CN115814686A - State monitoring method and system for laser gas mixing production system - Google Patents

Abstract

The invention provides a state monitoring method and system for a laser gas blending production system, and relates to the technical field of data processing. In the invention, a first state representative feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed are mined; matching second state focusing characteristic representations corresponding to each piece of reference production system operation data corresponding to the first state representative characteristic representation based on the first mapping relation information; calculating the matching degree of each second state focusing feature representation in the plurality of matched second state focusing feature representations and the first state focusing feature representation to determine the matching reference production system operation data; and determining a corresponding target operation state representative index according to a reference operation state representative index which is configured for the matched reference production system operation data in advance. Based on the above, the convenience and efficiency of state monitoring can be improved to a certain extent.

Description

State monitoring method and system for laser gas mixing production system

Technical Field

The invention relates to the technical field of data processing, in particular to a state monitoring method and system for a laser gas mixing production system.

Background

Laser gas belongs to radioactive gas and has application in various fields. The laser gas mixing production system is running, and the laser gas mixing production result is directly influenced. Therefore, abnormity monitoring needs to be performed on the laser gas mixing production process, but in the prior art, abnormity is monitored only when a large fault occurs, or more monitoring devices need to be deployed for comprehensive monitoring, so that the problems of convenience and low efficiency of state monitoring exist.

Disclosure of Invention

In view of the above, the present invention provides a method and a system for monitoring a status of a laser gas blending production system, so as to improve convenience and efficiency of status monitoring to a certain extent.

In order to achieve the above purpose, the embodiment of the invention adopts the following technical scheme:

a state monitoring method of a laser gas mixing production system comprises the following steps:

extracting the operation data of the production system to be analyzed;

loading the operation data of the production system to be analyzed to dig out a first state representation feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed by using a target operation data mining neural network, wherein the target operation data mining neural network is formed by network optimization by using an exemplary data combination and an exemplary state representation feature representation which have different operation state levels, the operation state levels corresponding to relevant exemplary data and non-relevant exemplary data in the exemplary data combination are consistent, the first state representation feature is used for reflecting the first operation state level corresponding to the operation data of the production system to be analyzed, and the first state focusing feature representation is used for reflecting operation data key information of the operation data of the production system to be analyzed at the first operation state level;

matching a second state focusing feature representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representative feature representation based on first mapping relation information corresponding to the reference production operation data, wherein the first mapping relation information is used for reflecting corresponding relation information between the state focusing feature representation and the state representative feature representation of each piece of reference production system operation data in the reference production operation data at a corresponding operation state level;

calculating the matching degree of each second state focusing feature representation in the plurality of matched second state focusing feature representations and the first state focusing feature representation, and determining matched reference production system operation data corresponding to the to-be-analyzed production system operation data in the reference production operation data based on a feature representation matching degree calculation value;

and determining a target operation state representative index corresponding to the operation data of the production system to be analyzed as an operation state representative index of the target laser gas blending production system according to a reference operation state representative index configured for the operation data of the matching reference production system in advance, wherein the first operation state layer is used for reflecting a corresponding state type and the target operation state representative index is used for reflecting a state value of the corresponding state type.

In some preferred embodiments, in the method for monitoring the state of the laser gas blending production system, the method further includes a step of determining the first mapping relationship information, where the step of determining includes:

extracting operation data mark information corresponding to each piece of reference production system operation data included in the reference production operation data;

loading each piece of reference production system operation data respectively to mine a neural network by using the target operation data and analyze state representation characteristic representation and state focusing characteristic representation corresponding to each piece of reference production system operation data;

determining reference production system operation data with the same state representative feature representation to construct mark corresponding relation information between the state representative feature representation and the operation data mark information corresponding to each piece of reference production system operation data with the state representative feature representation;

forming an initial layer mapping relation of the reference production operation data based on the constructed corresponding relation information of the plurality of marks;

forming a depth level mapping relation of the reference production operation data based on operation data mark information corresponding to each piece of reference production system operation data and corresponding relation information between the state focusing feature representation of the corresponding reference production system operation data;

and forming first mapping relation information of each piece of reference production system operation data included in the reference production operation data based on the initial level mapping relation and the depth level mapping relation.

In some preferred embodiments, in the above method for monitoring a state of a laser gas mixing and production system, the step of matching, based on first mapping relationship information corresponding to reference production operation data, a second state focusing characteristic representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representative characteristic representation includes:

determining a plurality of target operation data mark information corresponding to the first state representative feature representation based on the initial level mapping relationship included in the first mapping relationship information;

and determining second state focusing feature representations respectively corresponding to the target operation data mark information based on the depth-level mapping relation included in the first mapping relation information.

In some preferred embodiments, in the above method for monitoring the state of a laser gas mixing and production system, the step of determining, based on the initial level mapping relationship included in the first mapping relationship information, a plurality of pieces of target operation data flag information corresponding to the first state representative feature representation includes:

respectively carrying out matching degree calculation on a plurality of state representative feature representations in the initial level mapping relation included in the first mapping relation information and the first state representative feature representation so as to output corresponding feature representation matching coefficients;

determining a plurality of running data mark information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient with the maximum value in the initial level mapping relation;

and marking the determined running data mark information to form target running data mark information corresponding to the running data of the reference production system corresponding to the first running state level in the reference production running data.

In some preferred embodiments, in the above method for monitoring the state of the laser gas mixing and production system, the step of performing matching degree calculation on a plurality of state representative feature representations in the initial level mapping relationship included in the first mapping relationship information and the first state representative feature representation respectively to output corresponding feature representation matching coefficients includes:

extracting a characteristic representation cosine value between each state representative characteristic representation in the initial level mapping relation included in the first mapping relation information and the first state representative characteristic representation;

the step of determining, in the initial layer mapping relationship, a plurality of pieces of operating data flag information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient having the maximum value includes:

and in the initial level mapping relationship, determining a plurality of pieces of running data mark information corresponding to the state representation characteristic representation, wherein the characteristic representation cosine value does not exceed a preset contrast coefficient.

In some preferred embodiments, in the method for monitoring the state of the laser gas hybrid production system, the method for monitoring the state of the laser gas hybrid production system further includes an optimization step of the target operation data mining neural network, where the optimization step includes:

extracting a plurality of matching exemplary data sets;

analyzing the example production system operation data respectively included in the plurality of matching example data sets to output operation state level identification information corresponding to the example production system operation data;

performing feature space mapping processing on the running state level identification information of the target number to form a first exemplary state representative feature representation corresponding to each kind of running state level identification information;

determining relevant example data and non-relevant example data corresponding to first example data in example production system operation data included in the plurality of matching example data sets corresponding to the same operation state level identification information to form a plurality of example data combinations, wherein the first example data is randomly determined one example production system operation data in one matching example data set corresponding to the operation state level identification information;

and performing network optimization processing on the initial operation data mining neural network based on the exemplary data combination and the first exemplary state representative characteristic representation to form a corresponding target operation data mining neural network.

In some preferred embodiments, in the above method for monitoring the status of a laser gas blending production system, the step of determining relevant example data and non-relevant example data corresponding to the first example data from among example production system operation data included in the plurality of matching example data sets corresponding to the same operation status level identification information to form a plurality of example data combinations includes:

tagging first example production system operational data comprised by a first matching example data set belonging to a randomly determined one of the matching example data sets comprising the first matching example data set and a second matching example data set to form related example data corresponding to the first example data;

screening a second matching exemplary data subset corresponding to the same operating state level identification information as the first exemplary data from the various exemplary production system operating data included in the second matching exemplary data set;

matching out the non-relevant exemplary data with the maximum target number of data similarity with the first exemplary data in the second matching exemplary data subset;

and forming an exemplary data combination of the target number corresponding to the first exemplary data based on the first exemplary data and the related exemplary data included in the first matching exemplary data set and each non-related exemplary data corresponding to the matched first exemplary data.

In some preferred embodiments, in the above method for monitoring the state of a laser gas blending production system, the step of performing network optimization processing on the initial operation data mining neural network based on the exemplary data combination and the first exemplary state representative characteristic representation to form a corresponding target operation data mining neural network includes:

loading the example production system operational data to mine a second example state representative signature and a second example state focused signature representation corresponding to the example production system operational data using an initial operational data mining neural network;

respectively carrying out learning cost value determination processing on the first exemplary state representative characteristic representation, the second exemplary state representative characteristic representation and the second exemplary state focusing characteristic representation, and marking all determined learning cost values to form a first learning cost value of the operation data of the exemplary production system;

determining whether the first learning cost value matches a configured network optimization rule;

optimizing the network variables of the initial operation data mining neural network based on the first learning cost value under the condition that the first learning cost value is not matched with the configured network optimization rule, and further optimizing the optimized initial operation data mining neural network again through the operation data of the exemplary production system;

and under the condition that the first learning cost value is matched with the configured network optimization rule, marking the current initial operation data mining neural network to form a corresponding target operation data mining neural network.

In some preferred embodiments, in the above method for monitoring the state of a laser gas blending production system, the step of performing learning cost value determination processing on the first exemplary state representative feature representation, the second exemplary state representative feature representation and the second exemplary state focusing feature representation respectively, and marking all determined learning cost values to form a first learning cost value of the operation data of the laser gas blending production system includes:

based on the first exemplary state representative characteristic representation, performing learning cost value determination processing on the second exemplary state representative characteristic representation of the operation state level corresponding to the operation data of the exemplary production system, and outputting a state level learning cost value of the operation data of the exemplary production system;

performing combined learning cost value determination processing based on a second exemplary state focus characteristic representation of each of the exemplary production system operating data in the exemplary data combination corresponding to the operating state level corresponding to the exemplary production system operating data to output a combined learning cost value corresponding to the exemplary production system operating data;

and fusing the state level learning cost value corresponding to the exemplary production system operation data and the corresponding combined learning cost value to output a first learning cost value corresponding to the exemplary production system operation data.

The embodiment of the invention also provides a state monitoring system of the laser gas mixing production system, which comprises a processor and a memory, wherein the memory is used for storing the computer program, and the processor is used for executing the computer program so as to realize the state monitoring method of the laser gas mixing production system.

The method and the system for monitoring the state of the laser gas blending production system can extract the operation data of the production system to be analyzed; mining a first state representative feature representation and a first state focus feature representation of the production system operation data to be analyzed; matching a second state focusing feature representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representation feature representation based on the first mapping relation information; calculating the matching degree of each second state focusing feature representation in the plurality of matched second state focusing feature representations and the first state focusing feature representation to determine the matching reference production system operation data; and determining a corresponding target operation state representative index according to a reference operation state representative index which is configured for the matched reference production system operation data in advance. Based on the above content, after the first state representative feature representation and the first state focusing feature representation of the operation data of the production system to be analyzed are mined, matching search can be performed based on the first mapping relation information, so that the efficiency is higher.

In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

Fig. 1 is a block diagram of a state monitoring system of a laser gas blending production system according to an embodiment of the present invention.

Fig. 2 is a schematic flow chart illustrating steps included in a method for monitoring a state of a laser gas mixing production system according to an embodiment of the present invention.

Fig. 3 is a schematic diagram of modules included in a state monitoring device of a laser gas blending production system 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 only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the 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.

As shown in fig. 1, an embodiment of the present invention provides a state monitoring system for a laser gas blending production system. The state monitoring system of the laser gas blending production system can comprise a memory, a processor, other devices or components and the like which may be required.

In detail, the memory and the processor are electrically connected directly or indirectly to realize data transmission or interaction. For example, they may be electrically connected to each other via one or more communication buses or signal lines. The memory can have stored therein at least one software function (computer program) which can be present in the form of software or firmware. The processor may be configured to execute the executable computer program stored in the memory, so as to implement the method for monitoring the status of the laser gas blending production system according to the embodiment of the present invention (described later).

It should be understood that in some embodiments, the Memory may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Read Only Memory (EPROM), an electrically Erasable Read Only Memory (EEPROM), and the like. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), a System on Chip (SoC), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components.

It should be appreciated that in some embodiments, the condition monitoring system of the laser gas compounding production system can be a server with data processing capability.

With reference to fig. 2, an embodiment of the present invention further provides a method for monitoring a state of a laser gas blending production system, which can be applied to the state monitoring system of the laser gas blending production system. The method steps defined by the relevant flows of the state monitoring method of the laser gas mixing production system can be realized by the state monitoring system of the laser gas mixing production system.

The specific process shown in FIG. 2 will be described in detail below.

And step S110, extracting the operation data of the production system to be analyzed.

In the embodiment of the invention, the state monitoring system of the laser gas mixing production system can extract the operation data of the production system to be analyzed. The operation data of the production system to be analyzed includes production operation behavior record data (such as log data of equipment operation) of a plurality of production equipment in a plurality of time periods respectively, wherein the production equipment is included in the target laser gas blending production system.

Step S120, loading the operation data of the production system to be analyzed so as to mine a neural network by using the target operation data, and mining a first state representative feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed.

In an embodiment of the present invention, the state monitoring system of the laser gas blending production system may load the operation data of the production system to be analyzed, so as to mine a neural network by using the target operation data, and mine a first state representative feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed. The target operation data mining neural network is formed by performing network optimization by using an exemplary data combination with different operation state levels and an exemplary state representative feature representation, wherein the operation state levels corresponding to relevant exemplary data and non-relevant exemplary data in the exemplary data combination are consistent, the first state representative feature representation is used for reflecting a first operation state level (namely the operation state level of the operation data of the production system to be analyzed, and the operation state level can be defined according to actual requirements, such as good operation, normal operation, slight operation abnormality, abnormal operation, slight operation fault, serious operation fault and the like) corresponding to the operation data of the production system to be analyzed, and the first state focusing feature representation is used for reflecting operation data key information of the operation data of the production system to be analyzed under the first operation state level.

Step S130, matching a second state focusing feature representation corresponding to each piece of reference production system operation data in the plurality of pieces of reference production system operation data corresponding to the first state representative feature representation based on the first mapping relationship information corresponding to the reference production operation data.

In this embodiment of the present invention, the state monitoring system of the laser gas blending production system may match a second state focusing feature representation corresponding to each piece of reference production system operation data in the plurality of pieces of reference production system operation data corresponding to the first state representation feature representation based on first mapping relationship information corresponding to the reference production operation data. The first mapping relation information is used for reflecting corresponding relation information between the state focusing feature representation and the state representing feature representation of each piece of reference production system operation data in the reference production operation data at the corresponding operation state level.

Step S140, performing matching degree calculation on each of the second state focusing feature representations and the first state focusing feature representation in the plurality of matched second state focusing feature representations, and determining matching reference production system operation data corresponding to the to-be-analyzed production system operation data in the reference production operation data based on a feature representation matching degree calculation value.

In this embodiment of the present invention, the state monitoring system of the laser gas blending production system may perform matching degree calculation on each of the second state focusing feature representations in the plurality of matched second state focusing feature representations and the first state focusing feature representation, and then determine, based on a feature representation matching degree calculation value, matching reference production system operation data corresponding to the production system operation data to be analyzed in the reference production operation data (for example, the reference production system operation data corresponding to the second state focusing feature representation having the largest matching degree calculation value in the corresponding feature representation may be used as the matching reference production system operation data).

Step S150, determining a target operation state representative index corresponding to the operation data of the production system to be analyzed as the operation state representative index of the target laser gas mixed production system according to a reference operation state representative index configured for the operation data of the matching reference production system in advance.

In the embodiment of the present invention, the state monitoring system of the laser gas blending production system may determine, according to a reference operation state representative index configured for the operation data of the matching reference production system in advance, a target operation state representative index corresponding to the operation data of the production system to be analyzed, as the operation state representative index of the target laser gas blending production system. The first operation status layer is used for reflecting the corresponding status type and the target operation status representative index is used for reflecting the status value (for example, belonging to the aforementioned good degree in good operation, such as level 1, level 2, etc., and belonging to the aforementioned abnormal degree in abnormal operation, such as level 1, level 2, etc.) possessed by the corresponding status type.

Based on the above, after the first state representative feature representation and the first state focusing feature representation of the operation data of the production system to be analyzed are mined, matching search can be performed based on the first mapping relation information, so that the efficiency is higher, and in addition, analysis processing is directly performed based on the operation data of the production system to be analyzed.

It should be understood, however, that in some embodiments, the step S120 in the above description may further include the following specific sub-steps:

loading the production system operation data to be analyzed to utilize a target operation data mining neural network, and performing overall data mining processing and data association analysis on the production system operation data to be analyzed to output an overall data mining feature representation corresponding to the production system operation data to be analyzed (for example, the overall data mining processing can be realized by a key data mining sub-network included in the target operation data mining neural network, and the key data mining sub-network can include a plurality of filter matrixes to realize decimation filtering of data;

performing state level analysis on the operating data of the production system to be analyzed based on integral feature representation parameters of the integral data mining feature representation to output a first state representative feature representation corresponding to the operating data of the production system to be analyzed;

and performing knowledge screening processing on the overall data mining feature representation based on the first state representative feature representation to output a first state focusing feature representation which the to-be-analyzed production system operation data has at a corresponding first operation state level (for example, the first state representative feature representation and the overall data mining feature representation may be performed through a fully-connected sub-network included in the target operation data mining neural network, and the first state focusing feature representation which the to-be-analyzed production system operation data has at the first operation state level is output).

It should be understood that, in some embodiments, the step of performing state-level analysis on the to-be-analyzed production system operation data based on the overall feature representation parameters possessed by the overall data mining feature representation to output the first state representation corresponding to the to-be-analyzed production system operation data in the foregoing description may further include the following specific sub-steps:

performing state correspondence analysis on the overall data mining feature representation to analyze a first operation state level corresponding to the to-be-analyzed production system operation data based on a state possibility evaluation parameter reflected by a corresponding overall feature representation parameter (for example, a mapping output unit may map and output a body feature representation parameter included in the overall data mining feature representation to determine the first operation state level corresponding to the to-be-analyzed production system operation data, where the mapping output unit may include a softmax function having a classification function to classify and output the first operation state level corresponding to the to-be-analyzed production system operation data, that is, a prediction result);

based on the pre-configured target number, performing feature space mapping processing on the first operation state level to output a first state representative feature representation corresponding to the operation data of the production system to be analyzed (for example, the first operation state level may be embedded by an embedding unit to perform feature space mapping processing to output the first state representative feature representation corresponding to the operation data of the production system to be analyzed), where a parameter dimension of the first state representative feature representation is consistent with the target number, and the target number is determined based on the number of the operation state levels.

It should be understood that, in some embodiments, the method for monitoring the state of the laser gas blending production system may further include a step of determining the first mapping relationship information, and based on this, the determining step may further include the following specific sub-steps:

extracting operation data mark information corresponding to each piece of reference production system operation data included in the reference production operation data (the operation data mark information can be a configured number and the like);

loading each piece of reference production system operation data respectively to utilize the target operation data to mine a neural network and analyze state representation characteristic representations and state focusing characteristic representations (such as the related description above) corresponding to each piece of reference production system operation data;

carrying out deep classification processing on the state focusing feature representations corresponding to a plurality of pieces of reference production system operation data with the same state representative feature representation to form deep state representative feature representations corresponding to the plurality of pieces of reference production system operation data respectively;

based on the different state representative characteristic representations and the corresponding relationship information between the operation data mark information corresponding to the reference production system operation data with the corresponding state representative characteristic representation, forming an initial level mapping relationship corresponding to the reference production operation data;

based on the operation data flag information corresponding to the multiple pieces of reference production system operation data respectively having the same state representative feature representation and the corresponding relationship information between the different deep state representative feature representations formed by deep classification processing, forming a depth level mapping relationship corresponding to the reference production operation data (for example, the state focusing feature representations corresponding to the multiple pieces of reference production system operation data having the same state representative feature representation may be clustered to determine corresponding at least one clustering center data, so as to form the depth level mapping relationship based on the clustering center data, that is, the state focusing feature representations belonging to the same clustering center data have corresponding mapping relationship objects or corresponding relationship objects);

forming a bottom layer level mapping relation corresponding to the reference production operation data based on operation data mark information corresponding to each reference production operation data corresponding to the different deep layer state representative feature representations and corresponding relation information between the state focusing feature representations of the corresponding reference production system operation data;

based on the initial level mapping relationship, the depth level mapping relationship, and the bottom level mapping relationship, first mapping relationship information of each piece of reference production system operation data included in the reference production operation data is formed (that is, the first mapping relationship information may include the initial level mapping relationship, the depth level mapping relationship, and the bottom level mapping relationship), for example, in the mapping process, a first state representative feature representation of an operation state level corresponding to the piece of production system operation data to be analyzed may be determined first, that is, mapping is performed based on the initial level mapping relationship, then target operation data flag information corresponding to the target deep state representative feature representation may be determined, that is, mapping is performed based on the depth level mapping relationship, so as to obtain a second state focusing feature representation corresponding to each target operation data flag information, that is, mapping is performed based on the bottom level mapping relationship, and finally, based on matching of the feature representations, matching reference production system operation data that is most matched with the piece of production system operation data to be analyzed may be determined).

It should be understood that, in some embodiments, the method for monitoring the state of the laser gas compounding production system may further include a step of determining the first mapping relationship information, and based on this, the determining step may further include the following specific sub-steps:

extracting operation data mark information (as described above) corresponding to each piece of reference production system operation data included in the reference production operation data; loading each piece of the reference production system operation data respectively to utilize the target operation data to mine a neural network and analyze the state representation characteristic representation and the state focusing characteristic representation (as mentioned above) corresponding to each piece of the reference production system operation data; determining the reference production system operation data with the same state representative characteristic representation to construct mark correspondence information (as described above) between the state representative characteristic representation and the operation data mark information corresponding to each piece of reference production system operation data with the state representative characteristic representation; forming an initial level mapping relationship (as described above) of the reference production operation data based on the constructed corresponding relationship information of the plurality of marks; forming a depth level mapping relationship (as described above) of the reference production operation data based on operation data flag information corresponding to each piece of the reference production system operation data and corresponding relationship information between the state focusing feature representations of the corresponding reference production system operation data; based on the initial level mapping relationship and the depth level mapping relationship, first mapping relationship information of each piece of reference production system operation data included in the reference production operation data is formed (that is, the first mapping relationship information includes the initial level mapping relationship and the depth level mapping relationship).

It should be understood that, in some embodiments, the step S130 in the above description may further include the following specific sub-steps:

determining a plurality of target operation data mark information corresponding to the first state representative feature representation based on the initial level mapping relationship included in the first mapping relationship information;

and determining second state focusing feature representations respectively corresponding to the target operation data mark information based on the depth-level mapping relation included in the first mapping relation information.

It should be understood that, in some embodiments, the step of determining, based on the initial level mapping relationship included in the first mapping relationship information, a plurality of pieces of target operation data flag information corresponding to the first state representative feature representations may further include the following specific sub-steps:

respectively carrying out matching degree calculation on a plurality of state representative feature representations in the initial level mapping relation included in the first mapping relation information and the first state representative feature representation so as to output corresponding feature representation matching coefficients; determining a plurality of running data mark information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient with the maximum value in the initial level mapping relation; and marking the determined running data mark information to form target running data mark information corresponding to the running data of the reference production system corresponding to the first running state level in the reference production running data.

It should be understood that, in some embodiments, the step of performing matching degree calculation with the first state representative feature representation respectively for a plurality of state representative feature representations in the initial level mapping included in the pair of first mapping information in the foregoing to output corresponding feature representation matching coefficients may further include the following specific sub-steps:

extracting a characteristic representation cosine value (namely a cosine distance value between calculation vectors) between each state representative characteristic representation and the first state representative characteristic representation in the initial level mapping relation included in the first mapping relation information;

it should be understood that, in some embodiments, the step of determining, in the initial level mapping relationship, a plurality of pieces of operating data flag information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient having the maximum value may further include:

in the initial level mapping relationship, a plurality of pieces of operation data flag information corresponding to the state representative feature representation whose characteristic representation cosine value does not exceed a preset contrast coefficient are determined (or a plurality of pieces of operation data flag information corresponding to the state representative feature representation corresponding to the characteristic representation cosine value which is the smallest is selected).

It should be understood that, in some embodiments, the method for monitoring the state of the laser gas compounding production system may further include an optimization step of the target operation data mining neural network, and the optimization step may further include the following specific sub-steps:

extracting a plurality of matching exemplary data sets (for example, after obtaining a plurality of pieces of exemplary production system operation data, calculating data correlation degree, such as directly performing text data similarity calculation, for each two pieces of exemplary production system operation data in the plurality of pieces of exemplary production system operation data, or performing encoding processing to form corresponding exemplary encoding feature representations, then calculating matching degree between the exemplary encoding feature representations to obtain corresponding data correlation degree, and then combining the two pieces of exemplary production system operation data with the largest data correlation degree to form a corresponding one of the matching exemplary data sets);

analyzing the example production system operation data respectively included in the plurality of matching example data sets to output operation state level identification information corresponding to the example production system operation data (for example, the operation state level identification information may be used to reflect a real operation state level that the example production system operation data has, and may be formed based on a label);

performing feature space mapping processing on the running state level identification information of a target number to form a first exemplary state representative feature representation corresponding to each kind of running state level identification information (exemplarily, embedding processing may be performed based on an embedding unit to form a first exemplary state representative feature representation corresponding to each kind of running state level identification information);

determining relevant example data and non-relevant example data corresponding to first example data in example production system operation data included in the plurality of matching example data sets corresponding to the same operation state level identification information to form a plurality of example data combinations, wherein the first example data is randomly determined one example production system operation data in one matching example data set corresponding to the operation state level identification information;

and performing network optimization processing on the initial operation data mining neural network based on the exemplary data combination and the first exemplary state representative characteristic representation to form a corresponding target operation data mining neural network.

It should be understood that, in some embodiments, the step of determining relevant example data and non-relevant example data corresponding to the first example data from among the example production system operation data included in the plurality of matching example data sets corresponding to the same operation state level identification information in the above description to form a plurality of example data combinations may further include the following specific sub-steps:

marking first example production system operating data included in a first matching example data set to form first example data, and marking second example production system operating data included in the first matching example data set to form related example data corresponding to the first example data, wherein the first matching example data set belongs to one matching example data set randomly determined in the matching example data sets, and the matching example data sets include the first matching example data set and a second matching example data set (illustratively, other matching example data sets except the first matching example data set included in the matching example data sets can be used as second matching example data sets);

screening a second matching example data subset of the example production system operating data included in the second matching example data set that corresponds to the same operating state level identification information as the first example data (i.e., combining all example production system operating data that corresponds to the same operating state level identification information as the first example data to form the second matching example data subset);

in the second matching exemplary data subset, non-relevant exemplary data with the maximum target number of data similarity degrees with the first exemplary data are matched (namely, in the second matching exemplary data subset, the non-relevant exemplary data corresponding to the first exemplary data are all selected as the exemplary data with the maximum target number of data similarity degrees with the first exemplary data;

based on the first exemplary data and the relevant exemplary data included in the first matching exemplary data set and each non-relevant exemplary data corresponding to the matched first exemplary data, forming an exemplary data combination of a target number corresponding to the first exemplary data (exemplarily, in each exemplary data combination, the first exemplary data and the relevant exemplary data are included, and one piece of the non-relevant exemplary data is also included, where different exemplary data combinations include different non-relevant exemplary data).

It should be appreciated that in some embodiments, the step of performing a network optimization process on the initial operational data-mining neural network to form a corresponding target operational data-mining neural network based on the exemplary data combination and the first exemplary state representative characteristic representation in the above description may further include the following specific sub-steps:

loading the example production system operational data to mine a second example state representative signature and a second example state focused signature representation (as previously described) corresponding to the example production system operational data using an initial operational data mining neural network;

performing learning cost value determination processing on the first exemplary state representative feature representation, the second exemplary state representative feature representation and the second exemplary state focused feature representation respectively, and performing marking processing on all determined learning cost values to form a first learning cost value of the exemplary production system operation data;

determining whether the first learning cost value matches the configured network optimization rule (whether the first learning cost value matches the configured network optimization rule may include whether the first learning cost value is less than a pre-configured reference learning cost value, or whether the number of times of performing network optimization has exceeded a pre-configured reference number of times, etc.);

when the first learning cost value is not matched with the configured network optimization rule, optimizing the network variables of the initial operation data mining neural network (namely, performing loop optimization, which may refer to the related prior art and is not specifically limited and described herein) based on the first learning cost value, and further performing optimization processing again on the optimized initial operation data mining neural network through the operation data of the exemplary production system;

and under the condition that the first learning cost value is matched with the configured network optimization rule, marking the current initial operation data mining neural network to form a corresponding target operation data mining neural network (for example, the current initial operation data mining neural network can be directly used as the target operation data mining neural network, or the target operation data mining neural network with the same network variable value can be formed according to the current network variable value of the current initial operation data mining neural network).

It should be appreciated that in some embodiments, the steps of performing a learning cost value determination process on the first exemplary state representative feature representation, the second exemplary state representative feature representation, and the second exemplary state focused feature representation, respectively, and performing a labeling process on all determined learning cost values to form a first learning cost value of the exemplary production system operating data as described above may further include the following specific sub-steps:

performing a learning cost value determination process on the second exemplary state representative characteristic representation of the operation state level corresponding to the operation data of the exemplary production system based on the first exemplary state representative characteristic representation to output a state level learning cost value of the operation data of the exemplary production system (that is, the state level learning cost value may be determined based on a difference between the first exemplary state representative characteristic representation and the second exemplary state representative characteristic representation);

performing a combined learning cost value determination process based on the second exemplary state focused feature representation of each of the exemplary production system operating data in the exemplary data combination corresponding to the operating state level corresponding to the exemplary production system operating data to output a combined learning cost value corresponding to the exemplary production system operating data (illustratively, a degree of difference between the second exemplary state focused feature representation corresponding to the first exemplary data in the exemplary data combination and the second exemplary state focused feature representation corresponding to the related exemplary data may be calculated, and a degree of difference between the second exemplary state focused feature representation corresponding to the first exemplary data in the exemplary data combination and the second exemplary state focused feature representation corresponding to the non-related exemplary data may be calculated, and then a combined learning cost value may be determined based on the two degrees of difference, and illustratively, a difference between the former degree of difference and the latter degree of difference may be calculated, then superimposed with a configured bias parameter, and then a combined learning cost value corresponding to the exemplary production system operating data based on the superimposed result may be determined);

and performing fusion processing (such as weighted summation processing) on the state level learning cost value corresponding to the exemplary production system operation data and the corresponding combined learning cost value to output a first learning cost value corresponding to the exemplary production system operation data.

With reference to fig. 3, an embodiment of the present invention further provides a state monitoring apparatus for a laser gas mixing production system, which can be applied to the state monitoring system for the laser gas mixing production system. Wherein, the state monitoring device of the laser gas blending production system can comprise:

the system comprises an operation data extraction module, a data analysis module and a data analysis module, wherein the operation data extraction module is used for extracting the operation data of a production system to be analyzed, and the operation data of the production system to be analyzed comprises the production operation behavior record data of a plurality of production devices in a plurality of time periods respectively;

the operation data mining module is used for loading the operation data of the production system to be analyzed so as to mine a neural network by using target operation data, and mining a first state representative feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed, wherein the target operation data mining neural network is formed by performing network optimization by using an exemplary data combination and an exemplary state representative feature representation which have different operation state levels, the operation state levels corresponding to relevant exemplary data and non-relevant exemplary data in the exemplary data combination are consistent, the first state representative feature representation is used for reflecting the first operation state level corresponding to the operation data of the production system to be analyzed, and the first state focusing feature representation is used for reflecting operation data key information of the operation data of the production system to be analyzed under the first operation state level;

the characteristic representation mapping processing module is used for matching a second state focusing characteristic representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representative characteristic representation based on first mapping relation information corresponding to the reference production operation data, wherein the first mapping relation information is used for reflecting corresponding relation information between the state focusing characteristic representation and the state representative characteristic representation of each piece of reference production system operation data in the reference production operation data at a corresponding operation state level;

the characteristic representation matching module is used for calculating the matching degree of each second state focusing characteristic representation in the plurality of matched second state focusing characteristic representations and the first state focusing characteristic representation, and determining matched reference production system operation data corresponding to the to-be-analyzed production system operation data in the reference production operation data based on a characteristic representation matching degree calculation value;

and the running state representative index determining module is used for determining a target running state representative index corresponding to the running data of the production system to be analyzed according to a reference running state representative index configured for the running data of the matched reference production system in advance, and the target running state representative index is used as the running state representative index of the target laser gas mixing production system, and the first running state layer is used for reflecting a corresponding state type and the target running state representative index is used for reflecting a state value of the target laser gas mixing production system in the corresponding state type.

In summary, the method and the system for monitoring the state of the laser gas blending production system provided by the invention can extract the operation data of the production system to be analyzed; mining a first state representative feature representation and a first state focus feature representation of the production system operation data to be analyzed; matching a second state focusing feature representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representation feature representation based on the first mapping relation information; calculating the matching degree of each second state focusing feature representation in the plurality of matched second state focusing feature representations and the first state focusing feature representation to determine the matching reference production system operation data; and determining a corresponding target operation state representative index according to a reference operation state representative index which is configured for the matched reference production system operation data in advance. Based on the above content, after the first state representative feature representation and the first state focusing feature representation of the operation data of the production system to be analyzed are mined, matching search can be performed based on the first mapping relation information, so that the efficiency is higher.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A state monitoring method of a laser gas mixing production system is characterized by comprising the following steps:

extracting the operation data of the production system to be analyzed;

loading the operation data of the production system to be analyzed to dig out a first state representation feature representation and a first state focusing feature representation of the operation data of the production system to be analyzed by using a target operation data mining neural network, wherein the target operation data mining neural network is formed by network optimization by using an exemplary data combination and an exemplary state representation feature representation which have different operation state levels, the operation state levels corresponding to relevant exemplary data and non-relevant exemplary data in the exemplary data combination are consistent, the first state representation feature is used for reflecting the first operation state level corresponding to the operation data of the production system to be analyzed, and the first state focusing feature representation is used for reflecting operation data key information of the operation data of the production system to be analyzed at the first operation state level;

matching a second state focusing feature representation corresponding to each piece of reference production system operation data in a plurality of pieces of reference production system operation data corresponding to the first state representative feature representation based on first mapping relation information corresponding to the reference production operation data, wherein the first mapping relation information is used for reflecting corresponding relation information between the state focusing feature representation and the state representative feature representation of each piece of reference production system operation data in the reference production operation data at a corresponding operation state level;

calculating the matching degree of each second state focusing feature representation in the plurality of matched second state focusing feature representations and the first state focusing feature representation, and determining matched reference production system operation data corresponding to the to-be-analyzed production system operation data in the reference production operation data based on a feature representation matching degree calculation value;

and determining a target operation state representative index corresponding to the operation data of the production system to be analyzed as an operation state representative index of the target laser gas blending production system according to a reference operation state representative index configured for the operation data of the matching reference production system in advance, wherein the first operation state layer is used for reflecting a corresponding state type and the target operation state representative index is used for reflecting a state value of the corresponding state type.

2. The method for monitoring the status of a laser gas compounding production system of claim 1, further comprising a step of determining said first mapping relationship information, said step of determining comprising:

extracting operation data mark information corresponding to each piece of reference production system operation data included in the reference production operation data;

loading each piece of reference production system operation data respectively to mine a neural network by using the target operation data and analyze state representation characteristic representation and state focusing characteristic representation corresponding to each piece of reference production system operation data;

determining reference production system operation data with the same state representative feature representation to construct flag correspondence information between the state representative feature representation and the operation data flag information corresponding to each piece of reference production system operation data with the state representative feature representation;

forming an initial layer mapping relation of the reference production operation data based on the constructed corresponding relation information of the plurality of marks;

forming a depth level mapping relation of the reference production operation data based on operation data mark information corresponding to each piece of reference production system operation data and corresponding relation information between the state focusing feature representation of the corresponding reference production system operation data;

and forming first mapping relation information of each piece of reference production system operation data included in the reference production operation data based on the initial level mapping relation and the depth level mapping relation.

3. The method for monitoring the status of a laser gas mixing production system of claim 2, wherein the step of matching out the second status focus characterizations corresponding to each piece of the plurality of pieces of reference production system operating data corresponding to the first status representation characterization based on the first mapping relationship information corresponding to the reference production operating data comprises:

determining a plurality of pieces of target operation data mark information corresponding to the first state representative feature representation based on the initial layer mapping relationship included in the first mapping relationship information;

and determining second state focusing feature representations respectively corresponding to the plurality of target operation data mark information based on the depth level mapping relation included in the first mapping relation information.

4. The method of claim 3, wherein said step of determining a plurality of target operational data flag information corresponding to said first status representative feature representation based on an initial level mapping included in said first mapping information comprises:

respectively carrying out matching degree calculation on a plurality of state representative feature representations in the initial level mapping relation included in the first mapping relation information and the first state representative feature representation so as to output corresponding feature representation matching coefficients;

in the initial level mapping relation, determining a plurality of pieces of running data mark information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient with the maximum value;

and marking the determined running data mark information to form target running data mark information corresponding to the running data of the reference production system corresponding to the first running state level in the reference production running data.

5. The method for monitoring the status of a laser gas mixing production system according to claim 4, wherein the step of performing matching degree calculation on a plurality of status representative feature representations in the initial level mapping included in the first mapping information and the first status representative feature representation, respectively, to output corresponding feature representation matching coefficients includes:

extracting a characteristic representation cosine value between each state representative characteristic representation in the initial level mapping relation included in the first mapping relation information and the first state representative characteristic representation;

the step of determining, in the initial level mapping relationship, a plurality of pieces of operating data flag information corresponding to the state representation feature representation corresponding to the feature representation matching coefficient having the maximum value includes:

and in the initial level mapping relationship, determining a plurality of pieces of running data mark information corresponding to the state representation characteristic representation, wherein the characteristic representation cosine value does not exceed a preset contrast coefficient.

6. The method of claim 1 further comprising the step of optimizing said target operational data mining neural network, said optimizing step comprising:

extracting a plurality of matching exemplary data sets;

analyzing the example production system operation data respectively included in the plurality of matching example data sets to output operation state level identification information corresponding to the example production system operation data;

performing feature space mapping processing on the running state level identification information of the target number to form a first exemplary state representative feature representation corresponding to each kind of running state level identification information;

determining relevant example data and non-relevant example data corresponding to first example data in example production system operation data included in the plurality of matching example data sets corresponding to the same operation state level identification information to form a plurality of example data combinations, wherein the first example data is randomly determined one example production system operation data in one matching example data set corresponding to the operation state level identification information;

and performing network optimization processing on the initial operation data mining neural network based on the exemplary data combination and the first exemplary state representative characteristic representation to form a corresponding target operation data mining neural network.

7. The method of monitoring the status of a laser gas compounding production system of claim 6, wherein said step of determining relevant example data and non-relevant example data corresponding to a first example data from among the example production system operating data comprised in said plurality of sets of matching example data corresponding to the same said operating status level identification information to form a plurality of example data combinations comprises:

tagging first example production system operational data comprised by a first matching example data set belonging to a randomly determined one of the matching example data sets comprising the first matching example data set and a second matching example data set to form related example data corresponding to the first example data;

screening a second matching exemplary data subset corresponding to the same operating state level identification information as the first exemplary data from the various exemplary production system operating data included in the second matching exemplary data set;

matching out the non-relevant exemplary data with the maximum target number of data similarity with the first exemplary data in the second matching exemplary data subset;

and forming an exemplary data combination of the target number corresponding to the first exemplary data based on the first exemplary data and the related exemplary data included in the first matching exemplary data set and each non-related exemplary data corresponding to the matched first exemplary data.

8. The method of monitoring the status of a laser gas compounding production system of claim 6, wherein said step of network optimizing an initial operational data mining neural network based on said exemplary data combination and said first exemplary status representative characteristic representation to form a corresponding target operational data mining neural network comprises:

loading the example production system operational data to mine a second example state representative signature and a second example state focused signature representation corresponding to the example production system operational data using an initial operational data mining neural network;

respectively carrying out learning cost value determination processing on the first exemplary state representative characteristic representation, the second exemplary state representative characteristic representation and the second exemplary state focusing characteristic representation, and marking all determined learning cost values to form a first learning cost value of the operation data of the exemplary production system;

determining whether the first learning cost value matches a configured network optimization rule;

optimizing the network variables of the initial operation data mining neural network based on the first learning cost value under the condition that the first learning cost value is not matched with the configured network optimization rule, and further optimizing the optimized initial operation data mining neural network again through the operation data of the exemplary production system;

and under the condition that the first learning cost value is matched with the configured network optimization rule, marking the current initial operation data mining neural network to form a corresponding target operation data mining neural network.

9. The method of monitoring the status of a laser gas compounding production system of claim 8, wherein said step of processing said first exemplary state representative signature, said second exemplary state representative signature, and said second exemplary state focused signature to determine a learning cost value, respectively, and labeling all determined learning cost values to form a first learning cost value of said exemplary production system operating data comprises:

based on the first exemplary state representative characteristic representation, performing learning cost value determination processing on the second exemplary state representative characteristic representation of the operation state level corresponding to the operation data of the exemplary production system, and outputting a state level learning cost value of the operation data of the exemplary production system;

performing combined learning cost value determination processing based on a second exemplary state focus characteristic representation of each of the exemplary production system operating data in the exemplary data combination corresponding to the operating state level corresponding to the exemplary production system operating data to output a combined learning cost value corresponding to the exemplary production system operating data;

and fusing the state level learning cost value corresponding to the exemplary production system operation data and the corresponding combined learning cost value to output a first learning cost value corresponding to the exemplary production system operation data.

10. A condition monitoring system for a laser gas compounding production system comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program to perform the method of any one of claims 1-9.

Images