CN112464990A - Method and device for sensing vibration data based on current and voltage sensor - Google Patents
Method and device for sensing vibration data based on current and voltage sensor Download PDFInfo
- Publication number
- CN112464990A CN112464990A CN202011212653.8A CN202011212653A CN112464990A CN 112464990 A CN112464990 A CN 112464990A CN 202011212653 A CN202011212653 A CN 202011212653A CN 112464990 A CN112464990 A CN 112464990A
- Authority
- CN
- China
- Prior art keywords
- data
- network
- current
- sample data
- vibration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 79
- 238000006243 chemical reaction Methods 0.000 claims abstract description 24
- 238000005070 sampling Methods 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 25
- 238000013527 convolutional neural network Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 11
- 238000013075 data extraction Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 abstract description 13
- 238000004364 calculation method Methods 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 8
- 230000001133 acceleration Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The embodiment of the invention provides a method and a device for sensing vibration data based on a current-voltage sensor, wherein the method comprises the following steps: acquiring current and voltage data acquired by a current and voltage sensor; inputting the current and voltage data into a generator network to obtain virtual vibration data; the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data. According to the method, the generator network and the discriminator network form the countermeasure network to carry out countermeasure training, so that the calculation capacity of the generator network on data is gradually improved, a better data conversion process can be realized under the condition that the hidden state characteristics of the data are guaranteed, and the virtual sensing function of the generator network from current and voltage data to virtual vibration data is realized.
Description
Technical Field
The invention relates to the technical field of virtual sensing, in particular to a method and a device for sensing vibration data based on a current-voltage sensor, electronic equipment and a storage medium.
Background
Although sensors gradually form ubiquitous sensing through intellectualization and integration at the present stage, most of detected objects are complex in structure, the sensors are difficult to arrange in the sensors or on the surface side, or the operations of selecting positions, installing, wiring, maintaining and the like of the sensors are time-consuming and labor-consuming, and the operation cost of the equipment is greatly increased.
In general, most devices can be equipped with current and voltage sensors at the time of shipment, or the procedure of adding current and voltage sensors is relatively simple. Therefore, the vibration data of the equipment is calculated by the sensor data such as the current, the voltage and the like, so that the intelligent sensing and judgment of the state of the motor equipment are realized, and the virtual sensing is formed and is an effective equipment state sensing scheme.
In the prior art, although a virtual sensing scheme tries to perform a mapping process from current and voltage data to vibration data by adopting a regression algorithm such as machine learning, the mathematical error between the converted data and standard data is relatively small, the converted data basically loses the health state characteristics of equipment contained in the vibration data, and therefore the schemes are difficult to adapt to practical application requirements.
Therefore, how to provide a method to complete the conversion from the device current-voltage data to the vibration data without losing the device status characteristics of the data hiding, thereby forming a virtual sensing process, is a problem to be solved.
Disclosure of Invention
The embodiment of the invention provides a method and a device for sensing vibration data based on a current-voltage sensor, electronic equipment and a storage medium, which are used for solving the technical defects in the prior art.
The embodiment of the invention provides a method for sensing vibration data based on a current-voltage sensor, which comprises the following steps:
acquiring current and voltage data acquired by a current and voltage sensor;
inputting the current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
According to one embodiment of the invention, the generator network comprises: a down-sampling layer, a data conversion layer and an up-sampling layer;
inputting the current-voltage data into a generator network to obtain virtual vibration data, comprising:
the down-sampling layer receives input current and voltage data and generates first sampling data;
the data conversion layer receives input first sampling data and generates first conversion data;
the up-sampling layer receives the input first conversion data and generates virtual vibration data.
According to the method for sensing vibration data based on the current-voltage sensor, the countermeasure network is trained through the following methods:
forming a second training set by current and voltage sample data and real vibration sample data corresponding to the current and voltage sample data, and inputting the second training set to the generator network so as to train the generator network;
marking the vibration data output by the generator network as virtual vibration sample data, and forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data and inputting the third training set to the discriminator network so as to train the discriminator network;
and forming a countermeasure network by the obtained generator network and the discriminator network, and inputting the first training set into the countermeasure network to train the countermeasure network.
According to the method for sensing vibration data based on the current-voltage sensor, the discriminator network comprises a first discriminator network and a second discriminator network;
forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data, and inputting the third training set to the discriminator network so as to train the discriminator network, wherein the method comprises the following steps:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to train the first discriminator network;
and inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network so as to train the second discriminator network.
According to the method for sensing vibration data based on the current-voltage sensor, the obtained generator network and the discriminator network form a countermeasure network, the first training set is input to the countermeasure network to train the countermeasure network, and the method comprises the following steps:
inputting the first training set into the countermeasure network, and alternately locking the parameters of the discriminator network or the generator network to update the parameters of the generator network or the discriminator network so that the loss function of the countermeasure network is smaller than a first threshold value, wherein the loss function of the countermeasure network is the combination of the loss functions of the first discriminator network and the second discriminator network.
According to the method for sensing vibration data based on the current-voltage sensor, disclosed by the invention, the first discriminator network is a convolutional neural network;
inputting the virtual vibration sample data and the real vibration sample data to a first discriminator network to train the first discriminator network, including:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to obtain the probability that the currently input data is the real vibration sample data;
and adjusting parameters of the first discriminator network to enable the probability to be larger than a second threshold value when the input data is real vibration sample data.
According to one embodiment of the present invention, in the method for sensing vibration data based on a current-voltage sensor, the second discriminator network includes: a first convolution layer, a time sequence data extraction layer, a second convolution layer and a normalization layer;
inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network to train the second discriminator network, including:
inputting the real vibration sample data and the virtual vibration sample data into a first convolution layer to generate first convolution data;
inputting the first volume data into a time sequence data extraction layer to generate time sequence data;
inputting the time sequence data into the second convolution layer to generate second convolution data;
inputting the second convolution data into a normalization layer, and generating the prediction probability of the fault type of the real vibration sample data;
inputting the fault type of the real vibration sample data, comparing the prediction probability with the fault type of the real vibration sample data, and counting the prediction accuracy of a second discriminator;
and adjusting the parameters of the second discriminator network by taking the prediction accuracy rate greater than a third threshold value as a target.
The embodiment of the invention also provides a device for sensing vibration data based on the current-voltage sensor, which is characterized by comprising the following components:
the acquisition module is used for acquiring current and voltage data acquired by the current and voltage sensor;
the sensing module is used for inputting the current and voltage data into the generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
The embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, the steps of the method for sensing vibration data based on a current and voltage sensor as described in any one of the above are implemented.
Embodiments of the present invention further provide a non-transitory computer readable storage medium, on which a computer program is stored, the computer program, when executed by a processor, implementing the steps of the method for sensing vibration data based on a current-voltage sensor as described in any one of the above.
According to the method and the device for sensing the vibration data based on the current and voltage sensor, the generator network and the discriminator network are combined to form the countermeasure network for countermeasure training, so that the calculation capacity of the generator network on the data is gradually improved, a better data conversion process can be realized under the condition that the hidden state characteristics of the data are guaranteed, and the virtual sensing function of the generator network from the current and voltage data to the virtual vibration data is realized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for sensing vibration data based on a current-voltage sensor according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a generator network according to an embodiment of the present invention;
FIG. 3 is a flow chart of a training process of the countermeasure network provided by the embodiment of the invention;
FIG. 4 is a diagram of a second training set provided by an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a second arbiter network according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of an apparatus for sensing vibration data based on a current-voltage sensor according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be 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 some, but not all, embodiments of the present invention. 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.
The terminology used in the one or more embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the invention. As used in one or more embodiments of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present invention refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It should be understood that, although the terms first, second, etc. may be used herein to describe various information in one or more embodiments of the present invention, such information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first aspect may be termed a second aspect, and, similarly, a second aspect may be termed a first aspect, without departing from the scope of one or more embodiments of the present invention. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.
The embodiment of the invention provides a method for sensing vibration data based on a current-voltage sensor, which comprises the following steps of:
101. and acquiring current and voltage data acquired by the current and voltage sensor.
The current and voltage sensor is arranged on the target equipment, and the data of the current and voltage sensor, which operates in different states, of the target equipment are acquired after the target equipment is operated by starting a motor of the target equipment.
102. And inputting the current voltage data into a generator network to obtain virtual vibration data.
Specifically, referring to fig. 2, fig. 2 shows a structure of a generator network according to an embodiment of the present invention, the generator network including: a down-sampling layer, a data conversion layer, and an up-sampling layer.
Step 102 includes the following steps S121 to S123:
s121, the down-sampling layer receives input current and voltage data and generates first sampling data;
s122, the data conversion layer receives input first sampling data and generates first conversion data;
and S123, the up-sampling layer receives the input first conversion data and generates virtual vibration data.
Since the LSTM network is better at handling time series data, unlike the conventional countermeasure network, the present invention proposes to use the LSTM network as a basic network framework of the generator instead of the CNN network. In this embodiment, the down-sampling layer uses a CNN network, the data conversion layer uses an LSTM network, and the up-sampling layer uses a CNN network.
In the embodiment, the sensing data is converted for 1 second each time, and the sampling rate of the sensing data is 2000 times per second, so that the generator needs to input and output 2000-dimensional time sequence data per second, the input data of each dimension includes five data such as three-phase current, two-phase voltage and the like at the current moment, and the output data of each dimension includes three data such as three-axis vibration acceleration and the like at the current moment. For the LSTM network, data of 2000 length is too large, because the number of times of reverse calculation of the neural network increases with the increase of the length of time series data, thereby causing the problem of gradient explosion or gradient disappearance, and finally limiting the updating process of parameters in the network, thereby causing the disappearance of the correlation of the previous and subsequent data, and limiting the processing performance of the network on the overlong time series. Therefore, the length of the LSTM network used in this embodiment is 128 dimensions, the input data of each dimension is composed of 5 values, the output is 3 data, and meanwhile, a CNN down-sampling layer is added before the LSTM network, and a CNN up-sampling layer is added after the LSTM network. The CNN downsampling layer is composed of several convolutional layers and downsampling layers and is used to reduce 2000-dimensional input data into 128-dimensional input data to reduce the required length of the LSTM network. The CNN up-sampling layer is composed of a plurality of convolution layers and up-sampling layers and is used for restoring 128-dimensional output data to 2000-dimensional output data and ensuring that the vibration acceleration data output by the generator network is consistent with the real vibration acceleration data in format.
The embodiment of the invention has simple and easily-constructed network structure by constructing the generator network of the CNN-LSTM-CNN structure. The CNN network is used for performing dimensionality reduction and dimensionality enhancement operation on the data, so that time sequence data can be better sensed and processed by the LSTM network, and simultaneously, the format of finally output data and original data can be ensured to be consistent, and the data conversion process can be stably performed.
In addition, in the embodiment, the generator network and the arbiter network are combined into the countermeasure network, and then the countermeasure network is trained to improve the estimation capability of the generator network on the data. The countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data and fault types corresponding to the current and voltage sample data.
According to the method for sensing the vibration data based on the current and voltage sensor, the generator network and the discriminator network are combined to form the countermeasure network to carry out countermeasure training, the calculation capacity of the generator network on the data is gradually improved, and therefore a good data conversion process can be achieved under the condition that the data have hidden state characteristics, and the virtual sensing function of the generator network from the current and voltage data to the virtual vibration data is achieved.
The training process of the countermeasure network according to the present embodiment is explained in detail below. Referring to FIG. 3, the training step of the countermeasure network comprises 301-303:
301. and forming a second training set by the current and voltage sample data and the real vibration sample data corresponding to the current and voltage sample data, and inputting the second training set to the generator network so as to train the generator network.
And the second training set comprises a current and voltage sensor and a vibration acceleration sensor which are arranged aiming at the equipment, then a motor is started, sensor data of the equipment running for a certain time under different states are collected, and the sensor data are stored according to a specified format.
Fig. 4 is a diagram illustrating a second training set. As can be seen, the data collected includes: sampling time sequence, one-phase voltage, two-phase voltage, three-phase voltage, one-phase current, two-phase current, x-axis vibration acceleration, y-axis vibration acceleration and z-axis vibration acceleration.
In addition, the embodiment also collects data of 9 motor working conditions (normal, A1 rotor unbalance (outer ring), A2 rotor unbalance (outer ring), A1B1 rotor unbalance (outer ring), inner ring fault, outer ring fault, light balls, medium balls and heavy balls). Wherein 6 groups of data are collected under each working condition, each group is 5 minutes, and the sampling rate is 2000 times per second. All data were as per 8: and 2, setting a training/testing set in proportion for training and optimizing the algorithm model.
Training the generator network model for the purpose of enabling it to convert the current-voltage data into virtual vibration data that is substantially similar to the real vibration data. At this time, one second of sensor data is intercepted as a sample, wherein current and voltage sample data is used as an input sample, and the network calculates 2000-dimensional output data, namely virtual vibration sample data according to the input. And after the network calculation is finished, calculating the error between the real vibration data and the virtual vibration sample data by using the corresponding real vibration data as a label, and reversely updating the weight parameters of each part of the generator network by taking the error minimization as a target. The error is calculated by using a mean square error loss function, which is expressed by the following formula (1):
where MSE represents the value of the loss function and n is the number of samples used in a training process. y isi ′A value, y, representing the calculated output of the network from the current batch of i-th samplesiThe actual vibration data, i.e. the label, representing the current lot of the ith sample.
302. And marking the vibration data output by the generator network as virtual vibration sample data, and forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data and inputting the third training set into the discriminator network so as to train the discriminator network.
In this embodiment, step 302 includes:
s321, inputting the virtual vibration sample data and the real vibration sample data to a first discriminator network to train the first discriminator network.
And S322, inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network so as to train the second discriminator network.
In particular, the virtual sensor data should be close enough to the real vibration data, and furthermore, it should have the availability of real data, which can be used in situations such as condition monitoring, fault classification, equipment health assessment, etc. Therefore, the embodiment of the invention provides a countermeasure network using a multi-arbiter single generator structure, so that virtual data can cover more application scenes by increasing the types of the arbiters. For example, in the present embodiment, it is proposed to use a true-false data classification network and a fault classification network as a first discriminator network and a second discriminator network, respectively, the second discriminator network being as shown in fig. 5.
The first discriminator network is used to evaluate whether the received data is real or given by the generator. In this embodiment, the first discriminator network is a CNN type neural network, and is composed of several convolution layers, a down-sampling layer, a full-link layer, and a softmax classifier.
During training, generating a 2000-dimensional vector by using virtual vibration sample data and real vibration sample data, and inputting the vector into a first discriminator network to obtain the probability that the currently input data is the real vibration sample data, wherein the probability value is 0 to 1; adjusting parameters of the first discriminator network so that the probability is greater than a second threshold when the input data is true vibration sample data.
For the second discriminator network, the data output by the generator tends to realize the function of fault classification, thereby ensuring that the virtual data is not only like real data in form, but also has the capability of checking equipment faults. Specifically, the second discriminator includes: a first convolution layer, a timing sequence data extraction layer, a second convolution layer and a normalization layer. In order to realize better time sequence data processing performance, the second discriminator adopts an LSTM network as a time sequence data extraction layer, and the first convolution layer and the second convolution layer both adopt a CNN network layer.
During training, the method comprises the following steps: inputting real vibration sample data or the virtual vibration sample data into the first convolution layer to generate first convolution data; inputting the first volume data into a time sequence data extraction layer to generate time sequence data; inputting time series data into the second convolution layer to generate second convolution data; inputting the second convolution data into a normalization layer to generate the prediction probability of the fault type of the real vibration sample data; inputting the fault type of the real vibration sample data, comparing the prediction probability with the fault type of the real vibration sample data, and counting the prediction accuracy of a second discriminator; and adjusting the parameters of the second discriminator network by taking the prediction accuracy rate larger than a third threshold value as a target.
Specifically, the data is reduced from 2000 dimensions to 128 dimensions by using a first convolution layer, then the time sequence characteristics of the data are extracted through an LSTM time sequence data extraction layer, the time sequence data are input into a second convolution layer to further reduce the dimensions to 10 dimensions, and finally the second convolution data are input into a softmax normalization layer, so that the accuracy of the prediction fault type to which the data belong is estimated. Specific fault types include the above-mentioned 9 types, plus a "dummy data" class, totaling 10 classes.
In this embodiment, the purpose of the training is to enable the first discriminator to distinguish between real vibration data and virtual vibration data, and enable the second discriminator to distinguish between a fault type represented by the real vibration data, or whether the input data is virtual vibration data or real vibration data. Since both of them finally use the softmax function to give the type judgment result, both use the cross entropy loss function, and the formula is shown in the following formula (2):
wherein L represents the value of the loss function;
n is the number of samples used in one training process;
m is the total number of all data types, the value of M is 2 for the first discriminator network, and the value of M is 10 for the second discriminator network;
yicrepresenting the real type of the nth sample data, if the type is the type c, the value is 1, otherwise, the value is 0;
picand representing the probability that the nth sample data is considered as the class c data after being evaluated by the discriminator network.
303. And forming a countermeasure network by the obtained generator network and the discriminator network, and inputting the first training set into the countermeasure network to train the countermeasure network.
Specifically, step 303 includes: inputting a first training set into the countermeasure network, and alternately locking the parameters of the discriminator network or the generator network to update the parameters of the generator network or the discriminator network so that the loss function of the countermeasure network is smaller than a first threshold value, wherein the loss function of the countermeasure network is the combination of the loss functions of the first discriminator network and the second discriminator network.
The loss function formula of the countermeasure network is shown in the following formula (3):
wherein N is the number of samples used in one training process, LA、LBAnd a and b are weighting parameters of the loss values of the first discriminator network and the second discriminator network, and the generator network is emphasized to different conversion targets by adjusting the values of a and b.
Specifically, the weight value of the discriminator network is locked first, then the generator network is used to generate virtual vibration data according to the voltage and current data of the sample, the discriminator network is used to give a predicted value of the type to which the virtual vibration data belongs, and the weight value of the generator network is adjusted according to the prediction deviation. At this time, the weight update targets are: (1) the first discriminator evaluates the virtual vibration sample data as real vibration sample data as much as possible; (2) and enabling the second judging device to enable the fault prediction type of the virtual vibration sample data to be consistent with the fault type of the real current and voltage data as much as possible. Thus, for a certain virtual vibration sample data, its label changes from 0 to 1 for the first discriminator; for the second discriminator, its label changes from "virtual data class" to "failure class corresponding to real vibration sample data".
And then, locking the weight value of the generator network, and then training the discriminator network, wherein the training target is consistent with the training target in the step 302, namely the loss function of the countermeasure network is smaller than a first threshold value.
And through multiple iterations, the performances of the generator network and the discriminator network are greatly improved. And finally, independently taking out the generator network, namely a network model capable of carrying out virtual vibration data calculation based on current and voltage, and packaging the network model into a model for application and deployment.
According to the embodiment of the invention, a single-generator network-multi-discriminator network confrontation network structure is used, the function effectiveness of the generator network generated data is realized by adding a plurality of types of function discriminator networks, for example, the fault classification is used as an evaluation discriminator, so that the virtual vibration data and the real vibration data can be guaranteed not only to have small numerical errors, but also the fault identification performance can be guaranteed. Different functional emphasis of the virtual data is respectively realized by setting the weight of the discriminator network.
Based on any of the above embodiments, fig. 6 is a schematic structural diagram of an apparatus for sensing vibration data based on a current-voltage sensor according to an embodiment of the present invention, as shown in fig. 6, including:
the acquisition module 601 is used for acquiring current and voltage data acquired by a current and voltage sensor;
a sensing module 602, configured to input current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
Optionally, the generator network comprises: a down-sampling layer, a data conversion layer and an up-sampling layer;
the sensing module 602 is specifically configured to:
the down-sampling layer receives input current and voltage data and generates first sampling data;
the data conversion layer receives input first sampling data and generates first conversion data;
the up-sampling layer receives the input first conversion data and generates virtual vibration data.
Optionally, the apparatus for sensing vibration data based on a current-voltage sensor further includes:
the first training module is used for forming a second training set by current and voltage sample data and real vibration sample data corresponding to the current and voltage sample data and inputting the second training set into the generator network so as to train the generator network;
the second training module is used for marking the vibration data output by the generator network as virtual vibration sample data, and forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data and inputting the third training set into the discriminator network so as to train the discriminator network;
and the third training module is used for forming the obtained generator network and the discriminator network into a countermeasure network, and inputting the first training set into the countermeasure network so as to train the countermeasure network.
Optionally, the arbiter network comprises a first arbiter network and a second arbiter network;
the second training module is specifically configured to:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to train the first discriminator network;
and inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network so as to train the second discriminator network.
Optionally, the third training module is specifically configured to: inputting the first training set into the countermeasure network, and alternately locking the parameters of the discriminator network or the generator network to update the parameters of the generator network or the discriminator network so that the loss function of the countermeasure network is smaller than a first threshold value, wherein the loss function of the countermeasure network is the combination of the loss functions of the first discriminator network and the second discriminator network.
Optionally, the first discriminator network is a convolutional neural network;
the second training module is specifically configured to:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to obtain the probability that the currently input data is the real vibration sample data;
and adjusting parameters of the first discriminator network to enable the probability to be larger than a second threshold value when the input data is real vibration sample data.
Optionally, the second discriminator network comprises: a first convolution layer, a time sequence data extraction layer, a second convolution layer and a normalization layer;
the second training module is specifically configured to:
inputting the real vibration sample data and the virtual vibration sample data into a first convolution layer to generate first convolution data;
inputting the first volume data into a time sequence data extraction layer to generate time sequence data;
inputting the time sequence data into the second convolution layer to generate second convolution data;
inputting the second convolution data into a normalization layer, and generating the prediction probability of the fault type of the real vibration sample data;
inputting the fault type of the real vibration sample data, comparing the prediction probability with the fault type of the real vibration sample data, and counting the prediction accuracy of a second discriminator;
and adjusting the parameters of the second discriminator network by taking the prediction accuracy rate larger than a third threshold value as a target.
According to the device for sensing the vibration data based on the current and voltage sensor, the generator network and the discriminator network form the countermeasure network to carry out countermeasure training, so that the calculation capacity of the generator network on the data is gradually improved, a better data conversion process can be realized under the condition that the data have hidden state characteristics, and the virtual sensing function of the generator network from the current and voltage data to the virtual vibration data is realized.
Fig. 7 illustrates a physical structure diagram of an electronic device, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a method of sensing vibration data based on a current-voltage sensor, the method comprising:
acquiring current and voltage data acquired by a current and voltage sensor;
inputting the current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, an embodiment of the present invention further provides a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute the method for sensing vibration data based on a current-voltage sensor, provided by the above-mentioned method embodiments, where the method includes:
acquiring current and voltage data acquired by a current and voltage sensor;
inputting the current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
In yet another aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented by a processor to perform the method for sensing vibration data based on a current-voltage sensor provided in the foregoing embodiments, and the method includes:
acquiring current and voltage data acquired by a current and voltage sensor;
inputting the current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for sensing vibration data based on a current-voltage sensor is characterized by comprising the following steps:
acquiring current and voltage data acquired by a current and voltage sensor;
inputting the current and voltage data into a generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
2. The method for sensing vibration data based on current and voltage sensors of claim 1, wherein the generator network comprises: a down-sampling layer, a data conversion layer and an up-sampling layer;
inputting the current-voltage data into a generator network to obtain virtual vibration data, comprising:
the down-sampling layer receives input current and voltage data and generates first sampling data;
the data conversion layer receives the input first sampling data and generates first conversion data;
the up-sampling layer receives the input first conversion data and generates virtual vibration data.
3. The method for sensing vibration data based on current and voltage sensors of claim 1, wherein the countermeasure network is trained by:
forming a second training set by current and voltage sample data and real vibration sample data corresponding to the current and voltage sample data, and inputting the second training set to the generator network so as to train the generator network;
marking the vibration data output by the generator network as virtual vibration sample data, and forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data and inputting the third training set to the discriminator network so as to train the discriminator network;
and forming a countermeasure network by the obtained generator network and the discriminator network, and inputting the first training set into the countermeasure network to train the countermeasure network.
4. The method for sensing vibration data based on current-voltage sensors of claim 3, wherein the discriminator network comprises a first discriminator network and a second discriminator network;
forming a third training set by the virtual vibration sample data, the real vibration sample data corresponding to the virtual vibration sample data and the fault type of the real vibration sample data, and inputting the third training set to the discriminator network so as to train the discriminator network, wherein the method comprises the following steps:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to train the first discriminator network;
and inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network so as to train the second discriminator network.
5. The method for sensing vibration data based on current and voltage sensors of claim 4, wherein the obtained generator network and the discriminator network are combined into a countermeasure network, and the first training set is input to the countermeasure network to train the countermeasure network, comprising:
inputting the first training set into the countermeasure network, and alternately locking the parameters of the discriminator network or the generator network to update the parameters of the generator network or the discriminator network so that the loss function of the countermeasure network is smaller than a first threshold value, wherein the loss function of the countermeasure network is the combination of the loss functions of the first discriminator network and the second discriminator network.
6. The method for sensing vibration data based on current and voltage sensors according to claim 4, wherein the first discriminator network is a convolutional neural network;
inputting the virtual vibration sample data and the real vibration sample data to a first discriminator network to train the first discriminator network, including:
inputting the virtual vibration sample data and the real vibration sample data into a first discriminator network to obtain the probability that the currently input data is the real vibration sample data;
and adjusting parameters of the first discriminator network to enable the probability to be larger than a second threshold value when the input data is real vibration sample data.
7. The method for sensing vibration data based on current-voltage sensors of claim 4, wherein said second discriminator network comprises: a first convolution layer, a time sequence data extraction layer, a second convolution layer and a normalization layer;
inputting the virtual vibration sample data, the real vibration sample data and the fault type thereof into a second discriminator network to train the second discriminator network, including:
inputting the real vibration sample data and the virtual vibration sample data into the first convolution layer to generate first convolution data;
inputting the first volume data into a time sequence data extraction layer to generate time sequence data;
inputting the time sequence data into the second convolution layer to generate second convolution data;
inputting the second convolution data into a normalization layer, and generating the prediction probability of the fault type of the real vibration sample data;
inputting the fault type of the real vibration sample data, comparing the prediction probability with the fault type of the real vibration sample data, and counting the prediction accuracy of a second discriminator;
and adjusting the parameters of the second discriminator network by taking the prediction accuracy rate larger than a third threshold value as a target.
8. An apparatus for sensing vibration data based on a current-voltage sensor, comprising:
the acquisition module is used for acquiring current and voltage data acquired by the current and voltage sensor;
the sensing module is used for inputting the current and voltage data into the generator network to obtain virtual vibration data;
the generator network and the discriminator network form a countermeasure network, and the countermeasure network is obtained by training a first training set formed by pre-collected current and voltage sample data, real vibration sample data corresponding to the current and voltage sample data and a fault type corresponding to the current and voltage sample data.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the method for sensing vibration data based on a current and voltage sensor according to any of claims 1 to 7.
10. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for sensing vibration data based on a current-voltage sensor according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011212653.8A CN112464990B (en) | 2020-11-03 | 2020-11-03 | Method and device for sensing vibration data based on current-voltage sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011212653.8A CN112464990B (en) | 2020-11-03 | 2020-11-03 | Method and device for sensing vibration data based on current-voltage sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112464990A true CN112464990A (en) | 2021-03-09 |
| CN112464990B CN112464990B (en) | 2024-05-10 |
Family
ID=74835878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011212653.8A Active CN112464990B (en) | 2020-11-03 | 2020-11-03 | Method and device for sensing vibration data based on current-voltage sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112464990B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113552450A (en) * | 2021-07-07 | 2021-10-26 | 上海交通大学 | Partial discharge source positioning method based on virtual sensor |
| CN113946920A (en) * | 2021-10-22 | 2022-01-18 | 大连海事大学 | Rolling bearing fault diagnosis method with unbalanced data and data set deviation |
| CN116087814A (en) * | 2023-01-28 | 2023-05-09 | 上海玫克生储能科技有限公司 | Method and device for improving voltage sampling precision and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module method for diagnosing faults, system and equipment based on depth convolution confrontation network |
| WO2020038548A1 (en) * | 2018-08-20 | 2020-02-27 | Telefonaktiebolaget Lm Ericsson (Publ) | Improving immune system of sites using generative adversarial networks and reinforcement learning |
-
2020
- 2020-11-03 CN CN202011212653.8A patent/CN112464990B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020038548A1 (en) * | 2018-08-20 | 2020-02-27 | Telefonaktiebolaget Lm Ericsson (Publ) | Improving immune system of sites using generative adversarial networks and reinforcement learning |
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module method for diagnosing faults, system and equipment based on depth convolution confrontation network |
Non-Patent Citations (1)
| Title |
|---|
| 王德文;杨凯华;: "基于生成式对抗网络的窃电检测数据生成方法", 电网技术, no. 02 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113552450A (en) * | 2021-07-07 | 2021-10-26 | 上海交通大学 | Partial discharge source positioning method based on virtual sensor |
| CN113946920A (en) * | 2021-10-22 | 2022-01-18 | 大连海事大学 | Rolling bearing fault diagnosis method with unbalanced data and data set deviation |
| CN116087814A (en) * | 2023-01-28 | 2023-05-09 | 上海玫克生储能科技有限公司 | Method and device for improving voltage sampling precision and electronic equipment |
| CN116087814B (en) * | 2023-01-28 | 2023-11-10 | 上海玫克生储能科技有限公司 | Method and device for improving voltage sampling precision and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112464990B (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112464990A (en) | Method and device for sensing vibration data based on current and voltage sensor | |
| CN110377984B (en) | Industrial equipment residual effective life prediction method and system and electronic equipment | |
| CN110321603B (en) | Depth calculation model for gas path fault diagnosis of aircraft engine | |
| US9811718B2 (en) | Method and a system for face verification | |
| CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
| CN109671070B (en) | Target detection method based on feature weighting and feature correlation fusion | |
| CN112673381B (en) | Method and related device for identifying confrontation sample | |
| Gao et al. | ASM1D-GAN: An intelligent fault diagnosis method based on assembled 1D convolutional neural network and generative adversarial networks | |
| CN110298374B (en) | Driving track energy consumption analysis method and device based on deep learning | |
| CN113065581B (en) | Vibration fault migration diagnosis method for reactance domain self-adaptive network based on parameter sharing | |
| CN114677412B (en) | Optical flow estimation method, device and equipment | |
| CN114760098A (en) | CNN-GRU-based power grid false data injection detection method and device | |
| CN113408340B (en) | Dual-polarization SAR small ship detection method based on enhanced feature pyramid | |
| CN110503113B (en) | Image saliency target detection method based on low-rank matrix recovery | |
| CN112200114A (en) | Fault diagnosis model training method and device, electronic equipment and storage medium | |
| CN110472268A (en) | A kind of bridge monitoring data modality recognition methods and device | |
| CN113541985B (en) | Internet of things fault diagnosis method, model training method and related devices | |
| CN116304853A (en) | Data classification method and system combining extreme learning machine and feature extraction | |
| EP3671635B1 (en) | Curvilinear object segmentation with noise priors | |
| CN117349708A (en) | Motor fault intelligent diagnosis method based on shallow feature fusion | |
| CN114169396B (en) | Training data generation model construction method and application for aircraft fault diagnosis | |
| CN111881439A (en) | Recognition model design method based on antagonism regularization | |
| Srimani et al. | Adaptive data mining approach for PCB defect detection and classification | |
| CN112560894A (en) | Improved 3D convolutional network hyperspectral remote sensing image classification method and device | |
| CN116400168A (en) | Power grid fault diagnosis method and system based on depth feature clustering |
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 |