CN113963085B - State characterization method and device of industrial system and electronic equipment - Google Patents
State characterization method and device of industrial system and electronic equipment Download PDFInfo
- Publication number
- CN113963085B CN113963085B CN202111584989.1A CN202111584989A CN113963085B CN 113963085 B CN113963085 B CN 113963085B CN 202111584989 A CN202111584989 A CN 202111584989A CN 113963085 B CN113963085 B CN 113963085B
- Authority
- CN
- China
- Prior art keywords
- data
- index
- industrial system
- sequence
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/20—Drawing from basic elements, e.g. lines or circles
- G06T11/206—Drawing of charts or graphs
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
- G05B19/048—Monitoring; Safety
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
- General Factory Administration (AREA)
Abstract
The embodiment of the application provides a state representation method and device of an industrial system and electronic equipment, wherein the method comprises the following steps: reconstructing the training data through an unsupervised neural network model, and outputting a plurality of first reconstruction index sequences; determining first error data between each initial index sequence and the corresponding first reconstruction index sequence, and determining an early warning threshold according to the plurality of first error data; reconstructing the test data through an unsupervised neural network model, and outputting a plurality of second reconstruction index sequences; determining second error data between each index sequence to be characterized and the corresponding second reconstruction index sequence, and normalizing the plurality of second error data to obtain synthetic index data; and generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data. Therefore, the synthetic index obtained based on the unsupervised neural network model is used as the state representation of the industrial system, and the real-time state of the industrial system is indicated more accurately through the synthetic index.
Description
Technical Field
The present application relates to the field of communications technologies, and in particular, to a method and an apparatus for characterizing a state of an industrial system, and an electronic device.
Background
In the existing industrial production, an industrial production system often has multiple indexes, for example, when a shield machine works, the shield machine often needs to pay attention to the multiple indexes of an oil product, for example, indexes such as kinematic viscosity, dynamic viscosity, temperature, density and dielectric constant of the oil product, the indexes can determine the working state of the shield machine, the indexes are often independent indexes, monitoring personnel need to pay attention to real-time change data of each index independently, and certain risk is brought to industrial production.
In order to avoid risks, the prior art provides that each independent index is segmented and weighted and summed according to manual experience to obtain synthetic data to represent the state of the industrial production system. This approach often requires a large number of manual experience values, which are not only difficult to interpret, but also impossible to calculate quantitatively, resulting in a low accuracy of the synthesized data.
Disclosure of Invention
In order to solve the technical problem, embodiments of the present application provide a method and an apparatus for characterizing a state of an industrial system, and an electronic device.
In a first aspect, an embodiment of the present application provides a method for characterizing a state of an industrial system, where the method includes:
reconstructing training data through an unsupervised neural network model, and correspondingly outputting a plurality of first reconstruction index sequences, wherein the training data comprises a plurality of initial index sequences of an industrial system;
determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data;
reconstructing test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstructed index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system;
determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data;
and generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data.
In a second aspect, an embodiment of the present application provides a state characterization device for an industrial system, the device including:
the first reconstruction module is used for reconstructing training data through an unsupervised neural network model and correspondingly outputting a plurality of first reconstruction index sequences, wherein the training data comprises a plurality of initial index sequences of an industrial system;
the first determining module is used for determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function and determining an early warning threshold according to a plurality of first error data;
the second reconstruction module is used for reconstructing test data through the unsupervised neural network model and correspondingly outputting a plurality of second reconstruction index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system;
the second determining module is used for determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and normalizing the second error data to obtain synthesized index data;
and the generating module is used for generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory and a processor, where the memory is used to store a computer program, and the computer program executes the method for characterizing a state of an industrial system provided in the first aspect when the processor runs.
In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program runs on a processor, the computer program performs the method for characterizing the state of an industrial system provided in the first aspect.
According to the state characterization method and device for the industrial system and the electronic device, the unsupervised neural network model is used for reconstructing the training data, a plurality of first reconstruction index sequences are correspondingly output, and the training data comprises a plurality of initial index sequences of the industrial system; determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data; reconstructing test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstructed index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system; determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data; and generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data. Therefore, without depending on artificial experience, a plurality of data indexes of the industrial system are compressed into a synthetic index through the unsupervised neural network model, the synthetic index is used as the state representation of the industrial system, the accuracy of the synthetic index can be improved, and the real-time state of the industrial system can be more accurately indicated through the synthetic index.
Drawings
In order to more clearly explain the technical solutions of the present application, the drawings needed to be used in the embodiments are briefly introduced below, and it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope of protection of the present application. Like components are numbered similarly in the various figures.
FIG. 1 illustrates a flow diagram of a method for state characterization of an industrial system provided by an embodiment of the present application;
FIG. 2 is a block diagram illustrating error data provided by an embodiment of the present application;
FIG. 3 illustrates an index score distribution provided by an embodiment of the present application;
FIG. 4 is a diagram illustrating a comparison of index sequences provided in an embodiment of the present application;
FIG. 5 is a schematic diagram illustrating an iteration of training data and test data provided by an embodiment of the present application;
fig. 6 shows a schematic structural diagram of an industrial system state characterization device provided by an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments.
The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.
Hereinafter, the terms "including", "having", and their derivatives, which may be used in various embodiments of the present application, are intended to indicate only specific features, numbers, steps, operations, elements, components, or combinations of the foregoing, and should not be construed as first excluding the existence of, or adding to, one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing.
Furthermore, the terms "first," "second," "third," and the like are used solely to distinguish one from another and are not to be construed as indicating or implying relative importance.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the present application belong. The terms (such as those defined in commonly used dictionaries) should be interpreted as having a meaning that is consistent with their contextual meaning in the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein in various embodiments.
The prior art provides a method for representing the state of an industrial production system by using synthetic data obtained by dividing each independent index into sections according to manual experience and weighting and summing the sections. In this way, besides that the manual experience value is difficult to interpret and cannot be quantified, since the industrial production often involves a lot of links, and each experienced operator is only familiar with each link, the experience values of the whole system are difficult to be efficiently connected, and the experience values between the links may be inconsistent. The industrial production system is often influenced by natural conditions such as seasons, climate and the like, artificial experience often has certain hysteresis, and the change of experience values caused by the change of the natural conditions is difficult to capture in time. In summary, the effect of the comprehensive state characterization of the industrial production system in the prior art is poor.
Example 1
The embodiment of the disclosure provides a state representation method of an industrial system.
Specifically, referring to fig. 1, the method for characterizing the state of an industrial system includes:
and S101, reconstructing the training data through an unsupervised neural network model, and correspondingly outputting a plurality of first reconstruction index sequences.
In this embodiment, the training data includes a plurality of initial indicator sequences of the industrial system. It should be noted that the initial index sequence may be various index sequences for characterizing an industrial system, specifically, the index may be an oil index, and the oil index includes kinematic viscosity, dynamic viscosity, temperature, density, dielectric constant, and the like.
In this embodiment, the unsupervised neural network model may be a Long-Short Term Memory artificial neural network (LSTM) model. Specifically, step S101 includes the steps of:
coding each initial index sequence through a coding layer of the long-short term memory artificial neural network model to obtain a first compression index sequence;
and inputting the first compression index sequence into a hidden layer of the long-short term memory artificial neural network model, decompressing the compression index sequence through the hidden layer, and obtaining a corresponding first reconstruction index sequence.
For example, the initial index sequence S0 includes a first index X1, a second index X2, a. The long-short term memory artificial neural network model comprises an encoding layer (Encoder), a hidden layer (hidden representation) and a decoding layer (Decoder). Inputting the initial index sequence S0 into the long-short term memory artificial neural network model, and coding the initial index sequence S0 by the coding layer of the long-short term memory artificial neural network model to obtain a first compressed index sequence S1, wherein the first compressed index sequence S1 comprises first data Y1 and second data Y1The data Y2, the data-to-be-compressed and the data-to-be-compressed are the data Ym, the sequence length of the first compression indicator sequence S1 is m, and m is a positive integer smaller than n. Inputting the first compressed index sequence S1 into a hidden layer, decoding the first compressed index sequence S1 by the hidden layer to obtain a first reconstruction index sequence S2, wherein the sequence length of the first reconstruction index sequence S2 is n, and the first reconstruction index sequence S2 comprises indexes
Index of the composition
.
. In the embodiment, the long-short term memory artificial neural network model is an unsupervised learning neural network model, and the error between each index in the initial index sequence S0 and the corresponding index in the first reconstructed index sequence S2 is as small as possible.
Step S102, determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data.
In the present embodiment, the preset loss function includes the following formula 1:
equation 1:
wherein MAE represents error data of a sequence of index sequences, n represents the length of the initial index sequence,
representing the jth initial indicator in the initial indicator sequence,
representing the jth index of the first sequence of reconstruction indices.
The process of determining the first error data according to equation 1 includes: substituting the initial index sequence and the first reconstruction index sequence into the right side of the formula 1, and taking the calculated result as the value of the MAE.
In this embodiment, the determining the early warning threshold according to the plurality of first error data in step S102 includes the following steps:
obtaining the mean value and standard deviation of a plurality of first error data to determine a preset multiple;
and calculating a product value of the standard deviation and the preset multiple, and taking a sum value of the product value and the mean value as the early warning threshold value.
In this embodiment, one initial indicator sequence corresponds to one first reconstruction indicator sequence, each initial indicator sequence and the corresponding first reconstruction indicator sequence can determine a corresponding first error data through a preset loss function, and for a plurality of initial indicator sequences, a plurality of first error data can be correspondingly determined. In this embodiment, as can be seen from the theorem of large numbers, the error data calculated by the preset loss function basically follows normal distribution, and for the working state of the industrial system, the event needing attention is a small probability event, and the early warning threshold can be set to be data other than the multiple standard deviation by empirical data. For example, data outside of 3 standard deviations can be selected as the warning threshold.
In addition, the frequency of occurrence of each numerical value of the plurality of first error data can be counted, a corresponding histogram can be drawn, the mean value and the standard deviation of the first error data can be determined according to the drawn histogram, and the early warning threshold value can be determined by combining the mean value and the standard deviation.
Referring to fig. 2, fig. 2 is a histogram of error data, in which the abscissa represents the value of error data calculated according to a loss function, and the ordinate represents the frequency of occurrence of the value of each error data. If the mean value and the standard deviation of the error data are determined as a and b according to the histogram shown in fig. 2, the early warning threshold may be determined according to formula 2, where formula 2: j = a + b × 3, J denotes the warning threshold, a denotes the mean, b denotes the standard deviation. If J is 0.99, as calculated according to equation 2, the pre-warning threshold may be set to 0.99. In fig. 2, a straight line L1 indicates a position where the warning threshold is located, error data on the left side of the straight line L1 is approximate appearance data, and error data on the right side of the straight line L1 is small appearance data and can be regarded as abnormal data.
In this embodiment, to improve the data representation effect, normalization processing may be performed on the error data obtained by calculating the preset loss function, and the error data is mapped to corresponding data in 0 to 1. For example, the error data in the histogram shown in fig. 2 is normalized to obtain the index score distribution map shown in fig. 3. The abscissa in fig. 3 is the index score corresponding to the error data after normalization, and the ordinate in fig. 3 represents the frequency of occurrence of the index score. The index score on the left side of the straight line L2 is approximate probability occurrence data, and the index score on the right side of the straight line L2 is small probability occurrence data and is an abnormal event that needs attention.
In this embodiment, the determining the early warning threshold according to the plurality of first error data in step S102 includes the following steps:
generating a histogram according to the plurality of first error data, determining target error data with the occurrence probability smaller than a preset threshold value in the histogram, and determining the early warning threshold value according to the target error data.
In this embodiment, as can be seen from the theorem of majorities, the error data calculated by the preset loss function basically follows a normal distribution, and the event to be focused on is a small-probability event corresponding to the operating state of the industrial system. Referring to fig. 2 again, in fig. 2, a value with a low frequency of occurrence of error data may be selected as the warning threshold according to the distribution trend of the histogram.
And S103, reconstructing the test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstruction index sequences.
In this embodiment, the test data includes a plurality of index sequences to be characterized for the industrial system.
In the present embodiment, step S103 includes the following steps:
coding each index sequence to be characterized through the coding layer to obtain a second compression index sequence;
and inputting the second compression index sequence into the hidden layer, decompressing the second compression index sequence through the hidden layer, and obtaining a corresponding second reconstruction index sequence.
Step S104, determining second error data between each index sequence to be characterized and the corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data.
In this embodiment, the preset loss function includes the above formula 1, except that n in formula 1 represents the length of the index sequence to be characterized, and y representsjRepresents the jth initial index in the index sequence to be characterized,
the j-th index of the second reconstruction index sequence is represented.
And S105, generating state representation data of the industrial system according to the early warning threshold and the synthetic index data.
In this embodiment, the state representation data may be displayed through a display screen of the electronic device, so that a user can conveniently view the state representation data in time, and the display form may be a curve form or a histogram form, which is not limited herein.
In the present embodiment, step S105 includes:
drawing the synthetic index data into corresponding synthetic index curves according to a time sequence;
and marking the synthetic index curve according to the early warning threshold value to obtain the state representation data.
Referring to fig. 4, the time series data shown in fig. 4 are a density series C1, a dielectric constant series C2, a temperature series C3, a 40-degree kinematic viscosity series C4, a dynamic viscosity series C5, a real-time kinematic viscosity series C6, and a state characterization data series C7 of a certain industrial raw material collected by a sensor. The scattered point part in each sequence is the time of the industrial system which needs to be paid attention by the user, and as can be seen from the vertical dashed line box in fig. 4, each time the state representation data of the industrial system is increased, abnormality always occurs corresponding to at least one system index, and it can be seen that the state of the system can be well represented by the state representation data provided by the embodiment.
In this embodiment, to improve the adaptability of the training data and the test data to changes over time, the training data and the test data are updated over time.
Specifically, the obtaining of the training data and the test data includes:
determining subdata sets in corresponding time periods from the data set respectively according to a preset training time window and a plurality of preset test time windows to serve as a training set and a plurality of test sets;
determining a current training time point from the preset training time windows and the end time points of the preset test time windows according to a time sequence;
and taking the data of which the acquisition time is before the current training time point in the training set and the plurality of test sets as the training data, and taking the data of which the acquisition time is after the current training time point in the training set and the plurality of test sets as the test data.
In this embodiment, the unsupervised neural network model is continuously iterated, and is updated to adapt to different working condition changes according to the continuous iteration of the training data and the test data, and similarly, new working conditions occurring in the industrial system can be captured in time. Referring to fig. 5, the specific manner of model iteration is: the window for acquiring data each time slides by the length of the whole preset window, the length of the preset window comprises 1 preset training time window and 3 preset testing time windows, as shown in the direction marked by the arrow head in fig. 4, according to the time sequence, the data on the left of the current arrow time point is taken as training data each time, and the set on the right is taken as testing data. That is, at the time point corresponding to the first vertical arrow, the training set is used as training data, the test set 1 is used as test data, at the time point corresponding to the second vertical arrow, the training set plus the test set 1 is used as training data, the test set 2 is used as test data, at the time point corresponding to the third vertical arrow, the training set plus the test set 1 plus the test set 2 is used as training data, and the test set 3 is used as test data. Therefore, the training data and the test data can be determined in a real-time iteration mode, and the accuracy of the synthetic index can be ensured.
In the embodiment, without depending on artificial experience, a plurality of data indexes of the industrial system are compressed into one synthetic index through the unsupervised neural network model, and the synthetic index is used as the state representation of the industrial system, so that the accuracy of the synthetic index can be improved. Because the condition that experience values of all links are contradictory with each other does not occur due to no need of manual experience, and because the index data input into the unsupervised neural network model is updated in a updating mode, the method has better data real-time performance, can accurately capture the change of the index data, and better obtains more accurate synthetic indexes. In addition, for the manual experience value, when the experience value needs to be adjusted, multiple links are often linked to perform long-time debugging, the debugging process is low in efficiency, test data of the adjusted system are few, and many hidden dangers often exist.
In the state characterization method of the industrial system provided by this embodiment, the unsupervised neural network model is used to reconstruct training data, and a plurality of first reconstruction index sequences are correspondingly output, where the training data includes a plurality of initial index sequences of the industrial system; determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data; reconstructing test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstructed index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system; determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data; and generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data. Therefore, without depending on artificial experience, a plurality of data indexes of the industrial system are compressed into a synthetic index through the unsupervised neural network model, the synthetic index is used as the state representation of the industrial system, the accuracy of the synthetic index can be improved, and the real-time state of the industrial system can be more accurately indicated through the synthetic index.
Example 2
In addition, the embodiment of the disclosure provides a state representation device of an industrial system.
Specifically, as shown in fig. 6, the state characterization device 600 of the industrial system includes:
the first reconstruction module 601 is configured to reconstruct training data through an unsupervised neural network model, and correspondingly output a plurality of first reconstruction index sequences, where the training data includes a plurality of initial index sequences of an industrial system;
a first determining module 602, configured to determine, through a preset loss function, first error data between each initial indicator sequence and a corresponding first reconstruction indicator sequence, and determine an early warning threshold according to a plurality of the first error data;
the second reconstruction modeling block 603 is configured to reconstruct test data through the unsupervised neural network model, and correspondingly output a plurality of second reconstructed indicator sequences, where the test data includes a plurality of indicator sequences to be characterized of the industrial system;
a second determining module 604, configured to determine, through the preset loss function, second error data between each to-be-characterized index sequence and a corresponding second reconstruction index sequence, and perform normalization processing on a plurality of second error data to obtain synthesized index data;
and a generating module 605, configured to generate state representation data of the industrial system according to the early warning threshold and the synthetic index data.
In this embodiment, the state characterization device 600 of the industrial system further includes:
an acquisition module: determining subdata sets in corresponding time periods from the data set respectively according to a preset training time window and a plurality of preset test time windows to serve as a training set and a plurality of test sets;
determining a current training time point from the preset training time windows and the end time points of the preset test time windows according to a time sequence;
and taking the data of which the acquisition time is before the current training time point in the training set and the plurality of test sets as the training data, and taking the data of which the acquisition time is after the current training time point in the training set and the plurality of test sets as the test data.
In this embodiment, the unsupervised neural network model is a long-short term memory artificial neural network model, and the first reconstruction module 601 is further configured to encode each initial index sequence through an encoding layer of the long-short term memory artificial neural network model to obtain a first compressed index sequence;
and inputting the first compression index sequence into a hidden layer of the long-short term memory artificial neural network model, decompressing the compression index sequence through the hidden layer, and obtaining a corresponding first reconstruction index sequence.
In this embodiment, the second re-modeling block 603 is further configured to encode each index sequence to be characterized through the encoding layer to obtain a second compressed index sequence;
and inputting the second compression index sequence into the hidden layer, decompressing the second compression index sequence through the hidden layer, and obtaining a corresponding second reconstruction index sequence.
In this embodiment, the generating module 605 is further configured to draw the synthetic index data into a corresponding synthetic index curve according to a time sequence;
and marking the synthetic index curve according to the early warning threshold value to obtain the state representation data.
In this embodiment, the first determining module 602 is further configured to obtain a mean and a standard deviation of a plurality of the first error data to determine a preset multiple;
and calculating a product value of the standard deviation and the preset multiple, and taking a sum value of the product value and the mean value as the early warning threshold value.
In this embodiment, the first determining module 602 is further configured to generate a histogram according to a plurality of first error data, determine target error data with a probability of occurrence smaller than a preset threshold in the histogram, and determine the early warning threshold according to the target error data.
The state representation apparatus 600 of the industrial system according to the embodiment of the present invention may perform the steps of the state representation method of the industrial system according to embodiment 1, and is not described again to avoid repetition.
In the state characterization device of the industrial system provided by this embodiment, the unsupervised neural network model is used to reconstruct training data, and a plurality of first reconstruction index sequences are correspondingly output, where the training data includes a plurality of initial index sequences of the industrial system; determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data; reconstructing test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstructed index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system; determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data; and generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data. Therefore, without depending on artificial experience, a plurality of data indexes of the industrial system are compressed into a synthetic index through the unsupervised neural network model, the synthetic index is used as the state representation of the industrial system, the accuracy of the synthetic index can be improved, and the real-time state of the industrial system can be more accurately indicated through the synthetic index.
Example 3
Furthermore, an embodiment of the present disclosure provides an electronic device, which includes a memory and a processor, where the memory stores a computer program, and the computer program, when running on the processor, executes the state characterization method of the industrial system provided in the foregoing method embodiment.
The electronic device provided in the embodiment of the present invention may execute the steps of the state characterization method of the industrial system in embodiment 1, and is not described again to avoid repetition.
Example 4
The present application also provides a computer-readable storage medium on which a computer program is stored, which, when executed by a processor, implements the method of state characterization of an industrial system provided in embodiment 1.
In this embodiment, the computer-readable storage medium may be a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.
In this embodiment, the computer-readable storage medium may be the state characterization method of the industrial system provided in embodiment 1, and is not described herein again to avoid repetition.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative and, for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, each functional module or unit in each embodiment of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention or a part of the technical solution that contributes to the prior art in essence can 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 smart phone, a personal computer, a server, or a network device, etc.) 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.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention.
Claims (9)
1. A method of state characterization for an industrial system, the method comprising:
reconstructing training data through an unsupervised neural network model, and correspondingly outputting a plurality of first reconstruction index sequences, wherein the training data comprises a plurality of initial index sequences of an industrial system;
determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function, and determining an early warning threshold according to a plurality of first error data;
reconstructing test data through the unsupervised neural network model, and correspondingly outputting a plurality of second reconstructed index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system;
determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and performing normalization processing on a plurality of second error data to obtain synthetic index data;
generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data;
the generating of state characterization data of the industrial system according to the early warning threshold and the synthetic index data includes:
drawing the synthetic index data into corresponding synthetic index curves according to a time sequence;
and marking the synthetic index curve according to the early warning threshold value to obtain the state representation data.
2. The method of claim 1, wherein the obtaining of the training data and the test data comprises:
determining subdata sets in corresponding time periods from the data set respectively according to a preset training time window and a plurality of preset test time windows to serve as a training set and a plurality of test sets;
determining a current training time point from the preset training time windows and the end time points of the preset test time windows according to a time sequence;
and taking the data of which the acquisition time is before the current training time point in the training set and the plurality of test sets as the training data, and taking the data of which the acquisition time is after the current training time point in the training set and the plurality of test sets as the test data.
3. The method of claim 1, wherein the unsupervised neural network model is a long-short term memory artificial neural network model, and the reconstructing the training data by the unsupervised neural network model comprises:
coding each initial index sequence through a coding layer of the long-short term memory artificial neural network model to obtain a first compression index sequence;
and inputting the first compression index sequence into a hidden layer of the long-short term memory artificial neural network model, decompressing the compression index sequence through the hidden layer, and obtaining a corresponding first reconstruction index sequence.
4. The method of claim 3, wherein reconstructing test data through the unsupervised neural network model comprises:
coding each index sequence to be characterized through the coding layer to obtain a second compression index sequence;
and inputting the second compression index sequence into the hidden layer, decompressing the second compression index sequence through the hidden layer, and obtaining a corresponding second reconstruction index sequence.
5. The method of claim 1, wherein determining an early warning threshold from the plurality of first error data comprises:
obtaining the mean value and standard deviation of a plurality of first error data to determine a preset multiple;
and calculating a product value of the standard deviation and the preset multiple, and taking a sum value of the product value and the mean value as the early warning threshold value.
6. The method of claim 1, wherein determining an early warning threshold from the plurality of first error data comprises:
generating a histogram according to the plurality of first error data, determining target error data with the occurrence probability smaller than a preset threshold value in the histogram, and determining the early warning threshold value according to the target error data.
7. A state characterization device for an industrial system, the device comprising:
the first reconstruction module is used for reconstructing training data through an unsupervised neural network model and correspondingly outputting a plurality of first reconstruction index sequences, wherein the training data comprises a plurality of initial index sequences of an industrial system;
the first determining module is used for determining first error data between each initial index sequence and a corresponding first reconstruction index sequence through a preset loss function and determining an early warning threshold according to a plurality of first error data;
the second reconstruction module is used for reconstructing test data through the unsupervised neural network model and correspondingly outputting a plurality of second reconstruction index sequences, wherein the test data comprises a plurality of index sequences to be characterized of the industrial system;
the second determining module is used for determining second error data between each index sequence to be characterized and a corresponding second reconstruction index sequence through the preset loss function, and normalizing the second error data to obtain synthesized index data;
the generating module is used for generating state representation data of the industrial system according to the early warning threshold value and the synthetic index data; the generating module is further configured to draw the synthetic index data into corresponding synthetic index curves according to a time sequence;
and marking the synthetic index curve according to the early warning threshold value to obtain the state representation data.
8. An electronic device, comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, performs the method of state characterization of an industrial system according to any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that it stores a computer program which, when run on a processor, performs the method of state characterization of an industrial system according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111584989.1A CN113963085B (en) | 2021-12-23 | 2021-12-23 | State characterization method and device of industrial system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111584989.1A CN113963085B (en) | 2021-12-23 | 2021-12-23 | State characterization method and device of industrial system and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113963085A CN113963085A (en) | 2022-01-21 |
| CN113963085B true CN113963085B (en) | 2022-03-29 |
Family
ID=79473689
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111584989.1A Active CN113963085B (en) | 2021-12-23 | 2021-12-23 | State characterization method and device of industrial system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113963085B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022244A (en) * | 2016-05-16 | 2016-10-12 | 广东工业大学 | Unsupervised crowd abnormity monitoring and positioning method based on recurrent neural network modeling |
| CN110705692A (en) * | 2019-09-25 | 2020-01-17 | 中南大学 | Method for predicting product quality of industrial nonlinear dynamic process by long-short term memory network based on space and time attention |
| WO2021189844A1 (en) * | 2020-09-22 | 2021-09-30 | 平安科技（深圳）有限公司 | Detection method and apparatus for multivariate kpi time series, and device and storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6313730B2 (en) * | 2015-04-10 | 2018-04-18 | タタ コンサルタンシー サービシズ リミテッドＴＡＴＡ Ｃｏｎｓｕｌｔａｎｃｙ Ｓｅｒｖｉｃｅｓ Ｌｉｍｉｔｅｄ | Anomaly detection system and method |
| US20200104639A1 (en) * | 2018-09-28 | 2020-04-02 | Applied Materials, Inc. | Long short-term memory anomaly detection for multi-sensor equipment monitoring |
| US20210287071A1 (en) * | 2020-03-12 | 2021-09-16 | Morgan State University | Method and Apparatus for Augmented Data Anomaly Detection |
-
2021
- 2021-12-23 CN CN202111584989.1A patent/CN113963085B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022244A (en) * | 2016-05-16 | 2016-10-12 | 广东工业大学 | Unsupervised crowd abnormity monitoring and positioning method based on recurrent neural network modeling |
| CN110705692A (en) * | 2019-09-25 | 2020-01-17 | 中南大学 | Method for predicting product quality of industrial nonlinear dynamic process by long-short term memory network based on space and time attention |
| WO2021189844A1 (en) * | 2020-09-22 | 2021-09-30 | 平安科技（深圳）有限公司 | Detection method and apparatus for multivariate kpi time series, and device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113963085A (en) | 2022-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | Intelligent fault diagnosis method for rotating machinery via dictionary learning and sparse representation-based classification | |
| CN113095402B (en) | Code input-based generation countermeasure network fault detection method and system | |
| CN112257263B (en) | Equipment residual life prediction system based on self-attention mechanism | |
| CN111144548A (en) | Method and device for identifying working condition of pumping well | |
| CN103346797A (en) | Real-time compression method for gear fault signals | |
| Zhu et al. | Prediction of bearing remaining useful life based on DACN-ConvLSTM model | |
| Huang et al. | Application of optimized GM (1, 1) model based on EMD in landslide deformation prediction | |
| CN114648060A (en) | Fault signal standardization processing and classification method based on machine learning | |
| CN113963085B (en) | State characterization method and device of industrial system and electronic equipment | |
| CN117331015A (en) | Misalignment early warning method and system for CVT metering error | |
| Wang et al. | Fault diagnosis of rolling element bearing based on wavelet kernel principle component analysis-coupled hidden Markov model | |
| Wang et al. | Normal Sinkhorn Distance: A novel metric for evaluating generated signals and its application in mechanical fault diagnosis | |
| CN112836591B (en) | Method for extracting optical fiber early warning signal characteristics of oil gas long-distance pipeline | |
| CN115392715A (en) | Power utilization data risk assessment method, device, equipment and storage medium | |
| CN116776137A (en) | Data processing method and electronic equipment | |
| CN114565004A (en) | Method and device for eliminating abnormal scattered points of power curve of wind turbine generator | |
| CN109840479B (en) | Health state matching method and device | |
| CN111881259A (en) | Equipment fault probability evaluation method and system based on text mining | |
| CN112699608B (en) | Time sequence repairing method suitable for data loss caused by sensor power failure | |
| CN116125922B (en) | Complex industrial process monitoring method and system based on parallel dictionary learning | |
| CN110969180A (en) | Method and device for acquiring indicator diagram characteristic data | |
| CN116957361B (en) | Ship task system health state detection method based on virtual-real combination | |
| CN115267614B (en) | Detection method and system for intelligent transformer | |
| Yang et al. | Construction health indicator using physically-informed 1D-WGAN-GP joint attention LSTM-DenseNet method | |
| Wu et al. | Estimation of rolling bearing remaining useful life using DReLU-RA-ConvGRU model based on the encoder–decoder structure |
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 |