An AI agent assisted development method based on industrial big data

Abstract

The application discloses an AI intelligent agent assisted development method based on industrial big data, and relates to the technical field of big data processing, and comprises the following steps: S1, collecting historical mechanism model development records and preprocessing, and constructing an AI intelligent agent; S2, obtaining target equipment information and development requirements, executing task planning and mechanism reasoning, and constructing a mechanism characteristic sequence; S3, performing model planning on the mechanism characteristic sequence, matching a candidate model structure, and performing adaptive evaluation and iterative screening to generate a model construction scheme; S4, establishing an equipment mechanism model according to the model construction scheme, executing simulation deduction and deviation analysis, and obtaining a model deviation sequence; and S5, iteratively optimizing the equipment mechanism model based on the model deviation sequence until a preset convergence condition is met, and obtaining a target mechanism model. The application has the advantages of high model development efficiency, strong model structure adaptability and high model prediction accuracy by using an AI intelligent agent.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, and in particular to an AI-assisted development method based on industrial big data. Background Technology

[0002] With the development of industrial digitalization and intelligence, industrial equipment generates a large amount of equipment information, operating status information, model structure information, parameter configuration information, and model application results during operation and development. To analyze and predict the operating status of industrial equipment, it is usually necessary to establish corresponding mechanistic models based on the equipment's operating mechanism and simulate the state changes of the equipment under different operating conditions through these models.

[0003] The development of existing industrial equipment mechanism models mainly relies on developers determining the model structure based on the equipment's structure, operating rules, and development requirements, and then configuring parameters and repeatedly debugging based on historical experience. As equipment structures and operational relationships become increasingly complex, this manual development approach is heavily reliant on professional experience, resulting in low efficiency in model structure selection and parameter configuration. Furthermore, it is difficult to fully utilize the equipment information, development process, and model application results recorded in historical mechanism model development records.

[0004] Currently, some methods are beginning to utilize artificial intelligence to assist model development, but these typically focus on information retrieval, model recommendation, or parameter adjustment, lacking in-depth analysis of the relationships between the target equipment's operating patterns, model structure, and parameter configuration. This makes it difficult to automatically complete model planning and adaptation selection based on development needs. Furthermore, the simulation, deviation analysis, and model optimization processes after model establishment are usually independent of each other, making it difficult to continuously correct the model structure and parameter configuration based on the deviation between actual operating conditions and prediction results. This results in model development still requiring significant human intervention.

[0005] Therefore, how to provide an AI-assisted development method based on industrial big data is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an AI-assisted development method based on industrial big data. This invention fully utilizes historical mechanism model development records, AI agents, the ReAct inference framework, and Bayesian optimization algorithms to perform task planning and mechanism inference for the development requirements of target equipment. Based on the mechanism characteristics of the equipment, it completes adaptive screening of model structure and parameter configuration, and iteratively optimizes the equipment mechanism model by combining the deviation between simulation results and actual operating conditions. This achieves intelligent assisted development of industrial equipment mechanism models, and has the advantages of high model development efficiency, strong model structure adaptability, and high model prediction accuracy.

[0007] According to an embodiment of the present invention, an AI-assisted development method based on industrial big data includes the following steps:

[0008] S1. Collect and preprocess historical mechanism model development records, and construct AI agents based on the preprocessing results;

[0009] S2. Obtain target equipment information and development requirements, and use an AI agent to perform task planning and mechanism reasoning processing on the target equipment information and development requirements based on the ReAct reasoning framework to construct a mechanism feature sequence.

[0010] S3. The AI ​​agent performs model planning processing on the mechanism feature sequence, matches the corresponding candidate model structure from the standard structure library, and uses the Bayesian optimization algorithm to perform adaptation evaluation and iterative screening of each candidate model structure to determine the target model structure and target parameter configuration, and generate a model construction scheme.

[0011] S4. Establish a device mechanism model based on the model construction scheme, determine the prediction target according to the development requirements, perform simulation and deduction processing on the prediction target through the device mechanism model, generate a prediction state sequence, and perform deviation analysis on the prediction state sequence and the actual operating state of the target device at the corresponding time step to obtain the model deviation sequence.

[0012] S5. Optimize the equipment mechanism model based on the model deviation sequence, and iteratively execute simulation deduction, deviation analysis and model optimization until the preset convergence condition is met, so as to obtain the target mechanism model that meets the development requirements.

[0013] Optionally, the historical mechanism model development record represents a data set used to record equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information during the development process of historical equipment mechanism models. The preprocessing includes data cleaning, format unification, data encoding, and association organization operations. The AI ​​agent represents an intelligent development execution object with the capabilities of task planning, knowledge retrieval, mechanism reasoning, model planning, and model optimization. The target equipment information represents a set of information used to describe the operating status, structural characteristics, and associated parameters of the industrial equipment corresponding to the mechanism model to be developed. The development requirements represent a set of information used to describe the development goals, functional requirements, and model application requirements of the mechanism model corresponding to the target equipment.

[0014] Optionally, the construction process of the AI ​​agent specifically includes:

[0015] Feature extraction processing is performed on the preprocessed historical mechanism model development records. Corresponding features are extracted based on equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information.

[0016] Based on the correlation characteristics, the historical mechanism model development records are correlated and organized to establish the correspondence between development requirements and development process information, and to establish the correlation between equipment information and model structure information;

[0017] Based on the correspondence between development requirements and development process information, the task planning capability of the AI ​​agent is constructed. The development process information is processed by task decomposition according to development requirements, and the execution relationship between development tasks and model building process is established.

[0018] Based on the correlation between device information and model structure information, the ReAct inference framework is integrated to build the mechanism inference capability of AI agents;

[0019] Based on the model application results, an adaptation analysis is performed on the model structure information and parameter configuration information to establish a model structure evaluation relationship.

[0020] Based on the relationship between task execution and model structure evaluation, the historical mechanism model development records are linked and organized to build the knowledge retrieval capability of the AI ​​agent.

[0021] By linking and configuring task planning capabilities, mechanism reasoning capabilities, and knowledge retrieval capabilities, the construction of an AI agent can be completed.

[0022] Optionally, S2 specifically includes:

[0023] S21. Perform feature extraction processing on the target equipment information and development requirements, and construct equipment feature representation and requirement feature representation;

[0024] S22. Employing an AI agent, task planning and processing are performed based on device feature representations and demand feature representations to construct model development tasks, specifically including:

[0025] By using task planning capabilities to analyze the characteristic representation of requirements, the corresponding development goals and functional requirements are determined, and task planning information is generated.

[0026] By combining equipment feature representation with task planning information, the knowledge retrieval capabilities of the AI ​​agent are invoked to correlate and match development process information in the historical mechanism model development records, and to determine the task execution relationship corresponding to the development process information.

[0027] The development process information is broken down into multiple development tasks according to the task execution relationship, and the development tasks are then linked and organized to generate model development tasks.

[0028] S23. Based on the model development task, perform mechanism analysis and processing on the target equipment information and development requirements to determine the relationship between equipment operation rules and model influence;

[0029] S24. Based on the ReAct inference framework, perform mechanism inference operations according to the model development task, equipment operation rules and model influence relationship, and extract equipment mechanism features;

[0030] S25. Associate and organize the equipment mechanism characteristics with the model development tasks, and construct a mechanism characteristic sequence.

[0031] Optionally, S23 specifically includes:

[0032] S231. Based on the model development task, perform correlation analysis and processing on the target equipment information and development requirements, extract the corresponding operating status characteristics, equipment structure characteristics and functional requirement characteristics of the target equipment, and construct the equipment feature sequence.

[0033] S232. Use the variational mode decomposition algorithm to perform multi-scale feature decomposition on the equipment feature sequence, extract the state change features of the target equipment at different scales, and generate the operating feature sequence.

[0034] S233. Use the attention mechanism to perform feature association analysis on the running feature sequence, calculate the association weight between different running features, and select key running features based on the association weight;

[0035] S234. Calculate the Pearson correlation coefficient between key operating characteristics, and determine the equipment operating pattern based on the Pearson correlation coefficient;

[0036] S235. Based on the equipment operation rules and development requirements, perform correlation analysis on the model structure information and parameter configuration information in the historical mechanism model development record to determine the multiple model structures and parameter configurations corresponding to the equipment operation rules and establish model influence relationships.

[0037] Optionally, S24 specifically includes:

[0038] S241. Based on the model development task, equipment operation rules and model influence relationship, perform task parsing operation based on the ReAct inference framework to determine the inference target and inference content corresponding to the model development task.

[0039] S242. Call upon the mechanistic reasoning ability of the AI ​​agent, and perform correlation analysis on the relationship between equipment operation rules and model influence based on the reasoning goal and reasoning content, and determine the correlation characteristics between equipment operation rules and model influence.

[0040] S243. Use Bayesian inference to perform probabilistic inference processing on the association analysis results and calculate the association probability value of each association feature;

[0041] S244. Calculate the confidence level corresponding to each association feature based on the association probability value, and filter out valid association features with a confidence level greater than a preset confidence threshold. Specifically, this includes:

[0042] Based on the equipment operation rules, model structure information and parameter configuration information corresponding to each associated feature, the historical mechanism model development records are matched and retrieved to determine the number of times each associated feature appears in the historical mechanism model development records;

[0043] Based on the ratio between the frequency of occurrence and the total number of records, the historical correlation degree corresponding to each correlation feature is calculated, where the total number of records represents the total number of records developed by the historical mechanism model.

[0044] Analyze the distribution of each correlation feature in the historical mechanism model development record to determine the corresponding correlation stability.

[0045] The confidence level corresponding to each association feature is calculated by combining the association probability value, historical association degree, and association stability degree.

[0046] The confidence scores of each associated feature are compared with a preset confidence threshold, and valid associated features with confidence scores greater than the preset confidence threshold are selected.

[0047] S245. Normalize the correlation probability values ​​of each effective correlation feature, and perform weighted fusion processing on each effective correlation feature based on the normalization result to extract the equipment mechanism features in the equipment operation rules.

[0048] Optionally, S3 specifically includes:

[0049] S31. The AI ​​agent performs model planning processing on the mechanism feature sequence, and matches candidate model structures from the standard structure library based on the device mechanism features in the mechanism feature sequence.

[0050] S32. Use the Bayesian optimization algorithm to perform adaptation evaluation on the structure of each candidate model.

[0051] S33. Based on the adaptation evaluation results, perform screening processing on the candidate model structure, and iteratively perform adaptation evaluation and screening processing on the screening results until the preset iteration termination condition is reached, and determine the target model structure.

[0052] S34. Perform parameter configuration optimization processing based on the target model structure, and determine the target parameter configuration by combining the parameter configuration information corresponding to the target model structure in the standard structure library with the equipment mechanism characteristics.

[0053] S35. Generate a model construction scheme based on the target model structure and target parameter configuration.

[0054] Optionally, S32 specifically includes:

[0055] S321. Obtain the model structure information and parameter configuration information corresponding to each candidate model structure in the standard structure library, and match them with the equipment mechanism features in the mechanism feature sequence one by one. Calculate the structure adaptation value and parameter adaptation value of each candidate model structure based on the matching results.

[0056] S322. Call the knowledge retrieval capability of the AI ​​agent to match the model structure information corresponding to each candidate model structure with the model structure information in the historical mechanism model development record, and calculate the corresponding structure matching degree.

[0057] S323. Sort the historical mechanism model development records according to the degree of structural matching, and extract the historical mechanism model development records with the highest degree of structural matching with each candidate model.

[0058] S324. Match the parameter configuration information corresponding to each candidate model structure with the parameter configuration information extracted from the historical mechanism model development record, and calculate the corresponding parameter matching degree.

[0059] S325. Determine the historical evaluation value based on the model application result information extracted from the historical mechanism model development record, and perform weighted processing on the historical evaluation value in combination with the degree of structure matching and the degree of parameter matching to obtain the prior evaluation result corresponding to each candidate model structure.

[0060] S326. The structure fit value, parameter fit value and prior evaluation results of each candidate model structure are used as the evaluation data of the Bayesian optimization algorithm. The evaluation data are probabilistically fitted using a Gaussian process to obtain the prediction mean and prediction variance corresponding to each candidate model structure.

[0061] S327. Using the expected improvement criterion in the Bayesian optimization algorithm, the predicted mean and predicted variance corresponding to each candidate model structure are jointly calculated to obtain the fitness evaluation value corresponding to each candidate model structure as the fitness evaluation result.

[0062] Optionally, S4 specifically includes:

[0063] S41. Establish the equipment mechanism model according to the target model structure in the model construction scheme, and configure the target parameters to the corresponding positions in the equipment mechanism model;

[0064] S42. Determine the prediction target corresponding to the equipment mechanism model based on the development requirements, and extract the corresponding operating status from the target equipment information according to the prediction target;

[0065] S43. Input the operating status into the device mechanism model, perform simulation and deduction processing on the prediction target through the device mechanism model, obtain the prediction status corresponding to each time step, and arrange the prediction status in time order to generate a prediction status sequence.

[0066] S44. Based on the time steps of the predicted state sequence, collect the actual operating state of the target device at each corresponding time step, and perform deviation analysis processing on the predicted state and the actual operating state.

[0067] S45. Arrange the deviation analysis results in chronological order to obtain the model deviation sequence.

[0068] Optionally, S5 specifically includes:

[0069] S51. Based on the model deviation sequence, adjust the target model structure and target parameter configuration of the equipment mechanism model to optimize the equipment mechanism model;

[0070] S52. Iteratively perform simulation deduction, deviation analysis and model optimization processing through the updated equipment mechanism model until the preset convergence condition is met, and then stop the iteration.

[0071] The preset convergence condition means that the deviation analysis results in the model deviation sequence are all less than the preset deviation threshold, and the change amplitude of the corresponding deviation analysis results in the model deviation sequence obtained in two consecutive iterations is less than the preset change threshold.

[0072] S53. Use the device mechanism model after the iteration stops as the target mechanism model to meet the development requirements.

[0073] The beneficial effects of this invention are:

[0074] First, this invention develops and constructs an AI agent based on historical mechanism model development records, and combines the ReAct inference framework to perform task planning and mechanism inference on target equipment information and development requirements, thereby improving the accuracy of equipment mechanism feature extraction and reducing the dependence of mechanism model development on human experience.

[0075] Secondly, candidate model structures are matched with standard structure libraries, and Bayesian optimization algorithms are used for adaptation evaluation and iterative screening to determine the target model structure and target parameter configuration, thereby improving the compatibility between the model construction scheme and the target equipment operation mechanism.

[0076] Finally, a closed-loop iterative process is formed through simulation, deviation analysis, and model optimization. The equipment mechanism model is continuously corrected based on the actual operating status, thereby improving the development efficiency, prediction accuracy, and application reliability of the target mechanism model. Attached Figure Description

[0077] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0078] Figure 1This is a flowchart of an AI-assisted development method based on industrial big data proposed in this invention;

[0079] Figure 2 This is a flowchart illustrating the mechanism feature analysis and model construction scheme generation of an AI intelligent agent-assisted development method based on industrial big data proposed in this invention.

[0080] Figure 3 This is a flowchart illustrating the construction and iterative optimization of the equipment mechanism model for an AI-assisted development method based on industrial big data, as proposed in this invention. Detailed Implementation

[0081] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0082] refer to Figures 1-3 A method for AI-assisted development of intelligent agents based on industrial big data includes the following steps:

[0083] S1. Collect and preprocess historical mechanism model development records, and construct AI agents based on the preprocessing results;

[0084] S2. Obtain target equipment information and development requirements, and use an AI agent to perform task planning and mechanism reasoning processing on the target equipment information and development requirements based on the ReAct reasoning framework to construct a mechanism feature sequence.

[0085] S3. The AI ​​agent performs model planning processing on the mechanism feature sequence, matches the corresponding candidate model structure from the standard structure library, and uses the Bayesian optimization algorithm to perform adaptation evaluation and iterative screening of each candidate model structure to determine the target model structure and target parameter configuration, and generate a model construction scheme.

[0086] S4. Establish a device mechanism model based on the model construction scheme, determine the prediction target according to the development requirements, perform simulation and deduction processing on the prediction target through the device mechanism model, generate a prediction state sequence, and perform deviation analysis on the prediction state sequence and the actual operating state of the target device at the corresponding time step to obtain the model deviation sequence.

[0087] S5. Optimize the equipment mechanism model based on the model deviation sequence, and iteratively execute simulation deduction, deviation analysis and model optimization until the preset convergence condition is met, so as to obtain the target mechanism model that meets the development requirements.

[0088] In this embodiment, the historical mechanism model development record represents a data set used to record equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information during the development process of historical equipment mechanism models. Preprocessing includes data cleaning, format unification, data encoding, and association organization operations. The AI ​​agent represents an intelligent development execution object with the capabilities of task planning, knowledge retrieval, mechanism reasoning, model planning, and model optimization. The target equipment information represents a set of information used to describe the operating status, structural characteristics, and associated parameters of the industrial equipment corresponding to the mechanism model to be developed. The development requirements represent a set of information used to describe the development goals, functional requirements, and model application requirements of the mechanism model corresponding to the target equipment.

[0089] In this embodiment, the process of constructing the AI ​​agent specifically includes:

[0090] Feature extraction is performed on the preprocessed historical mechanism model development records. Corresponding features are extracted based on equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information. The associated features are then uniformly organized according to the data correspondence in the historical mechanism model development records to ensure that the equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information in the same historical mechanism model development record remain consistent.

[0091] Based on the correlation characteristics, the historical mechanism model development records are correlated and organized, the development requirements are matched with the corresponding development process information, the corresponding development process is determined according to the development goals and functional requirements recorded in the development requirements, and the correspondence between the development requirements and the development process information is established. At the same time, based on the equipment structure characteristics and equipment correlation parameters in the equipment information, the equipment information is correlated with the model structure information that adopts the corresponding structural form, and the correlation between the equipment information and the model structure information is established.

[0092] Based on the correspondence between development requirements and development process information, the task planning capability of the AI ​​agent is constructed. According to the development requirements, the corresponding development process information is processed by task decomposition. Development tasks are formed according to the execution order of model structure determination, parameter configuration, model construction, simulation and deviation analysis. Based on the dependencies between each development task in the model construction process, the execution relationship between development tasks and model construction process is established, so that the processing results of the previous development task participate in the execution of the next development task.

[0093] Based on the relationship between equipment information and model structure information, the ReAct inference framework is integrated into the AI ​​agent. The device structure features, equipment association parameters and model structure information corresponding to the equipment information are analyzed. During the mechanism inference process, the corresponding historical mechanism model development records are retrieved based on the current analysis results. The retrieved model structure information, parameter configuration information and model application result information are used to update the current inference content. Mechanism inference is continuously executed according to the processing relationship of analysis, retrieval and update, thus building the mechanism inference capability of the AI ​​agent.

[0094] Based on the model application results information, an adaptation analysis is performed on the model structure information and parameter configuration information. The model application results obtained by using the corresponding model structure information and parameter configuration information are correlated. The adaptation of the corresponding model structure information and parameter configuration information is determined based on the deviation analysis results between the predicted state and the actual operating state in the model application results. The model structure evaluation relationship between the model structure information, parameter configuration information and model application results information is established.

[0095] Based on the task execution relationship and the model structure evaluation relationship, the historical mechanism model development records are linked and organized. The development tasks are linked with the corresponding development process information, and the equipment information is linked with the corresponding model structure information, parameter configuration information and model application result information. The retrieval order of the historical mechanism model development records is determined according to the task execution relationship. Then, the model structure information and parameter configuration information that match the current development task are determined according to the model structure evaluation relationship, so as to build the knowledge retrieval capability of the AI ​​agent.

[0096] By configuring task planning capabilities, mechanism reasoning capabilities, and knowledge retrieval capabilities together, the development tasks generated by task planning capabilities serve as the retrieval basis for knowledge retrieval capabilities. The historical mechanism model development records obtained by knowledge retrieval capabilities serve as the analysis basis for mechanism reasoning capabilities. The reasoning results generated by mechanism reasoning capabilities are returned to task planning capabilities to participate in subsequent development task processing, thereby completing the construction of the AI ​​agent.

[0097] In this embodiment, S2 specifically includes:

[0098] S21. Perform feature extraction processing on the target equipment information and development requirements. Extract the corresponding equipment operating status, equipment structural features and equipment related parameters from the target equipment information, and extract the corresponding development goals, functional requirements and model application requirements from the development requirements. Associate and organize the extraction results according to the correspondence between the target equipment information and the development requirements, and construct equipment feature representation and requirement feature representation.

[0099] S22. Employing an AI agent, task planning and processing are performed based on device feature representations and demand feature representations to construct model development tasks, specifically including:

[0100] By using task planning capabilities to analyze the characteristic representation of requirements, the development goals are used as the task objectives of the model development task, and the functional requirements and model application requirements are used as the basis for task execution. The development goals, functional requirements and model application requirements are combined to form task planning information.

[0101] By combining equipment feature representation and task planning information, the knowledge retrieval capabilities of the AI ​​agent are invoked to match the equipment operating status, equipment structural features, and equipment associated parameters with the equipment information in the historical mechanism model development record. Based on the task planning information, the development goals and functional requirements are matched with the development requirements and development process information in the historical mechanism model development record. The development process information that matches the current target equipment and development requirements is extracted, and the task execution relationship is determined based on the model construction process corresponding to the development process information.

[0102] The development process information is broken down into tasks according to the task execution relationship. Development tasks with sequential execution dependencies are associated sequentially according to the model building process. Development tasks with common processing objects are organized accordingly, so that the processing result of the previous development task serves as the basis for the processing of the next development task, thus generating model development tasks.

[0103] S23. Based on the model development task, perform mechanism analysis and processing on the target equipment information and development requirements to determine the relationship between equipment operation rules and model influence;

[0104] S24. Based on the ReAct inference framework, perform mechanism inference operations according to the model development task, equipment operation rules and model influence relationship, and extract equipment mechanism features;

[0105] S25. Associate and organize the equipment mechanism characteristics with the model development tasks, and construct a mechanism characteristic sequence.

[0106] In this embodiment, S23 specifically includes:

[0107] S231. Based on the model development task, perform correlation analysis and processing on the target equipment information and development requirements. According to the development goal corresponding to the model development task, associate the operating status, equipment structural features and equipment correlation parameters in the target equipment information with the functional requirements in the development requirements. Extract operating status features that can reflect the changes in the target equipment status from the operating status, extract equipment structural features that can reflect the structural composition and correlation of the target equipment from the equipment structural features and equipment correlation parameters, and extract functional requirement features corresponding to the development goal from the functional requirements. Arrange them according to the correlation relationship between the operating status features, equipment structural features and functional requirement features to construct the equipment feature sequence.

[0108] S232. The variational mode decomposition algorithm is used to perform multi-scale feature decomposition on the equipment feature sequence. The features that change continuously with the operating state in the equipment feature sequence are decomposed, and the state change features with different rates of change are separated into corresponding scales. The correspondence between the state change features at each scale and the equipment structural features and functional requirements features is maintained. The state change features at each scale obtained by decomposition are arranged according to the arrangement relationship in the equipment feature sequence to generate the operating feature sequence.

[0109] S233. Use the attention mechanism to perform feature association analysis on the operating feature sequence, associate the state change features at different scales with the corresponding equipment structural features and functional requirements features respectively, calculate the association weight based on the consistency and correspondence of the state changes among the operating features, sort the operating features according to the association weight, and select the operating features with a high degree of association with the target equipment's operating state changes and functional requirements as key operating features.

[0110] S234. According to the correspondence between key operating features in the operating feature sequence, calculate the Pearson correlation coefficient for any two key operating features that are related. Determine the unidirectional and inverse relationship between key operating features based on the sign of the Pearson correlation coefficient. Determine the degree of correlation between key operating features based on the absolute value of the Pearson correlation coefficient. Organize key operating features with stable correlation according to equipment structure features and functional requirements features to determine the equipment operation rules corresponding to the target equipment.

[0111] S235. Based on the equipment operation rules and development requirements, invoke the knowledge retrieval capabilities of the AI ​​agent to associate and match the operation status characteristics, equipment structure characteristics, and functional requirement characteristics corresponding to the equipment operation rules with the equipment information and development requirements in the historical mechanism model development records. Based on the matching results, extract the corresponding model structure information and parameter configuration information, analyze the correspondence between different equipment operation rules and each model structure and parameter configuration, determine multiple model structures and parameter configurations corresponding to the equipment operation rules, and establish model influence relationships according to the correlation between equipment operation rules, model structure information, and parameter configuration information.

[0112] In this embodiment, S24 specifically includes:

[0113] S241. Based on the model development task, equipment operation rules and model influence relationship, perform task parsing operation based on the ReAct inference framework, take the development goal corresponding to the model development task as the current inference goal, and take the equipment operation rules, model structure information and parameter configuration information corresponding to the development goal as the inference content. According to the task execution relationship in the model development task, the inference content is associated and organized to determine the inference goal and inference content corresponding to the model development task.

[0114] S242. Call upon the mechanism reasoning ability of the AI ​​agent, perform correlation analysis processing on the relationship between equipment operation law and model influence based on the reasoning goal and reasoning content, match the key operation features corresponding to the equipment operation law with the model structure information and parameter configuration information in the model influence relationship, analyze the correspondence between the changes in key operation features and the model structure and parameter configuration, and extract the correlation features that can reflect the degree of influence of equipment operation law on model structure and parameter configuration based on the correspondence.

[0115] S243. The Bayesian inference method is used to perform probabilistic inference processing on the association analysis results. Based on the existing correspondence between the equipment operation rules, model structure information and parameter configuration information in the historical mechanism model development record, the probability of each association feature being true under the current inference objective is updated, and the probability update result is used as the association probability value corresponding to each association feature.

[0116] S244. Calculate the confidence level corresponding to each association feature based on the association probability value, and filter out valid association features with a confidence level greater than a preset confidence threshold. Specifically, this includes:

[0117] Based on the equipment operation rules, model structure information and parameter configuration information corresponding to each associated feature, the historical mechanism model development records are matched and retrieved. The historical mechanism model development records that simultaneously have the corresponding equipment operation rules, model structure information and parameter configuration information are identified as matching records, and the number of times the matching records corresponding to each associated feature appear is counted.

[0118] Based on the ratio between the frequency of occurrence and the total number of records, the historical correlation degree corresponding to each correlation feature is calculated. The total number of records represents the total number of records developed by the historical mechanism model. The higher the proportion of the frequency of occurrence to the total number of records, the higher the historical correlation degree of the corresponding correlation feature.

[0119] The distribution of each associated feature in the historical mechanism model development records was analyzed. The historical mechanism model development records were organized according to the development process information. The occurrence ratio of associated features in different historical mechanism model development records was compared. When the difference in the occurrence ratio of associated features in different historical mechanism model development records decreased, the corresponding association stability was increased. When the difference in the occurrence ratio increased, the corresponding association stability was decreased.

[0120] The association probability value, historical association degree, and association stability degree are converted to the same numerical range, and the confidence degree corresponding to each association feature is calculated based on the comprehensive result of the three, so that the association probability value, historical association degree, and association stability degree jointly participate in the determination of confidence degree.

[0121] The confidence level of each associated feature is compared with a preset confidence threshold. The associated features with a confidence level greater than the preset confidence threshold are retained as valid associated features, and the associated features that do not reach the preset confidence threshold are removed from the current mechanism inference results.

[0122] S245. Normalize the correlation probability values ​​of each effective correlation feature to ensure that the sum of the normalization results of each effective correlation feature is consistent. Determine the relative degree of each effective correlation feature's participation in the fusion based on the normalization results, and perform weighted fusion processing on each effective correlation feature according to the relative degree of participation. Associate the fusion results with the corresponding equipment operation rules and extract equipment mechanism features that can characterize the target equipment's operation mechanism and participate in subsequent model planning processing.

[0123] In this embodiment, S3 specifically includes:

[0124] S31. The AI ​​agent performs model planning processing on the mechanism feature sequence, reads the equipment mechanism features corresponding to each model development task in the mechanism feature sequence, and associates and matches the equipment operation rules, model influence relationships and corresponding development requirements reflected by the equipment mechanism features with the structural features and parameter configuration information corresponding to the model structure information in the standard structure library. Based on the matching results, the model structure that has a corresponding relationship with the equipment mechanism features is extracted as a candidate model structure, and the association relationship between each candidate model structure and the corresponding equipment mechanism feature is maintained.

[0125] S32. Use the Bayesian optimization algorithm to perform adaptation evaluation on the structure of each candidate model.

[0126] S33. Based on the adaptation evaluation results, perform screening processing on the candidate model structures, determine the candidate model structures to be retained according to the adaptation evaluation results corresponding to each candidate model structure, and continue to associate the screened candidate model structures with the corresponding device mechanism features. Iterate the adaptation evaluation and screening processing on the screening results, so that each adaptation evaluation is updated based on the candidate model structures retained in the previous round of screening, until the preset iteration termination condition is reached, stop the candidate model structure screening, and determine the model structure retained after stopping the screening as the target model structure.

[0127] S34. Perform parameter configuration optimization processing according to the target model structure, extract the parameter configuration information corresponding to the target model structure from the standard structure library, associate each parameter configuration information with the equipment mechanism characteristics corresponding to the target model structure, analyze the correspondence between different parameter configurations and equipment mechanism characteristics based on the equipment operation law and model influence relationship, and determine the configuration status of each parameter in the target model structure in combination with development requirements to form a target parameter configuration that is compatible with the target model structure.

[0128] S35. Associate the target model structure with the target parameter configuration, determine the structural composition and corresponding relationship of the equipment mechanism model based on the target model structure, determine the parameter settings corresponding to the target model structure based on the target parameter configuration, organize the target model structure and target parameter configuration according to the corresponding relationship, and generate a model construction scheme for subsequent establishment of the equipment mechanism model.

[0129] In this embodiment, S32 specifically includes:

[0130] S321. Obtain the model structure information and parameter configuration information corresponding to each candidate model structure in the standard structure library. Match the structural composition and structural relationship reflected by the model structure information with the equipment mechanism features in the mechanism feature sequence item by item. Match the parameter settings and parameter relationship reflected by the parameter configuration information with the corresponding equipment mechanism features item by item. Normalize and summarize the matching results between the model structure information and the equipment mechanism features. Calculate the structural adaptation value corresponding to each candidate model structure. Normalize and summarize the matching results between the parameter configuration information and the equipment mechanism features. Calculate the parameter adaptation value corresponding to each candidate model structure.

[0131] S322. Utilize the knowledge retrieval capabilities of the AI ​​agent, using the model structure information corresponding to each candidate model structure as the retrieval basis. Extract historical mechanism model development records that adopted the corresponding model structure for model development from the historical mechanism model development records. Match the structural composition and structural relationships of the candidate model structure with the model structure information in each historical mechanism model development record item by item. Calculate the corresponding structural matching degree based on the consistency of structural composition and structural relationships, so that the structural matching degree reflects the overall similarity between the candidate model structure and the historical model structure.

[0132] S323. Sort the historical mechanism model development records corresponding to each candidate model structure in descending order of structural matching degree, and maintain the original correspondence between the equipment information, development requirements, model structure information, parameter configuration information and model application result information in the historical mechanism model development records. Extract the historical mechanism model development record with the highest structural matching degree corresponding to the current candidate model structure from the sorting results, and use it as the basis for subsequent parameter matching and historical evaluation.

[0133] S324. Match the parameter configuration information corresponding to each candidate model structure with the parameter configuration information extracted from the historical mechanism model development record. Compare the parameter settings and parameter relationships according to the corresponding position of the parameters in the model structure. Normalize and summarize the consistency results of parameter settings and parameter relationships. Calculate the parameter matching degree corresponding to the candidate model structure so that the parameter matching degree reflects the adaptation relationship between the current parameter configuration and the historical parameter configuration.

[0134] S325. Read the model application result information from the extracted historical mechanism model development records. Determine the historical evaluation value based on the degree of conformity between the model operation results reflected in the model application result information and the corresponding development requirements. Then, determine the reference value of the historical evaluation value for the current candidate model structure based on the degree of structural matching and parameter matching. Perform weighted processing on the historical evaluation value so that the historical mechanism model development records with a high degree of matching with the current candidate model structure in terms of model structure and parameter configuration have a greater evaluation role, and obtain the prior evaluation results corresponding to each candidate model structure.

[0135] S326. Correlate the structure fit value and parameter fit value corresponding to each candidate model structure with the prior evaluation results as evaluation data for the Bayesian optimization algorithm. Use Gaussian process to establish the probabilistic correspondence between the evaluation data and the candidate model structure. Perform probability fitting on the candidate model structure that has not yet completed the subsequent evaluation. Estimate the central change level and uncertainty of the fit evaluation results corresponding to each candidate model structure based on the existing evaluation data. Obtain the prediction mean and prediction variance corresponding to each candidate model structure respectively.

[0136] S327. Using the expected improvement criterion in the Bayesian optimization algorithm, and taking the currently obtained fit evaluation results as the comparison basis, the predicted mean and predicted variance corresponding to each candidate model structure are jointly calculated. The predicted mean reflects the expected degree of the candidate model structure to achieve a better fit evaluation result, and the predicted variance reflects the uncertainty of the candidate model structure evaluation result. Based on the joint calculation results of the two, the expected improvement degree of each candidate model structure relative to the current fit evaluation result is determined, and the expected improvement degree is converted into the corresponding fit evaluation value as the fit evaluation result of each candidate model structure.

[0137] In this embodiment, S4 specifically includes:

[0138] S41. Establish the equipment mechanism model according to the target model structure in the model construction scheme, complete the structural configuration of the equipment mechanism model according to the structural composition and structural connection relationship corresponding to the target model structure, and read the target parameter configuration in the model construction scheme. According to the correspondence of each target parameter in the target model structure, configure the target parameter configuration to the corresponding position of the equipment mechanism model so that the equipment mechanism model maintains the correspondence between the target model structure and the target parameter configuration.

[0139] S42. Based on the development goals, functional requirements and model application requirements in the development needs, determine the prediction targets corresponding to the equipment mechanism model. According to the correspondence between the prediction targets and the target equipment information, extract the operating states participating in the simulation from the target equipment information, and organize the extraction results according to the correlation between the operating states and the prediction targets, so that the operating states are associated with the target parameters corresponding to the equipment mechanism model.

[0140] S43. Input the operating status into the equipment mechanism model, and perform continuous simulation and deduction processing on the operating status according to the structural relationship established by the target model structure and the parameter association relationship corresponding to the target parameter configuration. This allows the operating status of the current time step to be passed to the next time step according to the equipment operating law reflected by the equipment mechanism model, and the predicted status corresponding to each time step is obtained in sequence. The correspondence between each predicted status and the predicted target is maintained, and the predicted status is arranged in chronological order to generate a predicted status sequence.

[0141] S44. Based on the time steps corresponding to the predicted state sequence, collect the actual operating state of the target equipment at each corresponding time step, match each predicted state with the actual operating state corresponding to the same time step, calculate the state difference between the predicted state and the actual operating state, and analyze the prediction deviation of the equipment mechanism model at each time step based on the state difference to form the deviation analysis results for each time step.

[0142] S45. Maintain the correlation between each deviation analysis result and the predicted state, actual operating state and corresponding time step, arrange each deviation analysis result according to the time order of the predicted state sequence to obtain the model deviation sequence, and use the model deviation sequence as the basis for subsequent equipment mechanism model optimization processing.

[0143] In this embodiment, S5 specifically includes:

[0144] S51. Based on the model deviation sequence, perform optimization processing on the equipment mechanism model, associate the deviation analysis results corresponding to each time step with the target model structure and target parameter configuration in the equipment mechanism model, and determine the model structure information and parameter configuration information corresponding to the deviation analysis results based on the deviation direction and degree of deviation between the predicted state and the actual operating state reflected by the model deviation sequence, combined with the model influence relationship. Adjust the structural relationships in the target model structure that are related to the deviation analysis results, and update the corresponding parameter settings in the target parameter configuration. Reconfigure the adjusted target model structure and the updated target parameter configuration into the equipment mechanism model to complete the optimization of the equipment mechanism model in the current iteration process.

[0145] S52. Using the updated equipment mechanism model, the operating status of the target equipment information is re-extracted according to the prediction target corresponding to the development requirements. Simulation and deduction processing is performed for the prediction target to generate an updated prediction state sequence. The updated prediction state sequence and the actual operating status of the corresponding time step are compared again to obtain the model deviation sequence corresponding to the current iteration. The target model structure and target parameter configuration are then adjusted based on the current model deviation sequence. The simulation, deduction, and model optimization processes are executed in a loop. After each model optimization, the model deviation sequence obtained in the current iteration is compared with the preset deviation threshold and the model deviation sequence obtained in the previous iteration. When all the deviation analysis results in the model deviation sequence are less than the preset deviation threshold, and the change amplitude of the corresponding deviation analysis results in the model deviation sequences obtained in two consecutive iterations is less than the preset change threshold, the preset convergence condition is met and the iteration is stopped. If the preset convergence condition is not met, the next simulation, deduction, deviation analysis, and model optimization process is continued.

[0146] S53. Read the target model structure and target parameter configuration corresponding to the time when the iteration stops, maintain the correspondence between the target model structure, target parameter configuration and device mechanism model, and use the finally optimized device mechanism model as the target mechanism model that meets the development requirements.

[0147] Example 1: To verify the feasibility of this invention in practice, it was applied to the development of a mechanism model for centrifugal compression equipment in an industrial production scenario. This scenario involved 12 centrifugal compression devices participating in the operational analysis. The rated power of these devices ranged from 760kW to 1180kW, the rated speed from 7350r / min to 9860r / min, the normal operating inlet pressure from 0.31MPa to 0.47MPa, and the outlet pressure from 0.86MPa to 1.34MPa. Each device was configured with 18 operational status acquisition channels, with an acquisition cycle of 30s. The main data recorded included inlet pressure, outlet pressure, inlet temperature, outlet temperature, speed, flow rate, bearing temperature, vibration amplitude, and drive power. A total of 186 historical mechanism model development records were generated for constructing the AI ​​agent, corresponding to 34 similar or related industrial devices. These records contained equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application results. The standard structure library contains 17 callable mechanism model structures, each with corresponding parameter ranges and historical adaptation records.

[0148] This scenario requires the rapid development of mechanistic models for predicting outlet pressure and drive power for centrifugal compressors with different operating characteristics. Traditional development methods typically involve developers manually selecting the model structure based on equipment structure and historical experience, then repeatedly adjusting model parameters based on actual operating data. When the equipment's inlet state, load range, and speed range change, the applicability of the same model structure varies significantly across different devices. Developers need to modify the model structure and parameters multiple times, resulting in lengthy development times and a high risk of significant discrepancies between simulation results and actual operating conditions.

[0149] When applying this invention, the 186 historical mechanistic model development records are first cleaned, formatted, encoded, and linked. Relationships are established based on the development requirements, model structure, parameter configuration, development process, and model application results corresponding to different historical devices. An AI agent is then constructed based on this relationship. The AI ​​agent develops task planning, mechanistic reasoning, and knowledge retrieval capabilities based on historical records, enabling the historical model development experience to be reused in the development of new device mechanistic models, rather than simply being stored as independent historical files.

[0150] For the target equipment in the mechanism model to be developed, the equipment's operating status, structural features, and related parameters are used as target equipment information. Simultaneously, development requirements corresponding to outlet pressure prediction and drive power prediction are input. The AI ​​agent analyzes the development requirements based on the ReAct inference framework and determines the model development tasks by combining historical development records. For continuous equipment operation data, the variational mode decomposition algorithm is used to perform multi-scale decomposition of operating status changes. Then, an attention mechanism is used to analyze the correlation between different operating features, and the Pearson correlation coefficient is used to identify key operating features with strong relationships to changes in outlet pressure and drive power. In a verification device of this embodiment, 11 key operating features were selected from the features formed by 18 operating status acquisition channels. Among them, inlet pressure, inlet temperature, rotational speed, flow rate, and drive power showed relatively stable correlations with the prediction targets.

[0151] The AI ​​agent further combines equipment operating patterns and historical model structure information to perform mechanistic inference. Through Bayesian inference, it obtains the correlation probabilities between different operating patterns and model structures and parameter configurations, and filters effective correlation features based on historical correlation strength and correlation stability, ultimately constructing a mechanistic feature sequence. Subsequently, candidate model structures are selected from 17 standard model structures based on the mechanistic feature sequence. In actual processing, different devices typically yield 4 to 7 candidate model structures. A Bayesian optimization algorithm is then used to iteratively evaluate these structures, parameters, and historical model application results. A Gaussian process obtains the prediction mean and prediction variance based on the candidate model evaluation data, and then continuously filters new evaluation objects using the expected improvement criterion, ultimately determining the target model structure and target parameter configuration.

[0152] A device mechanism model is established based on the determined model construction scheme. The actual operating state of the target device is input into the device mechanism model for simulation and deduction, and the predicted outlet pressure and drive power states for each time step are obtained. A deviation sequence is formed by analyzing the deviation between the predicted states and the actual operating states collected at the same time step. When the deviation for some time steps still exceeds the preset deviation threshold, the target model structure and target parameter configuration are readjusted according to the model deviation sequence, and simulation and deviation analysis are repeated until the model deviation meets the preset convergence condition. To further verify the actual effect, eight devices with significant differences in operating conditions were selected from 12 devices for comparative verification. The model development results obtained by this invention are compared with those obtained by traditional manual modeling methods. Specific data are shown in the table below.

[0153] Table 1 Comparison of Development and Prediction Results of Mechanism Models for Different Centrifugal Compression Devices

[0154]

[0155] As shown in Table 1, the present invention maintains a relatively stable mechanistic model development effect under different rated power and operating speed conditions. Taking device A as an example, the traditional manual method requires 18.6 hours to complete the model structure selection, parameter configuration, and verification, while the present invention only requires 7.4 hours to complete the same development task. The average absolute error of the outlet pressure is reduced from 0.047 MPa to 0.021 MPa, and the average absolute error of the drive power is reduced from 34.8 kW to 15.7 kW. For device F with more complex operating conditions, the traditional manual method requires 25.1 hours, and the average absolute error of the outlet pressure reaches 0.067 MPa. However, after 9 model optimization iterations, the development time of the present invention is reduced to 10.4 hours, the average absolute error of the outlet pressure is reduced to 0.029 MPa, and the average absolute error of the drive power is reduced from 51.8 kW to 22.6 kW.

[0156] Further statistical analysis of data from eight verification devices revealed that the average model development time for this invention was approximately 8.49 hours, while the average model development time for traditional manual methods was approximately 21.25 hours. The average absolute error of the outlet pressure in the device mechanism model established by this invention was approximately 0.024 MPa, compared to approximately 0.055 MPa for traditional manual methods. The average absolute error of the drive power was reduced from 41.43 kW for traditional methods to 18.18 kW for this invention. Simultaneously, this invention automatically completes candidate model structure screening, target parameter configuration, and model deviation feedback optimization through an AI agent, reducing the process of repeatedly comparing model structures and adjusting parameters manually.

[0157] The above results demonstrate that the present invention can fully utilize the development experience formed by historical mechanism model development records, automatically complete mechanism feature extraction, model structure selection and parameter configuration based on the operating rules of the target equipment, and continuously correct the equipment mechanism model using the actual operating state. This reduces the time spent on mechanism model development, improves the model's adaptability to the operating characteristics of the target equipment and the accuracy of the prediction results, and verifies the feasibility and effectiveness of the present invention in the scenario of assisted development of industrial equipment mechanism models.

[0158] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for AI-assisted development of intelligent agents based on industrial big data, characterized in that, Includes the following steps: S1. Collect and preprocess historical mechanism model development records, and construct AI agents based on the preprocessing results; S2. Obtain target equipment information and development requirements, and use an AI agent to perform task planning and mechanism reasoning processing on the target equipment information and development requirements based on the ReAct reasoning framework to construct a mechanism feature sequence. S3. The AI ​​agent performs model planning processing on the mechanism feature sequence, matches the corresponding candidate model structure from the standard structure library, and uses the Bayesian optimization algorithm to perform adaptation evaluation and iterative screening of each candidate model structure to determine the target model structure and target parameter configuration, and generate a model construction scheme. S4. Establish a device mechanism model based on the model construction scheme, determine the prediction target according to the development requirements, perform simulation and deduction processing on the prediction target through the device mechanism model, generate a prediction state sequence, and perform deviation analysis on the prediction state sequence and the actual operating state of the target device at the corresponding time step to obtain the model deviation sequence. S5. Optimize the equipment mechanism model based on the model deviation sequence, and iteratively execute simulation deduction, deviation analysis and model optimization until the preset convergence condition is met, so as to obtain the target mechanism model that meets the development requirements.

2. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, The historical mechanism model development record represents a data set used to record equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information during the development process of historical equipment mechanism models. The preprocessing includes data cleaning, format unification, data encoding, and association organization operations. The AI ​​agent represents an intelligent development execution object with the capabilities of task planning, knowledge retrieval, mechanism reasoning, model planning, and model optimization. The target equipment information represents a set of information used to describe the operating status, structural characteristics, and associated parameters of the industrial equipment corresponding to the mechanism model to be developed. The development requirements represent a set of information used to describe the development goals, functional requirements, and model application requirements of the mechanism model corresponding to the target equipment.

3. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, The construction process of the AI ​​agent specifically includes: Feature extraction processing is performed on the preprocessed historical mechanism model development records. Corresponding features are extracted based on equipment information, development requirements, model structure information, parameter configuration information, development process information, and model application result information. Based on the correlation characteristics, the historical mechanism model development records are correlated and organized to establish the correspondence between development requirements and development process information, and to establish the correlation between equipment information and model structure information; Based on the correspondence between development requirements and development process information, the task planning capability of the AI ​​agent is constructed. The development process information is processed by task decomposition according to development requirements, and the execution relationship between development tasks and model building process is established. Based on the correlation between device information and model structure information, the ReAct inference framework is integrated to build the mechanism inference capability of AI agents; Based on the model application results, an adaptation analysis is performed on the model structure information and parameter configuration information to establish a model structure evaluation relationship. Based on the relationship between task execution and model structure evaluation, the historical mechanism model development records are linked and organized to build the knowledge retrieval capability of the AI ​​agent. By linking and configuring task planning capabilities, mechanism reasoning capabilities, and knowledge retrieval capabilities, the construction of an AI agent can be completed.

4. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, S2 specifically includes: S21. Perform feature extraction processing on the target equipment information and development requirements, and construct equipment feature representation and requirement feature representation; S22. Employing an AI agent, task planning and processing are performed based on device feature representations and demand feature representations to construct model development tasks, specifically including: By using task planning capabilities to analyze the characteristic representation of requirements, the corresponding development goals and functional requirements are determined, and task planning information is generated. By combining equipment feature representation with task planning information, the knowledge retrieval capabilities of the AI ​​agent are invoked to correlate and match development process information in the historical mechanism model development records, and to determine the task execution relationship corresponding to the development process information. The development process information is broken down into multiple development tasks according to the task execution relationship, and the development tasks are then linked and organized to generate model development tasks. S23. Based on the model development task, perform mechanism analysis and processing on the target equipment information and development requirements to determine the relationship between equipment operation rules and model influence; S24. Based on the ReAct inference framework, perform mechanism inference operations according to the model development task, equipment operation rules and model influence relationship, and extract equipment mechanism features; S25. Associate and organize the equipment mechanism characteristics with the model development tasks, and construct a mechanism characteristic sequence.

5. The AI-assisted development method based on industrial big data according to claim 4, characterized in that, S23 specifically includes: S231. Based on the model development task, perform correlation analysis and processing on the target equipment information and development requirements, extract the corresponding operating status characteristics, equipment structure characteristics and functional requirement characteristics of the target equipment, and construct the equipment feature sequence. S232. Use the variational mode decomposition algorithm to perform multi-scale feature decomposition on the equipment feature sequence, extract the state change features of the target equipment at different scales, and generate the operating feature sequence. S233. Use the attention mechanism to perform feature association analysis on the running feature sequence, calculate the association weight between different running features, and select key running features based on the association weight; S234. Calculate the Pearson correlation coefficient between key operating characteristics, and determine the equipment operating pattern based on the Pearson correlation coefficient; S235. Based on the equipment operation rules and development requirements, perform correlation analysis on the model structure information and parameter configuration information in the historical mechanism model development record to determine the multiple model structures and parameter configurations corresponding to the equipment operation rules and establish model influence relationships.

6. The AI-assisted development method based on industrial big data according to claim 4, characterized in that, S24 specifically includes: S241. Based on the model development task, equipment operation rules and model influence relationship, perform task parsing operation based on the ReAct inference framework to determine the inference target and inference content corresponding to the model development task. S242. Call upon the mechanistic reasoning ability of the AI ​​agent, and perform correlation analysis on the relationship between equipment operation rules and model influence based on the reasoning goal and reasoning content, and determine the correlation characteristics between equipment operation rules and model influence. S243. Use Bayesian inference to perform probabilistic inference processing on the association analysis results and calculate the association probability value of each association feature; S244. Calculate the confidence level corresponding to each association feature based on the association probability value, and filter out valid association features with a confidence level greater than a preset confidence threshold. Specifically, this includes: Based on the equipment operation rules, model structure information and parameter configuration information corresponding to each associated feature, the historical mechanism model development records are matched and retrieved to determine the number of times each associated feature appears in the historical mechanism model development records; Based on the ratio between the frequency of occurrence and the total number of records, the historical correlation degree corresponding to each correlation feature is calculated, where the total number of records represents the total number of records developed by the historical mechanism model. Analyze the distribution of each correlation feature in the historical mechanism model development record to determine the corresponding correlation stability. The confidence level corresponding to each association feature is calculated by combining the association probability value, historical association degree, and association stability degree. The confidence scores of each associated feature are compared with a preset confidence threshold, and valid associated features with confidence scores greater than the preset confidence threshold are selected. S245. Normalize the correlation probability values ​​of each effective correlation feature, and perform weighted fusion processing on each effective correlation feature based on the normalization result to extract the equipment mechanism features in the equipment operation rules.

7. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, S3 specifically includes: S31. Perform model planning processing on the mechanism feature sequence through AI intelligent agent, and match candidate model structures from the standard structure library according to the device mechanism features in the mechanism feature sequence; S32. Use the Bayesian optimization algorithm to perform adaptation evaluation on the structure of each candidate model. S33. Based on the adaptation evaluation results, perform screening processing on the candidate model structure, and iteratively perform adaptation evaluation and screening processing on the screening results until the preset iteration termination condition is reached, and determine the target model structure. S34. Perform parameter configuration optimization processing based on the target model structure, and determine the target parameter configuration by combining the parameter configuration information corresponding to the target model structure in the standard structure library with the equipment mechanism characteristics. S35. Generate a model construction scheme based on the target model structure and target parameter configuration.

8. The AI-assisted development method based on industrial big data according to claim 7, characterized in that, Specifically, S32 includes: S321. Obtain the model structure information and parameter configuration information corresponding to each candidate model structure in the standard structure library, and match them with the equipment mechanism features in the mechanism feature sequence one by one. Calculate the structure adaptation value and parameter adaptation value of each candidate model structure based on the matching results. S322. Call the knowledge retrieval capability of the AI ​​agent to match the model structure information corresponding to each candidate model structure with the model structure information in the historical mechanism model development record, and calculate the corresponding structure matching degree. S323. Sort the historical mechanism model development records according to the degree of structural matching, and extract the historical mechanism model development records with the highest degree of structural matching with each candidate model. S324. Match the parameter configuration information corresponding to each candidate model structure with the parameter configuration information extracted from the historical mechanism model development record, and calculate the corresponding parameter matching degree. S325. Determine the historical evaluation value based on the model application result information extracted from the historical mechanism model development record, and perform weighted processing on the historical evaluation value in combination with the degree of structure matching and the degree of parameter matching to obtain the prior evaluation result corresponding to each candidate model structure. S326. The structure fit value, parameter fit value and prior evaluation results of each candidate model structure are used as the evaluation data of the Bayesian optimization algorithm. The evaluation data are probabilistically fitted using a Gaussian process to obtain the prediction mean and prediction variance corresponding to each candidate model structure. S327. Using the expected improvement criterion in the Bayesian optimization algorithm, the predicted mean and predicted variance corresponding to each candidate model structure are jointly calculated to obtain the fitness evaluation value corresponding to each candidate model structure as the fitness evaluation result.

9. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, S4 specifically includes: S41. Establish the equipment mechanism model according to the target model structure in the model construction scheme, and configure the target parameters to the corresponding positions in the equipment mechanism model; S42. Determine the prediction target corresponding to the equipment mechanism model based on the development requirements, and extract the corresponding operating status from the target equipment information according to the prediction target; S43. Input the operating status into the device mechanism model, perform simulation and deduction processing on the prediction target through the device mechanism model, obtain the prediction status corresponding to each time step, and arrange the prediction status in time order to generate a prediction status sequence. S44. Based on the time steps of the predicted state sequence, collect the actual operating state of the target device at each corresponding time step, and perform deviation analysis processing on the predicted state and the actual operating state. S45. Arrange the deviation analysis results in chronological order to obtain the model deviation sequence.

10. The AI-assisted development method based on industrial big data according to claim 1, characterized in that, S5 specifically includes: S51. Based on the model deviation sequence, adjust the target model structure and target parameter configuration of the equipment mechanism model to optimize the equipment mechanism model; S52. Iteratively perform simulation deduction, deviation analysis and model optimization processing through the updated equipment mechanism model until the preset convergence condition is met, and then stop the iteration. The preset convergence condition means that the deviation analysis results in the model deviation sequence are all less than the preset deviation threshold, and the change amplitude of the corresponding deviation analysis results in the model deviation sequence obtained in two consecutive iterations is less than the preset change threshold. S53. Use the device mechanism model after the iteration stops as the target mechanism model to meet the development requirements.

Images