedding module is used for extracting characteristic representation of an input spectrum, and the relation module is used for calculating a relation score between two spectrum characteristic representations;
the built embedded modules have 7 layers and comprise 4 convolution modules and 3 pooling layers, wherein the 1 st layer, the 3 rd layer, the 5 th layer and the 6 th layer are convolution modules; the 2 nd layer and the 4 th layer are maximum pooling layers; the 7 th layer is a self-adaptive maximum pooling layer;
the constructed relation modules have 6 layers, including 2 convolution modules, 2 pooling layers and 2 full-connection layers; wherein, the 1 st layer and the 3 rd layer are convolution modules; the 2 nd layer and the 4 th layer are maximum pooling layers; the 5 th layer and the 6 th layer are all connected layers. The number of the neurons of the 5 th fully-connected layer is 1 × 8, and the number of the neurons of the 6 th fully-connected layer is 1 × 1;
each convolution module comprises a one-dimensional convolution layer, a batch normalization layer and an activation function;
in the relational network model, the sizes of convolution kernels of the 1 st layer of convolution of the division feature extraction module are 1 x 10, and the sizes of convolution kernels of the other layers of convolution are 1 x 3; the size of the self-adaptive maximum pooling layer is 1 × 25, and the sizes of the maximum pooling layers are 1 × 2;
and the 6 th layer of the full connection layer of the relation module in the activation function is a sigmoid function, and the rest are relu functions.
6. The relationship network based bearing health assessment method according to claim 1, wherein said method of obtaining relationship network model parameters is as follows:
the frequency spectrum F (S) in the support set S and the frequency spectrum F (Q) in the query set Q are compared1)，F(q2)，F(q3)，...，F(qm) Inputting the data into an embedding module f to obtain a feature representation f (S) of a support set S and a feature representation f (Q) in a query set Qi)；
Sequentially comparing the feature representation f (S) of the support set S with the m feature representations f (Q) in the query set Qi) Splicing together to obtain m characteristic spliced samples Z (f(s), f (q)i))；
Splicing m characteristics Z (f(s), f (q) after splicingi) Input into a relationship module g, and m relationship scores r are generated by the relationship modulei：
ri＝g(Z(f(s),f(qi))
In the formula: r isiRepresenting the relationship score, i.e. distance or similarity, r, between the samples in the support set S and the query set QiThe value range of (a) is 0-1, g represents a relationship module and is used for calculating a relationship score between two feature representations, f represents an embedding module and is used for extracting feature representations, and Z represents f(s) and f (q) respectivelyi) Performing characteristic splicing;
real health label '1' and prediction label r in query set QiThe mean square error between the two is used as a loss function, and the BP back propagation algorithm is used for training until the model is converged to obtain model parameters.
7. The relationship network based bearing health assessment method according to claim 6, wherein: the loss function value calculation formula is as follows:
wherein, the ratio of phi,
representing the parameters of the embedding module f and the relation module g, respectively.
8. The bearing health assessment method based on the relationship network as claimed in claim 1, wherein the method for calculating the relationship score between the current operation state sample y of the bearing and the support set S by using the trained relationship network model is as follows:
inputting the frequency spectrum F (S) of the support set S and the frequency spectrum F (y) of the current running state sample of the bearing into a trained embedding module f to obtain a feature representation f (S) of the support set S and a feature representation f (y) of the current running state sample;
splicing the feature representation f (S) of the support set S and the feature f (y) of the current operation state together to obtain a sample Z (f (S), f (y)) after feature splicing;
inputting the spliced characteristics Z (f(s), f (y)) into a trained relation module g, and obtaining a relation score r of the characteristic representation f (y) of the current operating state through the relation module gj：
rj＝g(Z(f(s),f(y))
In the formula: r isjRepresenting the score of the relationship, i.e. distance or similarity, r, between the support set S and the current operating state sample yjThe value range of (a) is 0-1, g represents a relationship module and is used for calculating a relationship score between two feature representations, f represents an embedding module and is used for extracting feature representations, and Z represents that f(s) and f (y) are subjected to feature splicing.
9. The relationship network-based bearing health assessment method according to claim 1, wherein said Savitzky-Golay smoothing filtering of the relationship score to obtain the bearing state health score is as follows:
assume a relationship score obtained by a relationship network model as riI ═ m,., 0.., m, with an n-th order fit polynomial of:
in the formula: b is the coefficient of a polynomial; k is the order of the polynomial; i is a weight of a polynomial; z is a polynomial function of order n;
relationship score riMean square error e between fitted polynomialsDComprises the following steps:
to minimize the mean square error, then εDThe derivatives of each term coefficient should all be 0, i.e.:
general formula mean square error epsilonDAfter being unfolded, the following formula is obtained:
by solving the above formula, the coefficients b of the polynomial can be obtainedkThus, a fitted n-term polynomial Z (n) is obtained, and the bearing state health score after smoothing can be obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110931532.7A CN113642465B (en) | 2021-08-13 | 2021-08-13 | Bearing health assessment method based on relational network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110931532.7A CN113642465B (en) | 2021-08-13 | 2021-08-13 | Bearing health assessment method based on relational network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113642465A CN113642465A (en) | 2021-11-12 |
| CN113642465B true CN113642465B (en) | 2022-07-08 |
Family
ID=78421569
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110931532.7A Active CN113642465B (en) | 2021-08-13 | 2021-08-13 | Bearing health assessment method based on relational network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113642465B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114417931A (en) * | 2022-01-21 | 2022-04-29 | 石家庄铁道大学 | Bearing fault diagnosis method based on prototype network |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145971A (en) * | 2018-08-07 | 2019-01-04 | 桂林电子科技大学 | Based on the single sample learning method for improving matching network model |
| CN109685135A (en) * | 2018-12-21 | 2019-04-26 | 电子科技大学 | A kind of few sample image classification method based on modified metric learning |
| CN109961089A (en) * | 2019-02-26 | 2019-07-02 | 中山大学 | Small sample and zero sample image classification method based on metric learning and meta learning |
| CN110108456A (en) * | 2019-04-16 | 2019-08-09 | 东南大学 | A kind of rotating machinery health evaluating method of depth convolutional neural networks |
| CN111476292A (en) * | 2020-04-03 | 2020-07-31 | 北京全景德康医学影像诊断中心有限公司 | Small sample element learning training method for medical image classification processing artificial intelligence |
| CN111709448A (en) * | 2020-05-20 | 2020-09-25 | 西安交通大学 | Mechanical fault diagnosis method based on migration relation network |
| CN111914705A (en) * | 2020-07-20 | 2020-11-10 | 华中科技大学 | Signal generation method and device for improving health state evaluation accuracy of reactor |
| CN112016392A (en) * | 2020-07-17 | 2020-12-01 | 浙江理工大学 | Hyperspectral image-based small sample detection method for soybean pest damage degree |
| CN112257862A (en) * | 2020-09-30 | 2021-01-22 | 重庆大学 | Semi-supervised identification method based on relational network marker sample expansion |
| CN112929380A (en) * | 2021-02-22 | 2021-06-08 | 中国科学院信息工程研究所 | Trojan horse communication detection method and system combining meta-learning and spatiotemporal feature fusion |
| CN113221964A (en) * | 2021-04-22 | 2021-08-06 | 华南师范大学 | Single sample image classification method, system, computer device and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108039987B (en) * | 2017-12-19 | 2020-09-22 | 北京航空航天大学 | Key infrastructure vulnerability assessment method based on multilayer coupling relation network |
| CN109508388A (en) * | 2018-11-28 | 2019-03-22 | 交通银行股份有限公司 | A kind of method and apparatus of relational network visualization map |
-
2021
- 2021-08-13 CN CN202110931532.7A patent/CN113642465B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109145971A (en) * | 2018-08-07 | 2019-01-04 | 桂林电子科技大学 | Based on the single sample learning method for improving matching network model |
| CN109685135A (en) * | 2018-12-21 | 2019-04-26 | 电子科技大学 | A kind of few sample image classification method based on modified metric learning |
| CN109961089A (en) * | 2019-02-26 | 2019-07-02 | 中山大学 | Small sample and zero sample image classification method based on metric learning and meta learning |
| CN110108456A (en) * | 2019-04-16 | 2019-08-09 | 东南大学 | A kind of rotating machinery health evaluating method of depth convolutional neural networks |
| CN111476292A (en) * | 2020-04-03 | 2020-07-31 | 北京全景德康医学影像诊断中心有限公司 | Small sample element learning training method for medical image classification processing artificial intelligence |
| CN111709448A (en) * | 2020-05-20 | 2020-09-25 | 西安交通大学 | Mechanical fault diagnosis method based on migration relation network |
| CN112016392A (en) * | 2020-07-17 | 2020-12-01 | 浙江理工大学 | Hyperspectral image-based small sample detection method for soybean pest damage degree |
| CN111914705A (en) * | 2020-07-20 | 2020-11-10 | 华中科技大学 | Signal generation method and device for improving health state evaluation accuracy of reactor |
| CN112257862A (en) * | 2020-09-30 | 2021-01-22 | 重庆大学 | Semi-supervised identification method based on relational network marker sample expansion |
| CN112929380A (en) * | 2021-02-22 | 2021-06-08 | 中国科学院信息工程研究所 | Trojan horse communication detection method and system combining meta-learning and spatiotemporal feature fusion |
| CN113221964A (en) * | 2021-04-22 | 2021-08-06 | 华南师范大学 | Single sample image classification method, system, computer device and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| 《Learning to Compare: Relation Network for Few-Shot Learning》;Flood Sung等;《2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition》;20181231;第1199-1208页 * |
| 《深度嵌入关系空间下齿轮箱标记样本扩充及其半监督故障诊断方法》;吕枫等;《仪器仪表学报》;20210228;第42卷(第2期);第55-65页 * |
| 《融合强化学习和关系网络的样本分类》;张碧陶等;《计算机工程与应用》;20191130;第55卷(第21期);第189-196,253页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113642465A (en) | 2021-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108426713B (en) | Rolling bearing weak fault diagnosis method based on wavelet transformation and deep learning | |
| CN112881942B (en) | Abnormal current diagnosis method and system based on wavelet decomposition and empirical mode decomposition | |
| CN112326210A (en) | Large motor fault diagnosis method combining sound vibration signals with 1D-CNN | |
| CN105206270A (en) | Isolated digit speech recognition classification system and method combining principal component analysis (PCA) with restricted Boltzmann machine (RBM) | |
| CN108435819B (en) | Energy consumption abnormity detection method for aluminum profile extruder | |
| CN111914703A (en) | Mechanical rotating part fault diagnosis method based on wavelet transformation and transfer learning GoogLeNet | |
| CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
| CN113642465B (en) | Bearing health assessment method based on relational network | |
| CN113488060A (en) | Voiceprint recognition method and system based on variation information bottleneck | |
| CN112380762A (en) | Power transmission line short-circuit fault diagnosis method based on VMD-WOA-LSSVM | |
| CN112151067B (en) | Digital audio tampering passive detection method based on convolutional neural network | |
| CN111428772B (en) | Photovoltaic system depth anomaly detection method based on k-nearest neighbor adaptive voting | |
| CN114420151B (en) | Speech emotion recognition method based on parallel tensor decomposition convolutional neural network | |
| CN116935892A (en) | Industrial valve anomaly detection method based on audio key feature dynamic aggregation | |
| CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network | |
| CN113740671A (en) | Fault arc identification method based on VMD and ELM | |
| CN112347917A (en) | Gas turbine fault diagnosis method, system, equipment and storage medium | |
| Xiao et al. | Health assessment for piston pump using LSTM neural network | |
| CN116796187A (en) | Power transmission line partial discharge detection method | |
| CN108827905B (en) | near-infrared model online updating method based on local weighting L asso | |
| CN110908365A (en) | Unmanned aerial vehicle sensor fault diagnosis method and system and readable storage medium | |
| CN116361724A (en) | Mechanical equipment fault diagnosis method, device, equipment and readable storage medium | |
| CN109903181A (en) | Line loss prediction technique under compressed sensing based missing data collection | |
| Li et al. | A robust fault diagnosis method for rolling bearings based on deep convolutional neural network | |
| CN116127302A (en) | Electric vehicle charging device fault arc identification method based on improved AlexNet algorithm |
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 |