Industrial agent collaborative decision-making method and equipment

Abstract

The invention discloses an industrial agent collaborative decision-making method and device, and the method comprises the steps: carrying out the semantic coding of industrial data through a semantic encoder, and obtaining a semantic vector; decoding the semantic vector through a semantic decoder to obtain understandable semantic information of the industrial agent; performing fault prediction on the industrial equipment through a fault prediction agent and the understandable semantic information; and / or carrying out resource allocation on the industrial equipment through a resource scheduling agent and the understandable semantic information. Decision information delay is reduced, and decision accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial management, and particularly relates to an industrial intelligent agent collaborative decision-making method and device. BACKGROUND

[0002] At present, the data interaction amount between industrial devices is growing explosively with the vigorous development of the industrial internet. When processing complex industrial data, the traditional communication mode often faces problems of large data transmission amount and low information processing efficiency.

[0003] At present, in the intelligent manufacturing scene, the instructions and state information transmitted between industrial devices contain a large amount of redundant data, which not only occupies network bandwidth but also causes information processing delay, thereby affecting production efficiency. In addition, when the industrial intelligent agent makes decisions according to the data, due to inaccurate understanding of the information, decision-making errors may occur, such as misjudgment in fault prediction, unreasonable resource scheduling, and the like. SUMMARY

[0004] Embodiments of the present application provide an industrial intelligent agent collaborative decision-making method and device, which are used to solve the problems of information transmission delay and low decision-making accuracy in the industrial intelligent agent collaborative decision-making process.

[0005] Embodiments of the present application adopt the following technical solutions:

[0006] On the one hand, the embodiments of the present application provide an industrial intelligent agent collaborative decision-making method, which comprises: performing semantic encoding on industrial data through a semantic encoder to obtain a semantic vector; decoding the semantic vector through a semantic decoder to obtain understandable semantic information of the industrial intelligent agent; performing fault prediction on the industrial device through a fault prediction intelligent agent and the understandable semantic information; and / or performing resource allocation on the industrial device through a resource scheduling intelligent agent and the understandable semantic information.

[0007] In one example, the semantic encoding of the industrial data through the semantic encoder to obtain the semantic vector specifically comprises: extracting first key semantic features of the industrial data through a pre-trained semantic model; the first key semantic features are related to parameters for identifying whether the industrial device is in normal operation and / or parameters for identifying whether the production resource is in a normal use state; performing semantic data format conversion on the first key semantic features to obtain second key semantic features; and converting the second key semantic features into a semantic vector through word embedding.

[0008] In an example, the converting the second key semantic feature into a semantic vector through the word embedding specifically comprises: converting the second key semantic feature into an initial semantic vector through the word embedding; calculating a feature importance weight of the initial semantic vector through a multi-head attention mechanism; removing features with a feature importance weight lower than a preset threshold in the initial semantic vector to obtain a final semantic vector.

[0009] In an example, the performing fault prediction on the industrial equipment through the fault prediction agent and the understandable semantic information specifically comprises: extracting equipment state semantic information from the understandable semantic information by the fault prediction agent; comparing the equipment state semantic information with preset semantic identifiers of fault prediction models in terms of similarity; determining a fault prediction model with the highest similarity as a current fault prediction model; and inputting the equipment state semantic information into the current fault prediction model to obtain a fault analysis result of the industrial equipment.

[0010] In an example, after the inputting the understandable semantic information into the current fault prediction model to obtain the fault analysis result of the industrial equipment, the method further comprises: receiving a feedback result of the fault analysis result; when a decision adaptation level of the feedback result is a medium level, fine-tuning a semantic model of the semantic encoder to strengthen the key semantic feature and performing a light-weight update on the current fault prediction model; and when the decision adaptation level of the feedback result is a low level, reconfiguring the semantic model of the semantic encoder to update the key semantic feature and performing a full update on the current fault prediction model.

[0011] In an example, after the performing fault prediction on the industrial equipment through the fault prediction agent and the understandable semantic information, the method further comprises: when the fault analysis result is a result verification, retrieving a historical semantic vector with a matching degree exceeding a preset threshold from the semantic vector; comparing a historical fault analysis result corresponding to the historical semantic vector with the highest matching degree with the fault analysis result; and when the comparison is consistent, determining that the fault analysis result passes the verification.

[0012] In an example, the method further comprises: when the comparison is inconsistent, extracting equipment running state industrial data from the industrial data, inputting the equipment running state industrial data into a general fault prediction model to obtain a fault verification analysis result; comparing the fault verification analysis result with the fault analysis result; and when the comparison is consistent, determining that the fault analysis result passes the verification.

[0013] In an example, the method further comprises: when the comparison is inconsistent, retrieving a historical industrial data set corresponding to the semantic vector within a preset historical time period according to semantic labeling of historical industrial data; inputting the historical industrial data set and the equipment operation state industrial data into a pre-trained fluctuation risk analysis network model to obtain a fluctuation coefficient of the industrial equipment; and when the fluctuation coefficient is greater than a preset coefficient threshold and the fault analysis result is risky, determining that the fault analysis result passes verification.

[0014] In an example, the resource allocation for the industrial equipment by the resource scheduling agent and the understandable semantic information specifically comprises: the resource scheduling agent extracts resource state information from the understandable semantic information; and the resource scheduling agent inputs the resource state information into a resource scheduling model to obtain a resource allocation scheme for the industrial equipment.

[0015] In another aspect, an embodiment of the present application provides an industrial agent collaborative decision-making device, comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any of the above-described industrial agent collaborative decision-making methods.

[0016] The above-mentioned at least one technical solution adopted by the embodiments of the present application can achieve the following beneficial effects:

[0017] By performing semantic encoding and decoding on industrial data, the industrial agent can more efficiently and accurately understand and process information. By utilizing the advantages of semantic communication, the industrial agent can quickly extract key information in decision-making scenarios such as fault prediction and resource scheduling, as follows:

[0018] Optimizing decision-making accuracy: With the help of semantic communication, the industrial agent can more accurately understand industrial data. It can deeply mine the key content of industrial data and make more reasonable decisions based on its decision-making model. In the fault prediction scenario, it can more sensitively detect potential fault risks, with an accuracy improvement of 15% compared to traditional methods. In terms of resource scheduling, resource utilization is improved by 10%, effectively reducing production costs.

[0019] Improving information processing efficiency: With the powerful feature extraction capability of deep learning, it can deeply analyze various information sources such as text and images, accurately capture the semantic features and context information of the text, and efficiently convert the text into semantic vectors, compactly and comprehensively representing the semantic content of the information, greatly reducing the amount of data required for transmission. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the present application, some embodiments of the present application will be described in detail below with reference to the accompanying drawings, in which:

[0021] Figure 1 A flowchart of an industrial agent collaborative decision-making method provided by an embodiment of the present application is shown in FIG. 1.

[0022] Figure 2 A schematic diagram of an industrial agent collaborative decision-making system provided by an embodiment of the present application is shown in FIG. 2.

[0023] Figure 3 A structural schematic diagram of an industrial agent collaborative decision-making device provided by an embodiment of the present application is shown in FIG. 3. DETAILED DESCRIPTION

[0024] In order to make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be described in detail below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0025] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0026] Figure 1 A flowchart of an industrial agent collaborative decision-making method provided by an embodiment of the present application is shown in FIG. 1. Some input parameters or intermediate results in the flowchart allow manual intervention for adjustment to help improve accuracy.

[0027] The implementation of the analysis method involved in the embodiments of the present application can be a terminal device or a server, and the present application does not make special limitations thereon. For the convenience of understanding and description, the following embodiments are described in detail with the server as an example.

[0028] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server, and the present application does not make specific limitations thereon.

[0029] Figure 1 The flowchart in FIG. 1 includes the following steps:

[0030] S101: Perform semantic encoding on the industrial data through a semantic encoder to obtain a semantic vector.

[0031] The industrial sensor acquisition device collects various data such as supply chain management data, device operation data, and product quality detection data.

[0032] It should be noted that various types of industrial sensors are deployed at appropriate locations, and the sensors start collecting data according to the set parameters, and the collected raw data is denoised and normalized.

[0033] In some embodiments of the present application, the semantic communication module is responsible for semantic encoding of industrial data, which requires the use of deep learning technology. In the encoding stage, deep learning has strong feature extraction capability, and can deeply analyze various information sources such as text and images. For example, the Transformer structure semantic model is good at processing sequence data, and can accurately capture the semantic features and context information of the text, efficiently convert the text data into semantic vectors, compactly and comprehensively represent the semantic content of the information, and greatly reduce the amount of data required for transmission.

[0034] Based on this, the process of converting semantic vectors is as follows:

[0035] First, the first key semantic features of the industrial data are extracted by the pre-trained semantic model.

[0036] It should be noted that the first key semantic features are related to parameters indicating whether the industrial equipment is in normal operation, and / or parameters indicating whether the production resources are in a normal use state.

[0037] For example, the first key semantic features are radial vibration surge, high-frequency component rise of noise spectrum, and insufficient material inventory.

[0038] Then, the first key semantic features are converted into second key semantic features through semantic data format conversion.

[0039] Finally, the second key semantic features are converted into semantic vectors through word embedding.

[0040] For example, the word embedding technology can be Word2Vec, GloVe, etc.

[0041] Further, the semantic vectors are optimized and encoded to realize effective compression and encryption of data, which not only improves the transmission efficiency, but also guarantees the security of the information.

[0042] Based on this, the second key semantic features are converted into initial semantic vectors through word embedding.

[0043] The feature importance weight of the initial semantic vector is calculated through the multi-head attention mechanism.

[0044] In the initial semantic vector, features with a feature importance weight lower than a preset threshold are removed to obtain a final semantic vector.

[0045] Therefore, by focusing on key semantic components through the attention mechanism, the vector can be compressed while retaining the core semantics, resulting in a smaller and more compact dimension, filtering out the most critical feature dimensions for semantic expression and discarding secondary information.

[0046] S102: decoding the semantic vector through the semantic decoder to obtain understandable semantic information of the industrial agent.

[0047] It should be noted that the semantic encoder extracts key feature information of the equipment state data according to the characteristics of the industrial data, and converts the original data into a simplified semantic representation. In the decoding stage, a model based on a generative adversarial network (GAN) or a variational autoencoder (VAE) can effectively deal with channel noise and interference during decoding, and can correct and recover the signal. At the same time, it can learn the latent distribution of semantic information, thereby generating high-quality and accurate decoding results, and restoring the received semantic data to detailed information that can be understood by the industrial agent.

[0048] Among them, the understandable semantic information is transmitted through a wireless network or a wired network to the industrial agent decision module.

[0049] It should be noted that during transmission, a reliable transmission protocol is used to ensure the integrity and accuracy of the data. The existing industrial communication protocol, such as OPC UA, is used to embed the semantic data transmission mechanism in its data transmission layer, achieving compatibility with existing industrial protocols.

[0050] S103: performing fault prediction on the industrial equipment through the fault prediction agent and the understandable semantic information; and / or, obtaining resource allocation of the industrial equipment through the resource scheduling agent and the understandable semantic information.

[0051] In some embodiments of the present application, the industrial agent decision module includes a plurality of agents, each agent receiving information transmitted by the semantic communication module according to its own function and task, and performing decision processing.

[0052] Among them, the fault prediction agent monitors the operation of the industrial equipment, detects potential faults, and avoids the impact of equipment sudden failure on production. The semantic communication module provides the fault prediction agent with analyzed and processed equipment state semantic information. After receiving the semantic information, the fault prediction agent quickly calls the fault prediction model trained by a large amount of historical data and machine learning algorithms (such as convolutional neural networks), identifies the subtle differences between normal and fault states of the equipment, comprehensively analyzes the current equipment state semantic information, and judges the fault risk according to the learned patterns and rules. Once it is determined that there is a potential fault, the agent immediately issues a warning and gives preliminary fault causes and treatment suggestions.

[0053] The resource scheduling agent can obtain the semantic state of the current resources, including the idle condition of the equipment, the inventory level of raw materials, the work arrangement of personnel, etc. Through comprehensive analysis of this information, the resource scheduling agent can develop an optimal resource allocation scheme, reasonably arrange the use of raw materials, and ensure the continuity and efficiency of the production process.

[0054] In some embodiments of the present application, the processing process for the resource scheduling agent is as follows:

[0055] The resource scheduling agent extracts resource state information from the understandable semantic information, inputs the resource state information into the resource scheduling model, and obtains the resource allocation scheme of the industrial equipment.

[0056] In some embodiments of the present application, the processing process for the fault prediction agent is as follows:

[0057] If the data characteristics are uniform and the model itself has strong adaptability, the optimization of a single fault prediction model is more concise and efficient, but when the scene is complex and the data is discrete, multi-model screening is an effective means to improve accuracy. The core is to make the model capability fit the diverse manifestations of actual faults.

[0058] Based on this, the fault prediction agent extracts device state semantic information from understandable semantic information.

[0059] Then, the device state semantic information is compared with the preset semantic identifier of the fault prediction model in terms of similarity.

[0060] It should be noted that the system pre-stores multiple fault prediction models, and each model has a corresponding preset semantic identifier (i.e., a description of the fault feature that the model is good at identifying).

[0061] For example, the fault identification feature of fault prediction model A is bearing wear, and the identification feature is a sudden increase in radial vibration and an increase in high-frequency components of noise spectrum. The fault identification feature of fault prediction model B is cooling system failure, and the identification feature is a continuous rise in winding temperature and abnormal rotation speed of the cooling fan.

[0062] Then, the fault prediction model with the highest similarity is determined as the fault prediction model for this time;

[0063] Finally, the device state semantic information is input into the fault prediction model for this time, and the fault analysis result of the industrial equipment is obtained.

[0064] In some embodiments of the present application, due to the diversity of industrial data, the pre-trained fault prediction model may not be able to accurately identify the situation, so the fault analysis result is compared with the historical fault analysis result under the condition of triggering result verification or iterative optimization, so that the decision of the agent is more reliable.

[0065] It should be noted that the data storage module plays the role of a data warehouse in the industrial intelligent system, and it is responsible for storing various important information generated in the industrial production process, including industrial data, semantic coding rules, and agent decision-making models, etc.

[0066] Based on this, when the fault analysis result is the result verification, the historical semantic vector with a matching degree exceeding a preset threshold is searched.

[0067] It should be noted that when there is no historical semantic vector exceeding the preset threshold, the operation of inputting the equipment running state industrial data into the general fault prediction model is performed.

[0068] The historical fault analysis result corresponding to the historical semantic vector with the highest matching degree is compared with the fault analysis result.

[0069] When the comparison is consistent, it is determined that the fault analysis result verification is passed.

[0070] It should be noted that when the comparison is inconsistent, the possible reason is that the current semantic vector has not appeared before, so when the current semantic vector and the historical semantic vector are not completely the same, there is a possibility that the analysis results are inconsistent. Therefore, when the two are inconsistent, it cannot be directly determined that the fault analysis result is wrong.

[0071] Based on this, when the comparison is inconsistent, the equipment running state industrial data is extracted from the industrial data, the equipment running state industrial data is input into the general fault prediction model, and the fault verification analysis result is obtained;

[0072] Then, the fault verification analysis result is compared with the fault analysis result.

[0073] Finally, when the comparison is consistent, it is determined that the fault analysis result verification is passed.

[0074] It should be noted that when the above two are inconsistent, since the general fault prediction model may not be able to accurately identify the new semantic vector without training, it is also impossible to directly determine that the fault analysis result is wrong.

[0075] In addition, when storing historical industrial data, the data storage module adopts a semantic annotation method, so that the data has a clear semantic meaning, facilitating subsequent data retrieval and utilization.

[0076] Taking the device operation history data as an example, the data storage module will perform detailed semantic annotation on these data. The annotation content includes the device operation stage, such as temperature, pressure, humidity and other environmental factors, and the operation parameters of the device. Through these semantic annotations, the agent can quickly locate and obtain relevant data support when making decisions. When the fault prediction agent needs to analyze the running state of the industrial equipment under a specific working condition, it can quickly filter out the historical data that meets the conditions through semantic annotation, and perform in-depth analysis and comparison. This efficient data retrieval method greatly improves the decision-making efficiency and accuracy of the agent.

[0077] Based on this, when the comparison is inconsistent, the historical industrial data set corresponding to the semantic vector within the preset historical time length is retrieved according to the semantic annotation of the historical industrial data.

[0078] For example, when the historical industrial data A is the winding temperature, the semantic annotation is the winding temperature rise, and if the semantic vector a includes the winding temperature rise, the target historical industrial data corresponding to the semantic vector a includes the historical industrial data A.

[0079] When the historical industrial data B is the cooling fan speed, the semantic annotation is the cooling fan speed anomaly, and if the semantic vector includes the cooling fan speed anomaly, the target historical industrial data corresponding to the semantic vector a also includes the historical industrial data B.

[0080] Then, the historical industrial data set and the device running state industrial data are input into the pre-trained fluctuation hidden danger analysis network model to obtain the fluctuation coefficient of the industrial equipment.

[0081] It should be noted that the fluctuation hidden danger analysis network model can be obtained by training the convolutional neural network structure.

[0082] When the fluctuation coefficient is greater than the preset coefficient threshold and the fault analysis result is related to the risk, it is determined that the fault analysis result verification passes.

[0083] It should be noted that each type of data of the industrial equipment usually does not have a large fluctuation, therefore, when the fluctuation coefficient of the industrial equipment is large, it can be determined that the industrial equipment has an abnormality, therefore, if the fault analysis result at this time is that the industrial equipment has a certain risk, it is determined that the verification passes and an alarm is sent in time. Otherwise, it is determined that the fault analysis result verification does not pass, and the industrial data and the semantic vector are sent to the management user terminal, so as to combine human judgment and processing.

[0084] In summary, the actual fault of the industrial equipment can be found in time without missing or misjudging, and the participation of manpower can be reduced.

[0085] It should be noted that the semantic coding rule is also an important part of the data storage module, and the semantic coding rule defines how to convert the original data into information with semantic meaning, which is the basis of semantic communication and agent decision-making. The data storage module will properly save these rules for calling and updating when needed to ensure the performance and accuracy of the model.

[0086] In some embodiments of the present application, according to the decision result of the agent, the corresponding operation is performed. For example, when the fault prediction agent judges that the equipment has a fault risk, the system issues an alarm and starts the corresponding maintenance process. In the process of executing the decision, the data of the actual execution effect is collected and fed back to the semantic communication module and the industrial agent decision-making module, which is used to optimize the semantic coding rule and the agent decision-making model.

[0087] Based on this, the feedback result of the fault analysis result is received.

[0088] When the decision adaptation level of the feedback result is medium, the semantic model of the semantic encoder is fine-tuned to strengthen the key semantic features, and the fault prediction model this time is updated.

[0089] When the decision adaptation level of the feedback result is low, the semantic model of the semantic encoder is reconstructed to update the key semantic features, and the fault prediction model this time is updated.

[0090] It should be noted that although the embodiments of the present application are introduced and described in sequence with reference to steps S101 to S103, Figure 1 but this does not mean that steps S101 to S103 must be executed in strict sequence. The reason why the embodiments of the present application introduce and describe steps S101 to S103 in sequence as shown in Figure 1 is to facilitate the understanding of the technical solution of the embodiments of the present application by those skilled in the art. In other words, in the embodiments of the present application, the sequence between steps S101 to S103 can be adjusted appropriately according to actual needs.

[0091] Through the method, Figure 1 by introducing semantic communication technology, the efficiency and accuracy of industrial data processing are improved, the decision-making process of industrial agents is optimized, and the overall efficiency of industrial production is improved. The specific goals include reducing data transmission volume and network load; enhancing the understanding ability of industrial agents to information and making more scientific and reasonable decisions; improving the real-time performance and reliability of the system to meet the stringent requirements of industrial scenarios.

[0092] Based on this, by encoding and decoding the industrial data semantically, the industrial agent can more efficiently and accurately understand and process information, and using the advantages of semantic communication, the industrial agent can quickly extract key information and make better decisions in decision-making scenarios such as fault prediction and resource scheduling. The system includes a semantic communication module, an industrial agent decision-making module, and the like, and the method covers data semantic processing, decision execution, and the like, effectively improving industrial production efficiency and reducing energy consumption, solving the problems of communication and decision-making in industrial environments, as follows:

[0093] Optimize decision accuracy: With the help of semantic communication, the industrial agent can better understand the industrial data. It can deeply mine the key content of the industrial data and make more reasonable decisions based on its decision-making model. In the fault prediction scenario, it can more sensitively detect potential fault risks, with an accuracy of 15% higher than traditional methods; in terms of resource scheduling, the resource utilization rate is increased by 10%, effectively reducing production costs.

[0094] Improve information processing efficiency: With the powerful feature extraction capability of deep learning, it can deeply analyze various information sources such as text and images, accurately capture the semantic features and context information of the text, and efficiently convert the text into semantic vectors, compactly and comprehensively representing the semantic content of the information, greatly reducing the amount of data required for transmission.

[0095] Further, the semantic vectors of industrial data are optimized and encoded to remove redundant information, greatly reducing data transmission volume. This not only speeds up data transmission, but also reduces network burden, making industrial data interaction more smooth and efficient, ensuring timely transmission and reception of information between industrial production links.

[0096] Facilitate data retrieval and utilization: When storing historical industrial data, the data storage module uses semantic annotation, making the data have clear semantic meaning, facilitating subsequent data retrieval and utilization.

[0097] Enhance system adaptability: The system and method of the present application can be compatible with existing industrial protocols, facilitating application in existing industrial environments. At the same time, through real-time feedback mechanism, the semantic encoding and decoding mechanism and the agent decision-making model can be continuously optimized according to the actual production situation, adapting to different industrial scene needs.

[0098] More intuitively, Figure 2 A schematic diagram of an industrial agent collaborative decision-making system provided by an embodiment of the present application.

[0099] In Figure 2 , the collaborative process between industrial data collection, industrial data semantic transmission, industrial agent decision-making tasks, and data storage modules is shown.

[0100] Based on the same idea, some embodiments of the present application also provide a device and a non-volatile computer storage medium corresponding to the above method.

[0101] Figure 3 A structural schematic diagram of an industrial agent collaborative decision-making device provided by some embodiments of the present application includes:

[0102] at least one processor; and

[0103] a memory in communication with the at least one processor; wherein

[0104] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the industrial agent collaborative decision-making method according to any one of the above.

[0105] Some embodiments of the present application provide a non-volatile computer storage medium of an industrial agent collaborative decision-making method, which stores computer executable instructions, and the computer executable instructions can perform the industrial agent collaborative decision-making method according to any one of the above.

[0106] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts of each of the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0107] The device and medium provided by the embodiments of the present application correspond to the method, so the device and medium also have the similar beneficial technical effects as the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here.

[0108] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0109] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0110] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0111] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0112] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0113] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory. The memory is an example of computer-readable media.

[0114] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0115] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0116] The above only describes the embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modification, equivalent replacement, improvement, etc. made within the technical principles of the present application shall fall within the protection scope of the present application.

Claims

1. An industrial agent collaborative decision-making method, characterized in that, The method comprises: semantic encoding of industrial data by a semantic encoder to obtain a semantic vector; decoding the semantic vector by a semantic decoder to obtain understandable semantic information of an industrial agent; fault prediction of industrial equipment by a fault prediction agent and the understandable semantic information; and / or, resource allocation of industrial equipment by a resource scheduling agent and the understandable semantic information.

2. The method of claim 1, wherein, The semantic encoding of industrial data by a semantic encoder to obtain a semantic vector specifically comprises: extracting first key semantic features of industrial data by a pre-trained semantic model; the first key semantic features are related to parameters indicating whether the industrial equipment is in normal operation, and / or parameters indicating whether the production resources are in normal use state; performing semantic data format conversion on the first key semantic features to obtain second key semantic features; converting the second key semantic features into a semantic vector by word embedding.

3. The method of claim 2, wherein, The conversion of the second key semantic features into a semantic vector by word embedding specifically comprises: converting the second key semantic features into an initial semantic vector by word embedding; calculating feature importance weights of the initial semantic vector by a multi-head attention mechanism; removing features with feature importance weights lower than a preset threshold in the initial semantic vector to obtain a final semantic vector.

4. The method of claim 1, wherein, The fault prediction of industrial equipment by a fault prediction agent and the understandable semantic information specifically comprises: the fault prediction agent extracts equipment state semantic information from the understandable semantic information; comparing the equipment state semantic information with preset semantic identifiers of fault prediction models; determining the fault prediction model with the highest similarity as the current fault prediction model; inputting the equipment state semantic information into the current fault prediction model to obtain a fault analysis result of the industrial equipment.

5. The method of claim 4, wherein, After inputting the understandable semantic information into the current fault prediction model to obtain a fault analysis result of the industrial equipment, the method further comprises: receiving a feedback result of the fault analysis result; when the decision adaptation level of the feedback result is medium, fine-tuning the semantic model of the semantic encoder to strengthen the key semantic features and perform lightweight update on the current fault prediction model; when the decision adaptation level of the feedback result is low, reconfiguring the semantic model of the semantic encoder to update the key semantic features and perform full update on the current fault prediction model.

6. The method of claim 1, wherein, After the fault prediction of industrial equipment by a fault prediction agent and the understandable semantic information, the method further comprises: when the fault analysis result is result verification, retrieving a historical semantic vector with a matching degree exceeding a preset threshold with the semantic vector; comparing the historical fault analysis result corresponding to the historical semantic vector with the highest matching degree with the fault analysis result; when the comparison is consistent, determining that the fault analysis result passes the verification.

7. The method of claim 6, wherein, The method further comprises: When the comparison is inconsistent, extracting equipment operation state industrial data from the industrial data, inputting the equipment operation state industrial data into a general fault prediction model to obtain a fault verification analysis result; Comparing the fault verification analysis result with the fault analysis result; When the comparison is consistent, determining that the fault analysis result is verified.

8. The method of claim 7, wherein, The method further comprises: When the comparison is inconsistent, retrieving a historical industrial data set corresponding to the semantic vector within a preset historical time length according to semantic annotation of historical industrial data; Inputting the historical industrial data set and the equipment operation state industrial data into a pre-trained fluctuation hidden danger analysis network model to obtain a fluctuation coefficient of the industrial equipment; When the fluctuation coefficient is greater than a preset coefficient threshold and the fault analysis result is risky, determining that the fault analysis result is verified.

9. The method of claim 1, wherein, The resource scheduling intelligent agent and the understandable semantic information are used to allocate resources to the industrial equipment, specifically comprising: The resource scheduling intelligent agent extracts resource state information from the understandable semantic information; Inputting the resource state information into a resource scheduling model to obtain a resource allocation scheme of the industrial equipment.

10. An industrial agent collaborative decision making device, characterized in that, Comprise: At least one processor; And The memory is in communication connection with the at least one processor; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the industrial intelligent agent cooperative decision method of any one of claims 1-9.

Images