CN109186973A - A kind of mechanical failure diagnostic method of unsupervised deep learning network - Google Patents
A kind of mechanical failure diagnostic method of unsupervised deep learning network Download PDFInfo
- Publication number
- CN109186973A CN109186973A CN201810949099.8A CN201810949099A CN109186973A CN 109186973 A CN109186973 A CN 109186973A CN 201810949099 A CN201810949099 A CN 201810949099A CN 109186973 A CN109186973 A CN 109186973A
- Authority
- CN
- China
- Prior art keywords
- training sample
- fault
- norm
- training
- udln
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000002405 diagnostic procedure Methods 0.000 title abstract 2
- 238000012549 training Methods 0.000 claims abstract description 153
- 238000003745 diagnosis Methods 0.000 claims abstract description 50
- 238000000034 method Methods 0.000 claims abstract description 49
- 238000012360 testing method Methods 0.000 claims abstract description 22
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 238000001914 filtration Methods 0.000 claims abstract description 9
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000013016 damping Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 abstract 1
- 238000005457 optimization Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 239000002131 composite material Substances 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 230000001050 lubricating effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
- G01M7/02—Vibration-testing by means of a shake table
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
It the invention discloses a kind of mechanical failure diagnostic method of unsupervised deep learning network, comprises the following steps that (1) is installed corresponding sensor near the components such as the bearing of mechanical equipment and is acquired to vibration signal, obtains mechanical vibration signal;(2) hybrid domain fault signature data set is converted by the vibration signal of acquisition, and is divided into test and training sample character subset；(3) training sample character subset is input to study and training in constructed unsupervised deep learning network (UDLN) model, which is made of two layers of improved unsupervised feature extraction layer of sparseness filtering (L12SF) and one layer of weighted euclidean distance similar affine (WE-AP) cluster layer；(4) test sample is input to trained diagnostic model, realizes whole unsupervised feature learning and fault cluster.(5) its discrimination is calculated according to the subjection degree of test sample clustering, realizes fault identification and diagnosis.The invention is simple and feasible, and this method can carry out adaptive unsupervised fault diagnosis to all kinds of failures of mechanical equipment.
Description
Technical Field
The invention relates to the technical field of fault diagnosis in industrial production, in particular to a mechanical fault diagnosis method of an unsupervised deep learning network.
Background
At present, industrial equipment is being developed in a large-scale, high-speed, and automated manner. Mechanical equipment is the most common component in the field of industrial equipment and is widely applied to important fields such as aviation, aerospace, transportation, intelligent manufacturing and the like. At present, the parts such as bearings are the most important power transmission and support parts of the machine, and according to statistics, 30% of mechanical faults are caused by local damage or defects of the parts such as the bearings, so that effective state monitoring and fault diagnosis are required to be carried out on mechanical equipment and key parts thereof.
Mechanical equipment such as cranes, rotating machinery and the like constantly generate a large amount of real-time information reflecting the operating conditions of the equipment, and the information comprises various physical quantities such as vibration, sound, temperature and the like. Once the mechanical equipment is abnormal, the corresponding physical information change is brought. In order to describe the operation condition of the mechanical equipment from different angles, more mechanical information needs to be collected, and effective intelligent monitoring and diagnosis are carried out on the mechanical equipment through data or information-based analysis. Therefore, the purpose of fault diagnosis is to realize effective diagnosis and analysis of equipment through the acquired mechanical information so as to reduce or reduce loss or harm caused by the fault of the mechanical equipment.
Because the equipment cluster for monitoring and diagnosing the mechanical equipment is large in scale, multiple in measuring points, high in sampling frequency and long in service life, the monitoring and diagnosing system acquires massive data, so that the field of mechanical health monitoring and management enters a big data era. However, in an actual industrial field, the phenomenon of "mechanical big data" also brings about a problem that the label information of the mechanical equipment is difficult to obtain. This is because the operating time of the mechanical equipment is much longer than the time when the fault occurs, sparsity of mechanical data is inevitable, and it is difficult to manually mark information. For this reason, the lack of tag information of the machine data and the sparsity of the data information present a series of challenges for fault diagnosis. Therefore, effective condition monitoring and fault identification of mechanical equipment in the limited or lack of training sample label information is a key issue which is always pending in the field of fault diagnosis.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a mechanical fault diagnosis method of an unsupervised deep learning network, aims to realize the whole-process unsupervised fault diagnosis of mechanical equipment from a characteristic extraction stage to a mode identification stage, and can provide a solvable effective scheme for the mechanical fault diagnosis under the condition of class label deficiency.
The invention adopts the following technical scheme for solving the technical problems:
the invention provides a mechanical fault diagnosis method of an unsupervised deep learning network, which comprises the following steps:
step 1, pre-selecting a tested part on mechanical equipment, and collecting a vibration signal of the tested part on the mechanical equipment;
step 2, converting the acquired vibration signals into a mixed domain fault feature data set, and dividing the mixed domain fault feature data set into a test sample feature subset and a training sample feature subset, wherein the test sample feature subset is used as a test sample, and the training sample feature subset is used as a training sample;
step 3, initializing parameters of an unsupervised deep learning network UDLN model, inputting training samples into the UDLN model for pre-training, and obtaining parameters of the UDLN model;
the pre-training is as follows:
the UDLN model consists of two-phase learning: firstly, inputting a training sample into an L12 norm sparse filtering L12SF feature extraction layer, wherein an L12 norm is a norm expression integrating a 1 norm and a 2 norm, and a feature competition special effect is generated by utilizing the L12SF so as to extract a feature value; then, the extracted characteristic value is sent to a weighted Euclidean distance similar affine WE-AP clustering layer to obtain parameters of the trained UDLN model;
step 4, inputting the test sample into the UDLN model after training to obtain a fault clustering and identifying result;
and 5, calculating various fault recognition rates according to the membership degree condition of the fault cluster and the cluster center, and realizing fault diagnosis of the mechanical equipment.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the tested parts in the step 1 comprise a crane, a bearing of a rotating machine and a gear.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the characteristics of the mixed domain fault characteristic data set in the step 2 comprise time domain, frequency domain and time-frequency domain characteristics.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the process of inputting training samples into the UDLN model for pre-training is as follows:
step 3.1, initializing parameters of the UDLN model;
step 3.2, taking the training sample feature subset as the original input of two layers of L12SF, and extracting the low-dimensional unsupervised features of the mixed domain fault feature data set layer by layer; wherein the objective function of L12SF is as follows:
wherein | | xi | purple1Representing a 1 norm, | × | | luminance2Representing 2 norm for M training samplesDenotes the i-th characteristic value, x, of the i-th training sampleiDenotes the ith training sample, wiIs the weight parameter of the ith training sample,represents the minimization of wi,，Wl TIs the transpose of the weight parameter of the I characteristic value of the UDLN model, | | | count the luminanceL12L12 fused 1-norm to 2-norm; the L12 norm r (w) fused to the 1 norm and the 2 norm is:
wherein,for regularizing norm adjustment coefficient, extracting l characteristic values f of new ith training sample by optimizing objective functionl iIs composed of
fl i＝G(Wl Txi)
Wherein G () is an activation function;
will f isl i＝G(Wl Txi) Optimizing the objective function of L12SF by adopting an L-BFGS algorithm until convergence;
step 3.3, inputting the characteristic values extracted in the step 3.2 into a clustering layer of the WE-AP for fault clustering learning and fault division, wherein the WE-AP clustering firstly initializes a weighting similarity matrix S with N training samplesw
Wherein x is1kAnd x2kRepresenting the kth feature, S, of two different training sampleskRepresenting the variance of two training samples, N being the number of eigenvalues, N ═ M;
meanwhile, the credibility and the availability of the ith training sample are calculated in the specific calculation mode
R(i,k)＝S(i,k)-max{A(i,j)+S(i,j)}
St.j＝1,2,...,N and j≠i,k
St.j＝1,2,...,N and j≠i,k
R(k,k)＝B(k)-max{A(k,j)+S(k,j)}
St.j＝1,2,...,N and j≠k
Wherein, b (k) is a prior value representing the tendency of each training sample to be selected as a cluster center point; s (i, K) represents a weighted similarity matrix of the ith training sample and the kth training sample, A (i, j) represents the availability of the ith training sample for selecting the jth training sample as the clustering center of the ith training sample, S (i, j) represents the weighted similarity matrix of the ith training sample and the jth training sample, A (i, K) represents the availability of the ith training sample for selecting the kth training sample as the clustering center of the ith training sample, A (i, j) represents the availability of the ith training sample for selecting the jth training sample as the clustering center of the ith training sample, R (K, K) represents the reliability of the kth training sample for selecting the kth training sample as the clustering center of the ith training sample, R (i, K) represents the reliability of the ith training sample for selecting the kth training sample as the clustering center of the ith training sample, and R (K, K) represents the reliability of the kth training sample for selecting the kth training sample as the clustering center of the kth training sample, a (k, j) represents the availability of the kth training sample to select the jth training sample as the clustering center of the kth training sample, and S (k, j) represents the weighted similarity matrix of the kth training sample and the jth training sample.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, in step 3.2, G (×) is an activation function using soft absolute values:
where σ is the activation threshold, thenThe L-BFGS algorithm is used to optimize the objective function of L12SF until convergence.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the sigma is 10-8。
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the condition and membership degree condition of the kth training sample as a clustering center are determined according to the following formula calculated by the UDLN model in step 5
R(k,k)+A(k,k)＞0
Wherein, R (k, k) represents the credibility of the k training sample selecting the k training sample as the clustering center of the k training sample, and A (k, k) represents the availability of the k training sample selecting the k training sample as the clustering center of the k training sample;
according to a preset maximum iteration number tmaxUpdating the reliability and availability t of each training samplemaxThen, the updating mode is
R(i,k)＝(1-lam)*R(i,k)+lam*R(i-1,k)
A(i,k)＝(1-lam)*A(i,k)+lam*A(i-1,k)
Wherein lam is a damping factor, wherein R (i-1, k) represents the reliability of selecting the kth training sample as the clustering center of the (i-1) th training sample, and A (i-1, k) represents the availability of selecting the kth training sample as the clustering center of the (i-1) th training sample;
and finally, calculating various fault recognition rates according to the membership degree condition of the fault cluster and the cluster center, and realizing fault diagnosis of the mechanical equipment.
As a further optimization scheme of the mechanical fault diagnosis method of the unsupervised deep learning network, the model parameters of the UDLN in the step 3 comprise the weight parameter w of the ith training sampleiAnd regularization norm adjustment coefficient
Compared with the prior art, the invention adopting the technical scheme has the following technical effects:
(1) the improved L12 norm sparse filtering (L12SF) unsupervised feature extraction network can extract unsupervised special diagnosis information of fault data layer by layer, and the newly defined L12 norm enhances the generalization capability of extracted low-dimensional features;
(2) the improved weighted Euclidean similar affine (WE-AP) clustering can more accurately realize the self-adaptive unsupervised clustering of fault categories, overcomes the defects that the traditional clustering method needs to preset the clustering number and the clustering center, and simultaneously highlights the contribution degree of different sample characteristics to clustering;
(3) on the basis of L12SF and WE-AP clustering, a brand-new unsupervised deep learning network-UDLN is constructed, and the neural network can realize whole-course unsupervised learning from feature extraction to pattern recognition;
(4) the invention discloses a novel mechanical fault diagnosis method based on a UDLN model, which realizes unsupervised intelligent mechanical fault diagnosis under the condition of no class label information and verifies the feasibility of the method by vibration signals of mechanical equipment. In addition, the method is simple and easy to implement, and is suitable for online or field fault diagnosis of mechanical equipment under the condition that the class labels are lacked.
Drawings
FIG. 1 is a flow chart of the present technique.
Fig. 2 is a schematic structural diagram of a multilayer L12SF network.
FIG. 3 is a schematic diagram of the UDLN model architecture.
FIG. 4 is a time domain and frequency domain waveform envelope of various types of fault signals of a bearing.
Fig. 5 is a confusion matrix for mechanical fault identification based on UDLN fault diagnosis.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
In actual industrial field diagnosis, multilayer unsupervised learning can provide one of solvable schemes for training samples and class labels thereof in the absence situation, and therefore the invention aims to construct a new Unsupervised Deep Learning Network (UDLN) model through the proposed L12 norm sparse filtering (L12SF) and weighted Euclidean distance similar affine (WE-AP) in a combined mode, and a mechanical fault diagnosis method based on the UDLN model is invented on the basis.
Fig. 1 shows a flow of a method for diagnosing a mechanical fault in an unsupervised deep learning network, which may be summarized as follows:
(1) firstly, arranging corresponding sensors at key positions of mechanical equipment (such as a crane and the like) to pick up vibration signals of the mechanical equipment;
(2) converting the vibration signal into a multi-domain mixed fault feature data set, and dividing the multi-domain mixed fault feature data set into a test sample feature subset and a training sample feature subset;
currently, the most common features include time domain, frequency domain, time-frequency domain, etc., which well characterize the health of a mechanical failure data set. A mixed multi-domain feature set shown in Table 1 was constructed from the bearing vibration signal samples and recorded as feature data set H.
TABLE 1 statistical characteristic parameters of time domain, frequency spectrum, time-frequency domain
(3) Initializing network parameters, and inputting the training sample feature subset into a UDLN model for training and learning;
the specific method is described as follows:
(3.1) description of L12SF method
Compared with other feature learning methods, the sparse filtering does not need to try to construct a model for data distribution, and only needs to optimize the sparsity of a simple loss function 2-norm standardization feature. Given a sample set with M training samples as follows
fl i＝Wl Txi
Wherein f isl iI-th characteristic value, W, representing the i-th sample (column)l TThe transpose of the weight matrix representing the l-th eigenvalue.
Sparse filtering uses a 2-norm to normalize the sparsity of the data. These features form a feature matrix. We first normalized each row of the feature matrix in all samples with a 2-norm.
Wherein,denotes the l characteristic value, f, of the 2 norm normalized training samplelRepresenting the ith eigenvalue of the training sample. Each column is then normalized by its 2 norm
Wherein,the ith sample representing 2-norm normalization,the ith sample representing the 2-norm normalized eigenvalue. Finally, the weight matrix WT lThe 1 norm can be constrained by optimizing the cost function for each sampleThe solution is as follows. The normalized features are optimized for sparsity by using a 1 penalty. Therefore, for a data with M samples, the objective function for a given sparse filter is:
wherein M represents the number of training samples, | Y calculation1Representing a 1 norm, | × | | luminance2Representing a 2 norm. Meanwhile, the L1 norm can make data more sparse, the 2 norm can prevent overfitting of the data, and the generalization capability of the model is improved. Therefore, if the advantages of the 1 norm and the 2 norm are fused, the sparse characteristic of the data and the generalization performance of the neural network can be fully adjusted. Therefore, the present invention redefines a norm R (w):
wherein,adjusting the coefficient for regularization norm, wiIs the weight parameter of the ith training sample of the model. Finally, the objective function of the above formula is changed to:
wherein | | xi | purpleL12The L12 norm is expressed, and the learning characteristic can find more nonlinear information from the input sample through the objective function in the optimization formula, so that the generalization capability is better.
fl i＝G(Wl Txi)
Where G (×) represents the activation function for soft absolute values. Therefore, the structural schematic diagram of L12SF is shown in fig. 2.
(3.2) description of WE-AP clustering method
The specific calculation steps of WE-AP clustering are shown below
(1): calculating a weighted similarity matrix SwDefining the maximum iteration tmax＝1000.
(2): calculating the credibility R and the availability A of each sample point in a specific way
R(i,k)＝S(i,k)-max{A(i,j)+S(i,j)}
St.j＝1,2,...,N and j≠i,k
St.j＝1,2,...,N and j≠i,k
R(k,k)＝B(k)-max{A(k,j)+S(k,j)}
St.j＝1,2,...,N and j≠k
Wherein, b (k) is a prior value representing the tendency of each training sample to be selected as a cluster center point; s (k, j) represents a weighted similarity matrix of the training samples, R (i, k) represents a degree to which the training sample i is suitable as a clustering center of the training sample k, and a (k, j) represents a degree to which the training sample k selects the training sample j as its clustering center.
(3): and (3) determining whether the kth training sample can be used as a clustering central point according to the calculation credibility, wherein the similarity value of the training sample is set as B (k):
R(k,k)+A(k,k)＞0
where R (k, k) represents the degree to which the training sample k is suitable as the clustering center of the training sample k, and a (k, k) represents the degree to which the training sample k selects the training sample k as its clustering center.
(4): and updating the credibility R and the availability A. The updating method is
R(i,k)＝(1-lam)*R(i,k)+lam*R(i-1,k)
A(i,k)＝(1-lam)*A(i,k)+lam*A(i-1,k)
Where lam is a damping factor, R (i-1, k) represents the degree to which the training sample i-1 is suitable as the clustering center of the training sample k, and A (i-1, k) represents the degree to which the training sample i-1 selects the training sample k as the clustering center. The purpose of the method is to balance the credibility R and the availability A in two iterations before and after the oscillation is avoided.
(5): checking whether a termination condition is met, i.e. the number of iterations reaches a maximum iteration tmaxAnd if not, jumping to the step (2) in 3.2.
(3.3) description of the UDLN method
UDLN consists of two phases of learning: firstly, inputting a sample into a corrected L12 sparse filter (L12SF), and extracting deep characteristic expression by using the algorithm; then, the deep features are sent to a WE-AP neural network, an internal distribution model of input data is further established through nonlinear function mapping, and a network weight of the whole learned UDLN model is obtained. During recognition, firstly, sparse filtering is carried out on a test sample, and then the extracted features are input into a trained WE-AP-based UDLN directly having a classification effect. Compared with deep learning and other fault diagnosis methods, the whole model does not need to use a class label, and whole-process unsupervised feature extraction and fault mode identification can be realized. The training process for the UDLN model is as follows: the algorithm pre-training stage consists of two layers of feature extraction layers consisting of an L12SF structure and one layer of unsupervised WE-AP clustering. The structure of the UDLN model is shown in FIG. 3.
The specific implementation mode is as follows:
example 1: in this section, to verify the performance of the proposed UDLN-based mechanical fault diagnosis model, it is necessary to simulate various rolling bearing faults. The experiment is carried out on an accelerated bearing life tester (ABLT-1A is provided by Hangzhou bearing test research center), and the main components of the experiment table are as follows: the test bench comprises a computer control system, a test headstock, an experimental head, a lubricating system, a transmission system, a loading system and a test and data acquisition system. The tester is designed with four bearings mounted on a shaft driven by an ac motor, with its drive system supported by a rubber belt for connecting the ac motor to the shaft using two pulleys. Meanwhile, various faults of the 6205 single-row deep groove ball bearing are simulated in an experiment. The experimental linear cutting respectively simulates five faults of 6205 rolling bearing, namely inner ring fault, inner ring and outer ring composite weak fault, rolling element outer ring composite weak fault and the like. And acquiring vibration data of a group of 20480 points every 5 minutes by using an NI9234 data acquisition card and a PCB acceleration sensor. The sampling frequency is 10240Hz, and the rotating speed is 1050 r/min. Each data acquisition system comprises four acceleration sensors and a data acquisition card. 6205 the test conditions and experimental data of the bearing are shown in table 2.
TABLE 26205 bearing test conditions
In the bearing fault data, there are weak faults and also serious faults in various bearing states, and also various fault conditions under different conditions such as composite faults, and the like, and the time domain and single-side spectrum frequency domain oscillograms of vibration signals are shown in fig. 4. It can be seen that, compared with a serious fault, a weak fault vibration signal is weak relatively and is greatly influenced by the noise interference degree, the impact characteristic is not obvious, the resonance frequency band in a frequency spectrum is not obvious, and the early weak fault is difficult to observe. The traditional time-frequency method is difficult to quantify the fault degree and the fault category, and needs to rely on a large amount of expert knowledge and field experience, so that the actual fault diagnosis is difficult. Therefore, an intelligent fault diagnosis method is required to quantify the fault diagnosis result, and the intelligent fault diagnosis method based on machine learning is widely applied. In order to display more diagnostic information, fig. 5 shows the processing result of the proposed fault diagnosis model on the bearing fault data feature set H according to the fault diagnosis flowchart 1, and the test sample confusion matrix of the fault diagnosis result is shown in fig. 5. As can be seen from fig. 5, the proposed method classifies the test samples of several second-class samples into the fourth class by mistake, because the discrimination between two classes of composite faults is not obvious, and all the faults including the inner ring are easy to be confused, and the clustering of other classes is successful.
To verify the proposed UDLN-based feature extraction and pattern recognition capabilities. The invention relates to a method for preparing L12SF + AP, L12SF + Kmeans; l12SF + FCM was used as a comparative test for the UDLN diagnostic model, and is referred to as { UDLN — FD 1; l12SF + AP FD 2; l12SF + Kmeans FD 3; l12SF + FCM FD4 }. According to the above fault diagnosis flowchart 1, using the other diagnosis models as a comparison test, respectively inputting the time domain, frequency domain and mixed domain characteristics into the 4 types of fault diagnosis models, and classifying according to membership and division to obtain corresponding recognition results, as shown in table 3:
TABLE 3 clustering results based on different feature data sets
In summary, to make intelligent fault diagnosis independent of a priori knowledge and the expertise of the diagnostician. The invention provides a mechanical fault diagnosis method based on an unsupervised learning neural network (ULDN). The ULDN model proposed by the method is composed of two layers of improved L12SF and one layer of improved AP clustering. In the ULDN approach, improved sparse filtering can adaptively capture features with discriminatory information from a vibration signal in an unsupervised manner; these features are then input to the improved AP cluster and the health is classified in an unsupervised manner. Example research of a bearing data set shows that the method not only can learn high-level features with discriminability, but also can effectively realize unsupervised full-automatic fault diagnosis. Under the condition of not using label information, the method can fully utilize the advantages of unsupervised learning, improve the accuracy of mechanical fault diagnosis and automatically identify fault health conditions.
The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.
Claims (8)
1. A mechanical fault diagnosis method of an unsupervised deep learning network is characterized by comprising the following steps:
step 1, pre-selecting a tested part on mechanical equipment, and collecting a vibration signal of the tested part on the mechanical equipment;
step 2, converting the acquired vibration signals into a mixed domain fault feature data set, and dividing the mixed domain fault feature data set into a test sample feature subset and a training sample feature subset, wherein the test sample feature subset is used as a test sample, and the training sample feature subset is used as a training sample;
step 3, initializing parameters of an unsupervised deep learning network UDLN model, inputting training samples into the UDLN model for pre-training, and obtaining parameters of the UDLN model;
the pre-training is as follows:
the UDLN model consists of two-phase learning: firstly, inputting a training sample into an L12 norm sparse filtering L12SF feature extraction layer, wherein an L12 norm is a norm expression integrating a 1 norm and a 2 norm, and a feature competition special effect is generated by utilizing the L12SF so as to extract a feature value; then, the extracted characteristic value is sent to a weighted Euclidean distance similar affine WE-AP clustering layer to obtain parameters of the trained UDLN model;
step 4, inputting the test sample into the UDLN model after training to obtain a fault clustering and identifying result;
and 5, calculating various fault recognition rates according to the membership degree condition of the fault cluster and the cluster center, and realizing fault diagnosis of the mechanical equipment.
2. The method for diagnosing the mechanical fault of the unsupervised deep learning network according to claim 1, wherein the tested parts in the step 1 comprise a crane, a bearing of a rotating machine and a gear.
3. The method as claimed in claim 1, wherein the characteristics of the mixed-domain fault feature data set in step 2 include time-domain, frequency-domain and time-frequency-domain characteristics.
4. The method for diagnosing the mechanical fault of the unsupervised deep learning network according to claim 1, wherein the process of inputting the training samples into the UDLN model for pre-training is as follows:
step 3.1, initializing parameters of the UDLN model;
step 3.2, taking the training sample feature subset as the original input of two layers of L12SF, and extracting the low-dimensional unsupervised features of the mixed domain fault feature data set layer by layer; wherein the objective function of L12SF is as follows:
wherein | | xi | purple1Representing a 1 norm, | × | | luminance2Representing 2 norm for M training samples Denotes the i-th characteristic value, x, of the i-th training sampleiDenotes the ith training sample, wiIs the weight parameter of the ith training sample,represents the minimization of wi,，Wl TIs the transpose of the weight parameter of the I characteristic value of the UDLN model, | | | count the luminanceL12L12 fused 1-norm to 2-norm; the L12 norm r (w) fused to the 1 norm and the 2 norm is:
wherein,for regularizing norm adjustment coefficient, extracting l characteristic values f of new ith training sample by optimizing objective functionl iIs composed of
fl i＝G(Wl Txi)
Wherein G () is an activation function;
will f isl i＝G(Wl Txi) Optimizing the objective function of L12SF by adopting an L-BFGS algorithm until convergence;
step 3.3, inputting the characteristic value extracted in the step 3.2 into a WE-AP clustering layer for failureLearning and fault division of fault cluster, wherein WE-AP cluster firstly initializes a weighted similarity matrix S with N training samplesw
Wherein x is1kAnd x2kRepresenting the kth feature, S, of two different training sampleskRepresenting the variance of two training samples, N being the number of eigenvalues, N ═ M;
meanwhile, the credibility and the availability of the ith training sample are calculated in the specific calculation mode
R(i,k)＝S(i,k)-max{A(i,j)+S(i,j)}
St.j＝1,2,...,Nandj≠i,k
St.j＝1,2,...,Nandj≠i,k
R(k,k)＝B(k)-max{A(k,j)+S(k,j)}
St.j＝1,2,...,Nandj≠k
Wherein, b (k) is a prior value representing the tendency of each training sample to be selected as a cluster center point; s (i, K) represents a weighted similarity matrix of the ith training sample and the kth training sample, A (i, j) represents the availability of the ith training sample for selecting the jth training sample as the clustering center of the ith training sample, S (i, j) represents the weighted similarity matrix of the ith training sample and the jth training sample, A (i, K) represents the availability of the ith training sample for selecting the kth training sample as the clustering center of the ith training sample, A (i, j) represents the availability of the ith training sample for selecting the jth training sample as the clustering center of the ith training sample, R (K, K) represents the reliability of the kth training sample for selecting the kth training sample as the clustering center of the ith training sample, R (i, K) represents the reliability of the ith training sample for selecting the kth training sample as the clustering center of the ith training sample, and R (K, K) represents the reliability of the kth training sample for selecting the kth training sample as the clustering center of the kth training sample, a (k, j) represents the availability of the kth training sample to select the jth training sample as the clustering center of the kth training sample, and S (k, j) represents the weighted similarity matrix of the kth training sample and the jth training sample.
5. The method according to claim 4, wherein in step 3.2, G () is an activation function using soft absolute values:
where σ is the activation threshold, thenThe L-BFGS algorithm is used to optimize the objective function of L12SF until convergence.
6. The method of claim 5, wherein σ is 10 ═ 10 for diagnosing mechanical failure in the unsupervised deep learning network-8。
7. The method as claimed in claim 1, wherein in step 5, the following formula is calculated according to the UDLN model to determine the condition and membership degree of the kth training sample as the clustering center
R(k,k)+A(k,k)＞0
Wherein, R (k, k) represents the credibility of the k training sample selecting the k training sample as the clustering center of the k training sample, and A (k, k) represents the availability of the k training sample selecting the k training sample as the clustering center of the k training sample;
according to a preset maximum iteration number tmaxUpdating the reliability and availability t of each training samplemaxThen, the updating mode is
R(i,k)＝(1-lam)*R(i,k)+lam*R(i-1,k)
A(i,k)＝(1-lam)*A(i,k)+lam*A(i-1,k)
Wherein lam is a damping factor, wherein R (i-1, k) represents the reliability of selecting the kth training sample as the clustering center of the (i-1) th training sample, and A (i-1, k) represents the availability of selecting the kth training sample as the clustering center of the (i-1) th training sample;
and finally, calculating various fault recognition rates according to the membership degree condition of the fault cluster and the cluster center, and realizing fault diagnosis of the mechanical equipment.
8. The method as claimed in claim 1, wherein the model parameters of UDLN in step 3 include weight parameter w of ith training sampleiAnd regularization norm adjustment coefficient
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810949099.8A CN109186973B (en) | 2018-08-20 | 2018-08-20 | Mechanical fault diagnosis method of unsupervised deep learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810949099.8A CN109186973B (en) | 2018-08-20 | 2018-08-20 | Mechanical fault diagnosis method of unsupervised deep learning network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109186973A true CN109186973A (en) | 2019-01-11 |
| CN109186973B CN109186973B (en) | 2020-03-31 |
Family
ID=64918447
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810949099.8A Active CN109186973B (en) | 2018-08-20 | 2018-08-20 | Mechanical fault diagnosis method of unsupervised deep learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109186973B (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110082106A (en) * | 2019-04-17 | 2019-08-02 | 武汉科技大学 | A kind of Method for Bearing Fault Diagnosis of the depth measure study based on Yu norm |
| CN110132554A (en) * | 2019-04-17 | 2019-08-16 | 东南大学 | A kind of rotary machinery fault diagnosis method that depth Laplce encodes certainly |
| CN110705114A (en) * | 2019-10-10 | 2020-01-17 | 辽宁工程技术大学 | Ventilation fault diagnosis method without training sample |
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN110864887A (en) * | 2019-11-19 | 2020-03-06 | 北京瑞莱智慧科技有限公司 | Method, device, medium and computing equipment for determining operating condition of mechanical equipment |
| CN111089720A (en) * | 2020-01-16 | 2020-05-01 | 山东科技大学 | Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed |
| CN111734669A (en) * | 2020-07-02 | 2020-10-02 | 重庆大学 | Multi-source information layered fusion centrifugal blower fault diagnosis method |
| CN111811818A (en) * | 2020-06-02 | 2020-10-23 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
| CN112131787A (en) * | 2020-09-18 | 2020-12-25 | 江西兰叶科技有限公司 | Unsupervised self-evolving motor design method and system |
| CN112488983A (en) * | 2019-09-11 | 2021-03-12 | 中信戴卡股份有限公司 | Defect identification network obtaining method, defect identification method and grade determining method |
| CN112631250A (en) * | 2020-12-15 | 2021-04-09 | 中国计量大学 | Fault isolation and identification method in nonlinear process based on denoising autoencoder |
| CN113642708A (en) * | 2021-08-13 | 2021-11-12 | 北京赛目科技有限公司 | Training method, recognition method and device for vehicle environment grade recognition model |
| CN113762005A (en) * | 2020-11-09 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Method, device, equipment and medium for training feature selection model and classifying objects |
| CN113935443A (en) * | 2021-12-20 | 2022-01-14 | 广东电网有限责任公司广州供电局 | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium |
| CN114414227A (en) * | 2021-12-29 | 2022-04-29 | 华电电力科学研究院有限公司 | Equipment collision sensing method, device, equipment and storage medium |
| CN115096627A (en) * | 2022-06-16 | 2022-09-23 | 中南大学 | Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment |
| CN115563575A (en) * | 2022-09-28 | 2023-01-03 | 江苏省特种设备安全监督检验研究院 | Crane fault prediction method, system and storage medium based on neural network |
| WO2023044978A1 (en) * | 2021-09-27 | 2023-03-30 | 苏州大学 | Adversarial-flow-model-based unsupervised fault diagnosis method for mechanical device |
| CN118551289A (en) * | 2024-07-29 | 2024-08-27 | 南京康尼机电股份有限公司 | Unsupervised diagnosis method and system for electromechanical equipment of rail transit |
| CN118551289B (en) * | 2024-07-29 | 2024-11-05 | 南京康尼机电股份有限公司 | Unsupervised diagnosis method and system for electromechanical equipment of rail transit |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102902961A (en) * | 2012-09-21 | 2013-01-30 | 武汉大学 | Face super-resolution processing method based on K neighbor sparse coding average value constraint |
| CN104751191A (en) * | 2015-04-23 | 2015-07-01 | 重庆大学 | Sparse self-adaptive semi-supervised manifold learning hyperspectral image classification method |
| CN105930629A (en) * | 2016-07-13 | 2016-09-07 | 大唐南京环保科技有限责任公司 | On-line fault diagnosis method based on massive amounts of operating data |
| CN105975443A (en) * | 2016-05-04 | 2016-09-28 | 西南大学 | Lasso-based anomaly detection method and system |
| CN106017876A (en) * | 2016-05-11 | 2016-10-12 | 西安交通大学 | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network |
| CN106295694A (en) * | 2016-08-05 | 2017-01-04 | 浙江工业大学 | Face recognition method for iterative re-constrained group sparse representation classification |
| CN106555788A (en) * | 2016-11-11 | 2017-04-05 | 河北工业大学 | Application of the deep learning based on Fuzzy Processing in hydraulic equipment fault diagnosis |
| CN107944454A (en) * | 2017-11-08 | 2018-04-20 | 国网电力科学研究院武汉南瑞有限责任公司 | A kind of semanteme marking method based on machine learning for substation |
-
2018
- 2018-08-20 CN CN201810949099.8A patent/CN109186973B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102902961A (en) * | 2012-09-21 | 2013-01-30 | 武汉大学 | Face super-resolution processing method based on K neighbor sparse coding average value constraint |
| CN104751191A (en) * | 2015-04-23 | 2015-07-01 | 重庆大学 | Sparse self-adaptive semi-supervised manifold learning hyperspectral image classification method |
| CN105975443A (en) * | 2016-05-04 | 2016-09-28 | 西南大学 | Lasso-based anomaly detection method and system |
| CN106017876A (en) * | 2016-05-11 | 2016-10-12 | 西安交通大学 | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network |
| CN105930629A (en) * | 2016-07-13 | 2016-09-07 | 大唐南京环保科技有限责任公司 | On-line fault diagnosis method based on massive amounts of operating data |
| CN106295694A (en) * | 2016-08-05 | 2017-01-04 | 浙江工业大学 | Face recognition method for iterative re-constrained group sparse representation classification |
| CN106555788A (en) * | 2016-11-11 | 2017-04-05 | 河北工业大学 | Application of the deep learning based on Fuzzy Processing in hydraulic equipment fault diagnosis |
| CN107944454A (en) * | 2017-11-08 | 2018-04-20 | 国网电力科学研究院武汉南瑞有限责任公司 | A kind of semanteme marking method based on machine learning for substation |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110132554B (en) * | 2019-04-17 | 2020-10-09 | 东南大学 | Rotary machine fault diagnosis method based on deep Laplace self-coding |
| CN110132554A (en) * | 2019-04-17 | 2019-08-16 | 东南大学 | A kind of rotary machinery fault diagnosis method that depth Laplce encodes certainly |
| CN110082106A (en) * | 2019-04-17 | 2019-08-02 | 武汉科技大学 | A kind of Method for Bearing Fault Diagnosis of the depth measure study based on Yu norm |
| CN110082106B (en) * | 2019-04-17 | 2021-08-31 | 武汉科技大学 | Bearing fault diagnosis method based on Yu norm deep measurement learning |
| CN112488983A (en) * | 2019-09-11 | 2021-03-12 | 中信戴卡股份有限公司 | Defect identification network obtaining method, defect identification method and grade determining method |
| CN110705114A (en) * | 2019-10-10 | 2020-01-17 | 辽宁工程技术大学 | Ventilation fault diagnosis method without training sample |
| CN110705114B (en) * | 2019-10-10 | 2023-04-07 | 辽宁工程技术大学 | Ventilation fault diagnosis method without training sample |
| CN110864887A (en) * | 2019-11-19 | 2020-03-06 | 北京瑞莱智慧科技有限公司 | Method, device, medium and computing equipment for determining operating condition of mechanical equipment |
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN111089720A (en) * | 2020-01-16 | 2020-05-01 | 山东科技大学 | Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed |
| CN111089720B (en) * | 2020-01-16 | 2021-10-08 | 山东科技大学 | Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed |
| CN111811818A (en) * | 2020-06-02 | 2020-10-23 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
| CN111811818B (en) * | 2020-06-02 | 2022-02-01 | 桂林电子科技大学 | Rolling bearing fault diagnosis method based on AP clustering algorithm of specified clustering number |
| CN111734669A (en) * | 2020-07-02 | 2020-10-02 | 重庆大学 | Multi-source information layered fusion centrifugal blower fault diagnosis method |
| CN112131787A (en) * | 2020-09-18 | 2020-12-25 | 江西兰叶科技有限公司 | Unsupervised self-evolving motor design method and system |
| CN112131787B (en) * | 2020-09-18 | 2022-05-27 | 江西兰叶科技有限公司 | Unsupervised self-evolving motor design method and system |
| CN113762005A (en) * | 2020-11-09 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Method, device, equipment and medium for training feature selection model and classifying objects |
| CN112631250A (en) * | 2020-12-15 | 2021-04-09 | 中国计量大学 | Fault isolation and identification method in nonlinear process based on denoising autoencoder |
| CN113642708A (en) * | 2021-08-13 | 2021-11-12 | 北京赛目科技有限公司 | Training method, recognition method and device for vehicle environment grade recognition model |
| CN113642708B (en) * | 2021-08-13 | 2024-05-14 | 北京赛目科技有限公司 | Training method, recognition method and device for vehicle environment level recognition model |
| WO2023044978A1 (en) * | 2021-09-27 | 2023-03-30 | 苏州大学 | Adversarial-flow-model-based unsupervised fault diagnosis method for mechanical device |
| CN113935443A (en) * | 2021-12-20 | 2022-01-14 | 广东电网有限责任公司广州供电局 | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium |
| CN114414227A (en) * | 2021-12-29 | 2022-04-29 | 华电电力科学研究院有限公司 | Equipment collision sensing method, device, equipment and storage medium |
| CN114414227B (en) * | 2021-12-29 | 2023-05-26 | 华电电力科学研究院有限公司 | Equipment collision sensing method, device, equipment and storage medium |
| CN115096627A (en) * | 2022-06-16 | 2022-09-23 | 中南大学 | Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment |
| CN115096627B (en) * | 2022-06-16 | 2023-04-07 | 中南大学 | Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment |
| CN115563575A (en) * | 2022-09-28 | 2023-01-03 | 江苏省特种设备安全监督检验研究院 | Crane fault prediction method, system and storage medium based on neural network |
| CN115563575B (en) * | 2022-09-28 | 2023-04-21 | 江苏省特种设备安全监督检验研究院 | Crane fault prediction method, system and storage medium based on neural network |
| CN118551289A (en) * | 2024-07-29 | 2024-08-27 | 南京康尼机电股份有限公司 | Unsupervised diagnosis method and system for electromechanical equipment of rail transit |
| CN118551289B (en) * | 2024-07-29 | 2024-11-05 | 南京康尼机电股份有限公司 | Unsupervised diagnosis method and system for electromechanical equipment of rail transit |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109186973B (en) | 2020-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109186973B (en) | Mechanical fault diagnosis method of unsupervised deep learning network | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN110110768B (en) | Rolling bearing fault diagnosis method based on parallel feature learning and multiple classifiers | |
| CN109582003B (en) | Bearing fault diagnosis method based on pseudo label semi-supervised kernel local Fisher discriminant analysis | |
| CN106980822B (en) | A kind of rotary machinery fault diagnosis method based on selective ensemble study | |
| CN112036301B (en) | Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion | |
| CN109781411B (en) | Bearing fault diagnosis method combining improved sparse filter and KELM | |
| CN112254964A (en) | Rolling bearing fault diagnosis method based on rapid multi-scale convolution neural network | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN112819059B (en) | Rolling bearing fault diagnosis method based on popular retention transfer learning | |
| CN108871762A (en) | A kind of gearbox of wind turbine method for diagnosing faults | |
| CN111753891B (en) | Rolling bearing fault diagnosis method based on unsupervised feature learning | |
| CN113837000A (en) | Small sample fault diagnosis method based on task sequencing meta-learning | |
| CN110132554A (en) | A kind of rotary machinery fault diagnosis method that depth Laplce encodes certainly | |
| CN112860183B (en) | Multisource distillation-migration mechanical fault intelligent diagnosis method based on high-order moment matching | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| CN114091504B (en) | Rotary machine small sample fault diagnosis method based on generation countermeasure network | |
| CN112257530A (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN111397902A (en) | Rolling bearing fault diagnosis method based on feature alignment convolutional neural network | |
| CN114755017B (en) | Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| CN111707472B (en) | Rolling bearing fault diagnosis method based on FFT and full-connection layer-SVM | |
| CN117475191A (en) | Bearing fault diagnosis method for feature alignment domain antagonistic neural network | |
| CN114139598A (en) | Fault diagnosis method and diagnosis framework based on deep cost sensitive convolution network | |
| CN111783941B (en) | Mechanical equipment diagnosis classification method based on probability confidence convolutional neural network |
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 |