An industrial simulation optimization method and system based on artificial intelligence

Abstract

The present invention discloses an industrial simulation optimization method and system based on artificial intelligence, including: firstly obtaining the task data of the target industrial simulation task module and determining the target task category, then determining the matching current industrial simulation request from the template library based on the data, providing the request to the simulation interface after the request passes the reliability evaluation, and executing the simulation request for the task data in response to the user call instruction. Such a design can improve the accuracy, reliability and convenience of simulation, and effectively optimize the industrial simulation process.

Description

Technical Field

[0001] The present invention relates to the field of industrial simulation, and in particular to an industrial simulation optimization method and system based on artificial intelligence. Background Art

[0002] In industrial production, simulation technology is of great significance for optimizing production processes and improving production efficiency and quality. However, existing industrial simulation methods often have problems such as inaccurate task category determination, low simulation request matching, and lack of reliability evaluation, resulting in poor simulation results and unable to fully meet the actual needs of industrial production. Summary of the invention

[0003] The purpose of the present invention is to provide an industrial simulation optimization method and system based on artificial intelligence.

[0004] In a first aspect, an embodiment of the present invention provides an industrial simulation optimization method based on artificial intelligence, comprising:

[0005] Acquire task data related to the target industrial simulation task module, and determine the target task category corresponding to the task data based on a preset task feature coding model set;

[0006] According to the target task category, determining a current industrial simulation request matching the target task category from a preset industrial simulation request template library;

[0007] On the basis that the current industrial simulation request passes the reliability evaluation, providing the current industrial simulation request to the simulation interface where the target industrial simulation task module is located;

[0008] In response to the user calling the target industrial simulation task module and the simulation instruction of the current industrial simulation request through the simulation interface, the industrial simulation request for the task data is executed.

[0009] In a second aspect, an embodiment of the present invention provides a server system, including a server, wherein the server is used to execute the method described in the first aspect.

[0010] Compared with the prior art, the beneficial effects provided by the present invention include: using an artificial intelligence-based industrial simulation optimization method and system disclosed by the present invention, first obtaining the task data of the target industrial simulation task module and determining the target task category, then determining the matching current industrial simulation request from the template library based on this, providing it to the simulation interface after the request passes the reliability evaluation, and executing the simulation request for the task data in response to the user call instruction. This method can improve the accuracy, reliability and convenience of simulation, and effectively optimize the industrial simulation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0012] Figure 1 A schematic diagram of the steps of an industrial simulation optimization method based on artificial intelligence provided by an embodiment of the present invention;

[0013] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0015] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0016] In order to solve the technical problems in the aforementioned background technology, Figure 1 A flow chart of an artificial intelligence-based industrial simulation optimization method provided in an embodiment of the present disclosure is provided below, and the artificial intelligence-based industrial simulation optimization method is introduced in detail.

[0017] Step S201, acquiring task data related to a target industrial simulation task module, and determining a target task category corresponding to the task data based on a preset task feature coding model set;

[0018] Step S202, according to the target task category, determining a current industrial simulation request matching the target task category from a preset industrial simulation request template library;

[0019] Step S203, on the basis that the current industrial simulation request passes the reliability evaluation, providing the current industrial simulation request to the simulation interface where the target industrial simulation task module is located;

[0020] Step S204, in response to the user calling the target industrial simulation task module and the simulation instruction of the current industrial simulation request through the simulation interface, executing the industrial simulation request for the task data.

[0021] In an embodiment of the present invention, illustratively, it is assumed that the server receives a target industrial simulation task module from an automobile manufacturing enterprise, and its task is to optimize the assembly process of automobile engines. The server first obtains task data related to this task module, such as detailed information such as specifications, dimensions, materials, etc. of engine parts, as well as steps of the assembly process, required working hours, possible problems, etc. The task feature coding model set preset in the server contains a variety of models for different industrial fields and task types. For the acquired automobile engine assembly task data, the server will input these data into the task feature coding model related to mechanical manufacturing. The model will analyze and process the task data and extract key features, such as the matching accuracy requirements between parts, the importance of assembly sequence, the impact of working hours on cost, etc. By comparing and calculating with various types of task features already in the model, it is finally determined that the target task category corresponding to the task data is "high-precision assembly optimization of automobile parts". After the server determines that the target task category is "high-precision assembly optimization of automobile parts", it will search and match in the preset industrial simulation request template library. Assume that the preset industrial simulation request template library contains different templates such as "simulation request for improving assembly efficiency", "simulation request for reducing assembly cost", and "simulation request for optimizing assembly accuracy". The server determines that the current industrial simulation request that best matches "optimization of high-precision assembly of automotive parts" is "simulation request for optimizing assembly accuracy" by analyzing and understanding the target task category. This request template can include specific simulation requirements such as adjustment of component matching tolerances, optimization of assembly tool selection, and re-planning of assembly sequence. The server performs a reliability assessment on the "simulation request for optimizing assembly accuracy". First, obtain the request content of the current industrial simulation request to be evaluated, such as the tolerance range, tool parameters, and sequence adjustment strategy specified in the request. Then, evaluate the execution status of the current industrial simulation request. For example, obtain the relevant standards and specifications for automobile engine assembly, compare the request content with them for feature similarity to evaluate the accuracy of the request description; check whether there are spelling errors, irregular descriptions, inaccurate timing or unconventional expressions in the request to evaluate the accuracy of the request description; analyze the data parameters in the request under preset indicators such as data volume, object richness, number of component levels and domain-specific vocabulary, and evaluate the request logic depth of the request. Then, the server calls the industrial simulation architecture based on the current industrial simulation request to perform related operations on multiple control parameter instances of automobile engine assembly, obtain the real simulation operation results, and compare them with the standard simulation operation results corresponding to each instance to evaluate the evaluation score of each instance on the industrial simulation architecture. The evaluation scores of multiple industrial control parameter instances are integrated to obtain the evaluation score of the current industrial simulation request on the industrial simulation architecture. Based on this evaluation score, the simulation operation reliability of the current industrial simulation request on the industrial simulation architecture is evaluated to obtain multiple second evaluation results.Poll multiple simulation architectures, average the reliability of the simulation operations in the second evaluation results corresponding to each simulation architecture model of each architecture, and obtain the average reliability score. Based on the average reliability score, perform a score evaluation on the simulation request completion of the current industrial simulation request on the target simulation architecture to obtain the evaluation score. After multiple simulation architectures are polled, integrate the evaluation scores of the current industrial simulation request on each simulation architecture to obtain the simulation request completion score. Assume that after the evaluation, the reliability score of the current industrial simulation request reaches the preset reliability standard. The server provides the "simulation request for optimizing assembly accuracy" to the simulation interface where the target industrial simulation task module is located for the user to view and operate. The user sees the "simulation request for optimizing assembly accuracy" on the simulation interface and decides to call the target industrial simulation task module and the simulation request for simulation. After receiving the simulation instruction from the user, the server starts the corresponding simulation program. According to the specific requirements in the "simulation request for optimizing assembly accuracy", the task data of the automobile engine assembly is simulated and optimized. During the simulation process, the relevant data and results are continuously monitored and collected, such as the collision situation during the assembly process, the stress distribution of the parts, and the change of the assembly time. When the simulation is completed, the server will feed back the simulation results to the user, who can further analyze and make decisions based on these results, such as whether the assembly process needs to be adjusted and whether the expected accuracy optimization goals have been achieved.

[0022] In an embodiment of the present invention, reliability evaluation is performed on the current industrial simulation request in the following manner, which can be implemented through the following examples.

[0023] Obtaining request content of a current industrial simulation request to be evaluated, wherein the current industrial simulation request is obtained based on an industrial control parameter operation item;

[0024] According to the request content of the current industrial simulation request, the execution status of the current industrial simulation request is evaluated to obtain a first evaluation result; wherein the execution status is used to characterize the reliability of the industrial simulation architecture in executing the industrial control parameter operation item based on the current industrial simulation request;

[0025] Calling the industrial simulation architecture to execute the industrial control parameter operation item on the industrial control parameter instance based on the current industrial simulation request to obtain a real simulation operation result; and based on the real simulation operation result, evaluating the simulation operation reliability of the current industrial simulation request on the industrial simulation architecture to obtain a second evaluation result;

[0026] A reliability score of the current industrial simulation request is generated based on the first evaluation result and the second evaluation result.

[0027] In an embodiment of the present invention, illustratively, it is assumed that the server receives a current industrial simulation request from an electronic manufacturing enterprise, which is intended to optimize the production process of the circuit board and improve production efficiency and product quality. The server first obtains the detailed content of this request, including but not limited to: the design specifications of the circuit board (such as the number of layers, circuit layout), the parameter settings of the production equipment (such as temperature, pressure), the arrangement of the production process (such as the order and time of patch, welding, and detection), etc. These request contents are formulated based on actual industrial control parameter operation items. The server obtains a detailed text description of the circuit board production process as standard content, and performs feature extraction on the current industrial simulation request to obtain a first content description feature. At the same time, feature extraction is also performed on the standard production process text description to obtain a second content description feature. By comparing these two features, the server calculates the similarity between them. If the similarity is high, it means that the description accuracy of the current industrial simulation request is good; if the similarity is low, it means that the description may be biased or inaccurate. For example, the description of the temperature of the production equipment in the current industrial simulation request is "During the welding process, the equipment temperature is maintained at about 250 degrees Celsius", while the standard description is "During the welding process, the equipment temperature is precisely controlled within the range of 250 degrees Celsius plus or minus 5 degrees Celsius". Through feature extraction and comparison, it is found that the feature similarity between the two is low, and the server evaluates that the request is insufficient in description accuracy. The server obtains valid industrial simulation requests under the description reliability attribute, which are based on the circuit board production process and meet the corresponding attribute requirements. Feature extraction is performed on these valid requests to obtain the third content description feature. Then, the first content description feature of the current industrial simulation request is compared with the third content description feature to determine the request overlap. The server sequentially polls the sub-attributes in the description reliability attribute, such as the spelling correctness sub-attribute, the description specification sub-attribute, the timing accuracy sub-attribute, and the description habit adaptability sub-attribute. Taking the spelling correctness sub-attribute as an example, if there is a spelling error in the current industrial simulation request, such as writing "temperature" as "temperature", the evaluation score under this sub-attribute will be low. For the description standardization sub-attribute, if non-standard terms or expressions are used in the request, such as "almost 250 degrees" instead of "about 250 degrees Celsius", the evaluation score will also be reduced. For the timing accuracy sub-attribute, if there is a logical error in the timing of the production process in the request, such as testing before welding, the evaluation score will also be affected. For the description habit adaptability sub-attribute, if the expression of the request does not conform to the usual description habits in the industry, such as using overly colloquial or uncommon expressions, it will also lead to a reduction in the evaluation score. The evaluation scores of each sub-attribute are combined, and the lowest score is selected as the description accuracy of the current industrial simulation request.Assuming that the evaluation scores of the spelling correctness sub-attribute, description standard sub-attribute, timing accuracy sub-attribute and description habit adaptability sub-attribute are 80 points, 70 points, 60 points and 75 points respectively, the description accuracy of the current industrial simulation request is finally evaluated as 60 points. The server obtains weighting factors of multiple preset indicators (such as data volume indicator, object richness indicator, component level number indicator and domain-specific vocabulary indicator). For the description of the circuit board production process in the current industrial simulation request, the server counts the total amount of request data (i.e., the number of text elements) as the data parameter under the data volume indicator; determines the object richness involved in the request (such as different types of electronic components, production equipment, etc.) as the data parameter under the object richness indicator; counts the number of component levels of the production process (such as raw material preparation, patch, welding, testing, packaging, etc.) as the data parameter under the component level number indicator; counts the domain-specific vocabulary (such as specific electronic component names, production process terms, etc.) as the data parameter under the domain-specific vocabulary indicator. Then, use the weighting factors of each indicator to linearly superimpose these data parameters to obtain the cumulative value of the data parameters. For example, the weighting factor of the data volume indicator is 0.2, the weighting factor of the object richness indicator is 0.3, the weighting factor of the component level number indicator is 0.3, and the weighting factor of the domain-specific vocabulary indicator is 0.2. Assuming that the data parameter under the data volume indicator is 1000, the data parameter under the object richness indicator is 8, the data parameter under the component level number indicator is 5, and the data parameter under the domain-specific vocabulary indicator is 50, the cumulative value of the data parameters obtained by weighted superposition calculation is: 0.2×1000+0.3×8+0.3×5+0.2×50=200+2.4+1.5+10=213.9. This cumulative value is standardized to obtain a standardized value. Assuming that the standardized value is 0.8, the server uses this value as the request logical depth of the current industrial simulation request. Combining the above three aspects of evaluation, the server obtains the first evaluation result of the current industrial simulation request, including the evaluation values ​​of description accuracy, description accuracy and request logic depth. The server calls a special industrial simulation architecture and simulates the industrial control parameter instances of multiple circuit board production according to the parameters and job items in the current industrial simulation request. For example, 100 circuit board production cases with different specifications and requirements are selected as parameter instances. During the simulation operation, the real simulation operation results such as production efficiency data (such as the number of circuit boards produced per unit time) and product quality data (such as defective rate, welding defect rate) of each instance are collected. Then, the server obtains the standard simulation operation results corresponding to each industrial control parameter instance. These standard results can be based on historical optimal production data, industry standards or verified theoretical models. For each instance, the server compares and evaluates its real simulation operation results with the standard simulation operation results.For example, the standard production efficiency of a certain instance is to produce 50 circuit boards per hour, while the actual simulation result is to produce 45 per hour. The server calculates the evaluation score of the instance according to the predetermined evaluation rules. The evaluation scores of 100 industrial control parameter instances are integrated and statistically analyzed, and the server obtains the evaluation score of the current industrial simulation request on the industrial simulation architecture. This score reflects the effect and reliability of the current industrial simulation request in actual execution. Assuming that the average value of the evaluation scores of these 100 instances is 85 points through calculation, the server uses it as the evaluation score of the current industrial simulation request on the industrial simulation architecture. The server obtains the evaluation values ​​of description accuracy, description accuracy and request logic depth in the first evaluation result, and the evaluation score on the industrial simulation architecture in the second evaluation result. Assuming that the description accuracy is 80 points, the description accuracy is 70 points, and the request logic depth is 85 points in the first evaluation result, the evaluation score on the industrial simulation architecture in the second evaluation result is 85 points. The server obtains the importance factor of each parameter value in these evaluation results, such as the importance factor of description accuracy is 0.2, the importance factor of description accuracy is 0.3, the importance factor of request logic depth is 0.2, and the importance factor of the evaluation score on the industrial simulation architecture in the second evaluation result is 0.3. Then, the server calculates based on these importance factors, each parameter value and the evaluation score to determine the comprehensive score of the current industrial simulation request. For example: 0.2×80+0.3×70+0.2×85+0.3×85=16+21+17+25.5=79.5. The server uses this comprehensive score as the reliability score of the current industrial simulation request to measure the reliability and effectiveness of the request.

[0028] In an embodiment of the present invention, the request content of the current industrial simulation request includes: a first content description feature obtained by performing a feature extraction operation on the current industrial simulation request; the execution status of the current industrial simulation request includes the description accuracy of the current industrial simulation request;

[0029] The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result may be implemented through the following example.

[0030] Acquire content text data of the industrial control parameter operation item, and perform a feature extraction operation on the content text data to obtain a second content description feature;

[0031] Determining feature similarity between the current industrial simulation request and the content text data based on the first content description feature and the second content description feature;

[0032] The description accuracy of the current industrial simulation request is evaluated according to the determined feature similarity, and the evaluated description accuracy is added to the first evaluation result.

[0033] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of a chemical production process, specifically, the optimization of the synthesis reaction conditions of a certain chemical product to improve the purity and yield of the product. First, the server performs a feature extraction operation on the current industrial simulation request. It analyzes the key information in the request, such as the setting range of the reaction temperature, the length of the reaction time, the concentration ratio of the reactants, the type and amount of the catalyst, etc., and converts these key information into digital feature vectors to obtain the first content description feature. Next, the server obtains the content text data of the industrial control parameter operation items. These data may include detailed chemical experiment records, past production data reports, related technical documents, etc. Then, the server performs a feature extraction operation on these content text data. For example, the specific reaction conditions of each experiment and the final product purity, yield and other results are extracted from the experimental records, and converted into feature vectors to obtain the second content description feature. Next, the server determines the feature similarity between the current industrial simulation request and the content text data based on the first content description feature and the second content description feature. The server uses a special algorithm to calculate the similarity between the two feature vectors. For example, it calculates the cosine similarity between the vectors. Assume that the first content description feature vector is [0.8, 0.5, 0.7, 0.6], indicating that the reaction temperature is high, the reaction time is long, the reactant concentration ratio is moderate, and the catalyst dosage is large; the second content description feature vector is [0.7, 0.4, 0.6, 0.5], indicating similar but slightly lower reaction conditions. By calculation, the cosine similarity between them is 0.85. Then, the description accuracy of the current industrial simulation request is evaluated according to the determined feature similarity. If the similarity is high, such as greater than 0.8, it means that the current industrial simulation request is more consistent with the actual operation items, and the description accuracy is high; if the similarity is between 0.6-0.8, it means that there is a certain degree of consistency, but there may be some differences in details; if the similarity is lower than 0.6, it may indicate that there is a large deviation between the current industrial simulation request and the actual operation items, and the description accuracy is low. In this example, since the calculated similarity is 0.85, the server evaluates that the description accuracy of the current industrial simulation request is high, and adds this evaluation result to the first evaluation result. This means that the request is more accurate in describing the reaction conditions and has a higher probability of achieving better results in practical applications. For example, in another scenario, the current industrial simulation request is about the optimization of the machining process of mechanical parts, aiming to reduce machining errors and improve production efficiency. The server performs feature extraction on the request to obtain the first content description feature containing parameters such as the rotation speed, feed speed, and cutting depth of the machining tool. The content text data of the industrial control parameter operation items obtained may include detailed process parameter records and corresponding machining quality assessments for previous different batches of parts processing.The server extracts the features of these text data to obtain the second content description features. Assuming that the calculated feature similarity is 0.7, the server evaluates that the description accuracy of the current industrial simulation request is at a medium level, and it may be necessary to further refine and adjust the description of certain parameters to improve its fit with the actual operation. This evaluation result is then added to the first evaluation result to provide a basis for subsequent comprehensive evaluation. Through the above specific scenario examples, the server can accurately evaluate the description accuracy of the current industrial simulation request based on the feature similarity, thereby providing important basic data for a comprehensive evaluation of the execution status and reliability of the industrial simulation request.

[0034] In an embodiment of the present invention, the request content of the current industrial simulation request includes: a first content description feature obtained by performing a feature extraction operation on the current industrial simulation request; the execution status of the current industrial simulation request includes the description accuracy of the current industrial simulation request;

[0035] The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result may be implemented through the following example.

[0036] Obtaining a valid industrial simulation request under a description reliability attribute, wherein the valid industrial simulation request is an industrial simulation request obtained based on the industrial control parameter operation item and meeting corresponding attribute requirements;

[0037] Performing a feature extraction operation on the valid industrial simulation request to obtain a third content description feature;

[0038] Determine, based on the first content description feature and the third content description feature, a request overlap degree between the current industrial simulation request and the valid industrial simulation request;

[0039] The description accuracy of the current industrial simulation request is evaluated according to the determined request overlap, and the evaluated description accuracy is added to the first evaluation result.

[0040] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of the blast furnace ironmaking process in a steel production process, aiming to reduce energy consumption and improve the quality of molten iron. First, the server obtains valid industrial simulation requests under the description reliability attribute. These valid industrial simulation requests are obtained based on the industrial control parameter operation items of blast furnace ironmaking and meet the corresponding attribute requirements. For example, these requirements may include: using accurate professional terms to describe various parameters and operations in the ironmaking process, following the common specifications and standards in the industry, without obvious spelling and grammatical errors, and the description of the ironmaking process is logical and practical. The server obtains multiple such valid industrial simulation requests, such as a valid request that describes the temperature control range of the blast furnace, the ratio of raw materials, the intensity and time of the blast, and other key parameters in detail, and the expression is clear, standardized and accurate. Then, the server performs feature extraction operations on these valid industrial simulation requests. The extracted features may include the numerical range of key parameters, the order and key nodes of the operation steps, the professional terms used, etc., so as to obtain the third content description feature. Next, the server compares the first content description feature obtained by performing feature extraction operation on the current industrial simulation request with these third content description features to determine the request overlap between the current industrial simulation request and the valid industrial simulation request. Assuming that the description of the blast furnace temperature in the current industrial simulation request is "control the blast furnace temperature between 1500 degrees Celsius and 1600 degrees Celsius", and the description in a valid industrial simulation request is "the blast furnace temperature should be strictly controlled within the range of 1550 degrees Celsius plus or minus 50 degrees Celsius", by comparing these features, the server can calculate the overlap in temperature description. The server will perform similar comparisons and calculations on the features of various aspects to comprehensively obtain the overall request overlap. Based on the determined request overlap, the server evaluates the description accuracy of the current industrial simulation request. If the request overlap is very high, for example, more than 90%, it means that the description accuracy of the current industrial simulation request is very high; if the overlap is between 70%-90%, it indicates that the description accuracy is good, but there is still room for improvement; if the overlap is less than 70%, it may mean that there are many inaccurate descriptions in the current industrial simulation request. Assume that after calculation, the overall request overlap between the current industrial simulation request and the valid industrial simulation request is 85%. The server evaluates that the description accuracy of the current industrial simulation request is high, and adds this evaluation result to the first evaluation result. For example, in another scenario, the current industrial simulation request is about the optimization of a food processing company's production line, with the goal of improving production efficiency and ensuring product hygiene quality. The valid industrial simulation request under the description reliability attribute obtained by the server contains a detailed description of each link of the production line, such as the cleaning frequency and method of processing equipment, the processing steps of raw materials, the specifications of the packaging link, etc., and meets the relevant attribute requirements.After feature extraction of these valid requests, the third content description feature is obtained. For example, features such as the specific time interval for equipment cleaning and the type of cleaning agent used. The description of equipment cleaning in the current industrial simulation request is "regularly clean the processing equipment and use appropriate cleaning agents". Compared with the "after every 5 hours of production, use a specific brand of cleaning agent to thoroughly clean the equipment" in the valid request, the server calculates the request overlap of this part. Based on the calculation results of the feature overlap in all aspects, if the overall request overlap is 75%, the server evaluates that the description accuracy of the current industrial simulation request is good, but some descriptions need to be further clarified and refined, such as clarifying the specific time interval for cleaning and the brand of the cleaning agent. This evaluation result is then added to the first evaluation result to provide a basis for subsequent comprehensive evaluation and decision-making of industrial simulation requests.

[0041] In an embodiment of the present invention, the description reliability attribute includes sub-attributes: spelling correctness sub-attribute, description standardization sub-attribute, timing accuracy sub-attribute and description habit adaptability sub-attribute;

[0042] Among them, the attribute requirements of the spelling correctness sub-attribute include: there are no spelling errors in the industrial simulation request; the attribute requirements of the description standardization sub-attribute include: there is no non-standard description in the industrial simulation request; the attribute requirements of the timing accuracy sub-attribute include: there is no inaccurate timing in the industrial simulation request; the attribute requirements of the description convention adaptability sub-attribute include: the expression of the industrial simulation request meets the description convention corresponding to the corresponding industrial simulation request.

[0043] In the embodiment of the present invention, for example, it is assumed that the current industrial simulation request received by the server is about the optimization of the drug production process of a pharmaceutical factory. The server first checks all the text expressions involved in the industrial simulation request. For example, the expression "drug ingredients" mentioned in the request should be "drug ingredients", and the server will determine it as a spelling error. For another example, "temperature adjustment valve" should be "temperature control valve", which is also considered a spelling error. The server ensures that there are no such spelling errors in the industrial simulation request by carefully comparing standard vocabulary and common professional terms. If no spelling errors are found in the entire request, the evaluation of the spelling correctness sub-attribute is qualified. The server then evaluates whether there are non-standard descriptions in the request. For example, the description of temperature control in drug production is "about 25 to 30 degrees", which is non-standard, and the accurate description should be "temperature control between 25 and 30 degrees". For another example, the description of production time uses vague words such as "a while" instead of a specific time range, which is also not in compliance with the specification. The server will review each description in the request according to the common specifications and standards in the industry to ensure that there are no non-standard expressions. When evaluating timing accuracy, the server will focus on whether the time sequence of each production step in the request is reasonable and accurate. For example, if the request first describes the packaging of the drug and then mentions the synthesis process of the drug, there is a timing error because the drug must be synthesized before it can be packaged. For another example, for a chemical reaction, it is required to add a catalyst first and then increase the temperature, but the request writes the temperature increase step before adding the catalyst, which is also inaccurate timing. The server determines whether there is such a timing error through logical analysis of the entire production process. The server also considers whether the expression of the request conforms to the pharmaceutical industry's habit of describing the drug production process. For example, in the pharmaceutical industry, the description of drug dosage usually uses units such as "milligrams" and "grams", and if the request uses vague expressions such as "a little" and "some", it does not conform to the description habit. For another example, there are specific names for the names of drug production equipment in the industry. If uncommon or non-standard names are used in the request, it does not meet the requirements of description habit adaptability. The server evaluates the performance of current industrial simulation requests in this regard by understanding industry practices and common description methods. Assume another scenario, the current industrial simulation request is about improving the fabric dyeing process of a textile factory. For the spelling correctness sub-attribute, the server will check for spelling errors such as whether "dye" should be "dye" and whether "color uniformity" should be "color uniformity". In terms of the description standardization sub-attribute, if the request contains an ambiguous description such as "put the fabric in the dye vat for a while" or uses non-professional colloquial expressions, the server will identify it as non-standard.Regarding the timing accuracy sub-attribute, for example, the request first mentions drying the fabric, followed by the dyeing step, which is obviously inconsistent with the actual production timing, and the server will point out this error. In the description habit adaptability sub-attribute evaluation, if the description of the fabric color uses an imprecise expression such as "very bright red", while the industry usually uses specific color numbers or standard color names, the server will consider it inconsistent with the habit. Through a detailed evaluation of these sub-attributes, the server can comprehensively and accurately judge the performance of the current industrial simulation request in terms of description reliability, thereby providing a reliable basis for subsequent simulation optimization and decision-making.

[0044] In the embodiment of the present invention, a valid industrial simulation request is configured under each sub-attribute in the reliability attribute;

[0045] The valid industrial simulation requests under each sub-attribute are subjected to feature extraction operations respectively to obtain the third content description features corresponding to each sub-attribute; and the third content description features corresponding to each sub-attribute are used to determine the request overlap between the current industrial simulation request and the valid industrial simulation requests under the corresponding sub-attribute.

[0046] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the production process improvement of an automobile parts manufacturer. The valid industrial simulation request configured under the spelling correctness sub-attribute is a description of the production process of automobile parts that has been strictly proofread and has no spelling errors. The server extracts features from this valid request, for example, extracts the correctly spelled professional terms, such as "casting mold", "CNC machine tool processing", "surface treatment process", etc. Then, the part of the production process description in the current industrial simulation request is compared with this valid request. If there are no errors in the spelling of these professional terms in the current industrial simulation request, and the spelling feature overlap with the valid request is high, it means that the performance in the spelling correctness sub-attribute is good. For example, the current industrial simulation request also spells the terms such as "casting mold" and "CNC machine tool processing" accurately, so the spelling overlap is high. Under the description specification sub-attribute, the configured valid industrial simulation request is a description of the production process of automobile parts that strictly follows industry specifications and is clearly and accurately expressed. The server extracts the standardized description features, such as the accurate and standardized description of dimensional accuracy, "the tolerance of the part size is controlled within ±0.1 mm". Then, the description of dimensional accuracy and other aspects in the current industrial simulation request is compared with it. If the current industrial simulation request also describes "the diameter of the part must be controlled within 50 mm, and the tolerance is ±0.1 mm" in a standardized manner, then the overlap under the description specification sub-attribute is high, indicating that the current request meets the requirements in terms of description specifications. For the timing accuracy sub-attribute, the configured valid industrial simulation request describes the production steps of automotive parts in the correct time sequence. For example, the raw materials are cut first, then stamped, and then welded and assembled. The server extracts this correct timing feature. When evaluating the current industrial simulation request, if the production step sequence described in it is also cutting first, then stamping, and then welding and assembly, which is highly overlapped with the timing features of the valid request, it means that the current request has no problem in terms of timing accuracy. Under the description habit adaptability sub-attribute, the configured valid industrial simulation request is a production process description that fully complies with the habits of the automotive parts manufacturing industry. For example, the processing technology is named after the common terms in the industry such as "turning" and "milling". The server extracts these customary descriptive features and compares them with the current industrial simulation request. If the current request also uses idiomatic terms such as "turning" and "milling", then the overlap in the description habit adaptability sub-attribute is high, indicating that the expression habits of the current request meet industry requirements. Assume another scenario, the current industrial simulation request is about the optimization of the production line of an electronic equipment assembly plant. For the spelling correctness sub-attribute, the configured valid industrial simulation request accurately spells professional terms such as "integrated circuit board", "resistor", and "capacitor".The server extracts these spelling features and compares them with the spelling of the corresponding terms in the current request to determine the degree of spelling overlap. In the description specification sub-attribute, the valid request describes in a specification that "the operation time of each step in the assembly process should be accurate to the minute." The server extracts this specification feature and compares it with the current request. For the timing accuracy sub-attribute, the valid request describes the correct timing of first mounting, then plugging in, and then soldering. The server extracts this timing feature to evaluate the current request. Under the description custom adaptability sub-attribute, the valid request uses industry customary terms such as "wave soldering" and "reflow soldering." The server extracts and compares the current request.

[0047] In the embodiment of the present invention, the step of evaluating the description accuracy of the current industrial simulation request according to the determined request overlap degree may be implemented through the following example.

[0048] Polling each sub-attribute in the attribute describing reliability to determine a target sub-attribute currently being polled;

[0049] According to the request overlap between the current industrial simulation request and the valid industrial simulation request under the target sub-attribute, performing score evaluation on the current industrial simulation request under the target sub-attribute to obtain an evaluation score of the current industrial simulation request under the target sub-attribute;

[0050] After each sub-attribute in the description reliability attribute is polled, the description accuracy of the current industrial simulation request is determined based on the evaluation score of the current industrial simulation request under each sub-attribute.

[0051] In the embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of the production process of a food processing plant. First, the server starts to poll each sub-attribute in the description reliability attribute. The server determines that the target sub-attribute currently polled is the spelling correctness sub-attribute. It obtains the request overlap between the current industrial simulation request and the valid industrial simulation request under the spelling correctness sub-attribute. For example, the valid industrial simulation request correctly spells professional terms such as "sterilization process" and "preservative addition amount", and the current industrial simulation request also spells these terms accurately, and the overlap between the two is high. The server performs a score evaluation on the current industrial simulation request under the spelling correctness sub-attribute according to the preset scoring rules. Assuming that the score is 90 points (out of 100 points), this is the evaluation score of the current industrial simulation request under the spelling correctness sub-attribute. The server then polls the description specification sub-attribute. The description of the food processing temperature in the valid industrial simulation request is "the food processing temperature should be kept between 80 and 90 degrees Celsius", which is a clear and standardized expression. However, the description of the same content in the current industrial simulation request is "the food processing temperature is about 80 to 90 degrees", which is not standardized. Based on this, the server calculates the request overlap and gives the current industrial simulation request a score evaluation under the description specification sub-attribute, assuming a score of 70 points. The server polls the timing accuracy sub-attribute. The order of food processing steps in the valid industrial simulation request is "raw material pretreatment-heating sterilization-packaging-storage", which is clear and accurate. The current industrial simulation request is also described according to this correct timing, and the overlap between the two is high. The server evaluates the current industrial simulation request under the timing accuracy sub-attribute, assuming a score of 95 points. Finally, the description habit adaptability sub-attribute is polled. The valid industrial simulation request uses common expressions in the industry, such as "high temperature instant sterilization" and "vacuum packaging". The current industrial simulation request also uses these customary expressions, with a high overlap. The server evaluates the current industrial simulation request under the description habit adaptability sub-attribute, assuming a score of 90 points. After each sub-attribute in the description reliability attribute is polled, the server comprehensively evaluates the current industrial simulation request under each sub-attribute, for example, the spelling correctness sub-attribute is 90 points, the description standard sub-attribute is 70 points, the timing accuracy sub-attribute is 95 points, and the description habit adaptability sub-attribute is 90 points.

[0052] The server will use different methods to determine the description accuracy of the current industrial simulation request. One possible method is to calculate the average of these evaluation scores. That is: (90+70+95+90)÷4=86.25 points. Alternatively, different weights are assigned to each sub-attribute, such as 20% for spelling correctness sub-attribute, 30% for description specification sub-attribute, 30% for timing accuracy sub-attribute, and 20% for description habit adaptability sub-attribute. Then the calculation method is: 90×0.2+70×0.3+95×0.3+90×0.2=86.5 points. Through such comprehensive calculation, the server finally determines that the description accuracy of the current industrial simulation request is 86.25 points or 86.5 points (depending on the calculation method used). Assume another scenario, the current industrial simulation request is about the production process improvement of a clothing manufacturing plant. In the spelling correctness sub-attribute polling, the valid industrial simulation request correctly spells the terms "sewing machine model" and "fabric material", and the current industrial simulation request has no spelling errors, and the score is 95 points. When polling the description specification sub-attribute, the valid request specification describes that "the cutting size error shall not exceed 1 cm", and the current request is vague and scored 75 points. In the timing accuracy sub-attribute polling, the production step sequence of the valid request is "design-proofing-cutting-sewing-quality inspection", and the current request is the same, with a score of 90 points. When polling the description habit adaptability sub-attribute, the valid request uses habitual expressions such as "locking process" and "ironing treatment", and the current request also conforms to it, with a score of 85 points. After comprehensive calculation, for example, the average value is 85 points, or 84 points are calculated according to different weights, the server determines that the description accuracy of the current industrial simulation request is the corresponding score.

[0053] In the embodiment of the present invention, the determination of the description accuracy of the current industrial simulation request based on the evaluation score of the current industrial simulation request under each sub-attribute can be implemented by the following example: From the evaluation scores of the current industrial simulation request under each sub-attribute, the lowest evaluation score is selected as the description accuracy of the current industrial simulation request.

[0054] In the embodiment of the present invention, for example, it is assumed that the current industrial simulation request received by the server is about the production process optimization of a new energy battery factory. First, the server evaluates each sub-attribute in the description reliability attribute and obtains the corresponding evaluation score. In terms of the spelling correctness sub-attribute, the effective industrial simulation request accurately spells professional terms such as "electrode material" and "electrolyte formula", and the current industrial simulation request has no spelling errors. The evaluation score given by the server is 90 points. In terms of the description specification sub-attribute, the description of the battery production temperature in the effective industrial simulation request is "the temperature during the battery production process must be strictly controlled between 25 and 30 degrees Celsius, with an error of no more than 1 degree Celsius", which is standardized, accurate and detailed. However, the description of the temperature in the current industrial simulation request is only "the battery production temperature must be maintained at 25 to 30 degrees", which is relatively not standardized and accurate. After evaluation, the server gives a score of 70 points under this sub-attribute. In terms of the timing accuracy sub-attribute, the order of steps for battery production in the effective industrial simulation request is "batch-coating-rolling-slitting-packaging-formation-testing", which is logically clear and accurate. The production step sequence of the current industrial simulation request is exactly the same as that of the current industrial simulation request, and the evaluation score given by the server is 95 points. In terms of the description habit adaptability sub-attribute, the effective industrial simulation request uses common expressions in the industry, such as "pole compaction density" and "active material ratio". The current industrial simulation request also uses these habitual expressions, and the evaluation score given by the server is 85 points. The server selects the lowest evaluation score of 70 points from the evaluation scores of 90 points, 70 points, 95 points, and 85 points of the current industrial simulation request under each sub-attribute as the description accuracy of the current industrial simulation request. This means that in the evaluation of the server, the current industrial simulation request has relatively large room for improvement in description specifications. Although it performs well in other sub-attributes such as spelling correctness, timing accuracy, and description habit adaptability, the overall description accuracy is determined by the worst-performing description specification sub-attribute. Assume another scenario, the current industrial simulation request is about the assembly process optimization of a smart home appliance factory. In terms of spelling correctness sub-attribute, the effective industrial simulation request correctly spells terms such as "circuit board welding" and "shell assembly". The current industrial simulation request is also correct and the server gives 92 points. In terms of the description specification sub-attribute, the description of assembly accuracy in the valid industrial simulation request is "the assembly gap between the outer shell and the internal components shall not be greater than 0.5 mm", which is very standard and accurate. However, the current industrial simulation request states that "the assembly gap between the outer shell and the contents cannot be too large, preferably within 0.5 mm", which is not standard enough, and the server gives 75 points. In terms of the timing accuracy sub-attribute, the assembly step sequence of the valid industrial simulation request is "install the internal key components first, then assemble the outer shell, and finally test", which is the same as the current industrial simulation request, and the server gives 90 points.In terms of the description of the customary adaptability sub-attribute, the valid industrial simulation request uses customary expressions such as "assembly line operation" and "modular assembly", and the current industrial simulation request also complies with the description, and the server gives 88 points. The server selects the lowest score of 75 from 92 points, 75 points, 90 points, and 88 points as the description accuracy of the current industrial simulation request. This shows that the current industrial simulation request needs to focus on improvement and optimization in the description specification sub-attribute to improve the overall description accuracy, so as to better support subsequent industrial simulation and optimization work.

[0055] In an embodiment of the present invention, the request content of the current industrial simulation request includes data parameters of the current industrial simulation request under multiple preset indicators; the execution state of the current industrial simulation request includes the request logic depth of the current industrial simulation request;

[0056] The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result includes:

[0057] Obtaining a weighting factor for each of the plurality of preset indicators;

[0058] Using the weighting factors of the preset indicators, linearly superimposing the data parameters of the current industrial simulation request under the multiple preset indicators to obtain the accumulated value of the data parameters;

[0059] The request logic depth of the current industrial simulation request is evaluated based on the accumulated value of the data parameter, and the evaluated request logic depth is added to the first evaluation result.

[0060] In the embodiment of the present invention, it is assumed that the current industrial simulation request received by the server is about the optimization of the production line of a mobile phone manufacturer. First, multiple preset indicators are set, including data volume indicators, object richness indicators, component level number indicators and domain-specific vocabulary indicators. For the data volume indicator, it measures the number of text elements in the current industrial simulation request, such as the number of words for the detailed description of each production link, the setting of equipment parameters, etc. The object richness indicator considers the types of different objects involved in the request, such as different models of mobile phones, various production equipment, different parts, etc. The component level number indicator focuses on the number of levels into which the production process is divided, such as from raw material procurement, parts processing, assembly, testing to packaging and shipping. The domain-specific vocabulary indicator counts the number of professional terms in the request that are specific to the field of mobile phone manufacturing, such as "chip mounting process", "packaging technology", "screen resolution", etc. Next, the weighting factors of each preset indicator are obtained. Assume that the weighting factor of the data volume indicator is 0.2, the weighting factor of the object richness indicator is 0.3, the weighting factor of the component level number indicator is 0.3, and the weighting factor of the domain-specific vocabulary indicator is 0.2. Then, the server analyzes the data parameters of the current industrial simulation request under these preset indicators. In terms of the data volume indicator, the server counts the total number of text elements in the request, assuming that it is 1500, as the data parameter. In terms of the object richness indicator, the server identifies the types of objects involved in the request, such as 10 different mobile phone models, 8 production equipment, and 15 parts, a total of 33 types, which is the data parameter of object richness. In terms of the component level number indicator, the server determines that the production process is clearly divided into 6 levels, that is, 6 is the data parameter of this indicator. In terms of the domain-specific vocabulary indicator, the server counts the number of domain-specific words appearing in the request as 80. These data parameters are linearly superimposed using the weighting factors of each preset indicator. The calculation is as follows: data parameter of data volume index × weighting factor of data volume index + data parameter of object richness index × weighting factor of object richness index + data parameter of component level number index × weighting factor of component level number index + data parameter of domain-specific vocabulary index × weighting factor of domain-specific vocabulary index = 1500 × 0.2 + 33 × 0.3 + 6 × 0.3 + 80 × 0.2 = 300 + 9.9 + 1.8 + 16 = 327.7. The accumulated value of the data parameter is 327.7. Based on this accumulated value of the data parameter, the request logic depth of the current industrial simulation request is evaluated. If the accumulated value is high, it means that the request has a richer and deeper description in many aspects, and the request logic depth is deeper; conversely, if the accumulated value is low, the request logic may be relatively shallow. Assume that based on experience and preset standards, the server determines that the request logic depth corresponding to this accumulated value is "deeper" and adds this evaluation result to the first evaluation result.For example, in another scenario, the current industrial simulation request is about improving the production efficiency of a car tire manufacturer. The preset indicators are still the data volume indicator, object richness indicator, component level number indicator, and domain-specific vocabulary indicator, with weighting factors of 0.2, 0.3, 0.3, and 0.2, respectively. Under the data volume indicator, the number of text elements in the request is 1,200. In terms of object richness, it involves 8 different specifications of tires, 6 raw materials, and 10 production equipment, a total of 24 objects. The number of component levels is 5, including raw material preparation, molding, vulcanization, testing, and packaging. The domain-specific vocabulary is 60. Perform linear superposition calculation: 1200×0.2+24×0.3+5×0.3+60×0.2=240+7.2+1.5+12=260.7. The server evaluates the request logic depth as "medium" based on this cumulative value and adds it to the first evaluation result.

[0061] In an embodiment of the present invention, the plurality of preset indicators include: a data volume indicator, an object richness indicator, a component level number indicator, and a domain-specific vocabulary indicator; wherein:

[0062] The data parameter of the current industrial simulation request under the data volume indicator is the total amount of request data of the current industrial simulation request, and the total amount of request data is the number of text elements in the current industrial simulation request;

[0063] The data parameter of the current industrial simulation request under the object richness index is the object richness in the current industrial simulation request, and the object is a property used to divide different operations and different operating systems;

[0064] The data parameter of the current industrial simulation request under the component level number indicator is the component level number in the current industrial simulation request, and the component level is the data obtained by dividing the current industrial simulation request into request levels;

[0065] The data parameter of the current industrial simulation request under the domain-specific vocabulary indicator is the domain-specific vocabulary in the current industrial simulation request.

[0066] In an embodiment of the present invention, for example, it is assumed that the current industrial simulation request received by the server is about the optimization of the production process of an aircraft parts manufacturer. The server first calculates the data parameters of the current industrial simulation request under the data volume index. It will count the number of text elements in the request, including detailed specifications of the parts, step descriptions of the production process, parameter settings of the equipment, etc. For example, the request describes in detail that "the material of the wing parts is high-strength aluminum alloy, and its dimensional accuracy must be controlled within plus or minus 0.1 mm. The processing technology includes milling, drilling, grinding and other steps. The processing time and tool speed of each process are clearly specified..." The server determines the number of text elements by analyzing and counting these texts. Assuming it is 2000, then this 2000 is the data parameter of the current industrial simulation request under the data volume index, that is, the total amount of requested data. The server then evaluates the object richness in the current industrial simulation request. Objects are attributes used to divide different operations and different operating systems. For example, in this request for aircraft parts manufacturing, objects may include different types of wing parts (such as leading edge flaps, trailing edge ailerons), various processing equipment (such as CNC milling machines, drilling machines), different production materials (such as aluminum alloys, titanium alloys), different operating systems (such as quality inspection systems, logistics distribution systems), etc. After careful analysis and identification, the server determines that there are 25 types of objects involved in the request, so 25 is the data parameter of the current industrial simulation request under the object richness index. The server divides the current industrial simulation request into request levels. For example, the entire production process is divided into the raw material procurement level, the parts rough processing level, the parts fine processing level, the assembly level, the quality inspection level, and the finished product delivery level, a total of 6 levels. Then 6 is the data parameter of the current industrial simulation request under the component level number index. The server counts the domain-specific vocabulary in the current industrial simulation request. In this request for aircraft parts manufacturing, domain-specific vocabulary may include "CNC machining center", "fatigue strength test", "surface roughness", "tooling fixture", etc. After identifying and counting each one, the server determines that the number of domain-specific words in the request is 120, so 120 is the data parameter of the current industrial simulation request under the domain-specific vocabulary indicator. Assume another scenario, the current industrial simulation request is about the capacity increase of a semiconductor chip manufacturer. The server counts the text elements in the request, such as "Chip manufacturing requires the use of photolithography technology, the exposure accuracy of the photolithography machine must reach 5 nanometers, the thickness of the photoresist is controlled within a specific range, and the parameters of the etching process are set...", and determines that the number of text elements is 1,800, which is the total amount of request data under the data volume indicator.Objects can include chips of different specifications (such as 7-nanometer chips, 14-nanometer chips), various manufacturing equipment (such as lithography machines, etching machines), raw materials (such as silicon wafers), operating systems (such as production management systems, quality traceability systems), etc. After identification, there are 18 types in total, which is the data parameter of object richness. The production process is divided into silicon wafer preparation level, lithography level, etching level, ion implantation level, packaging and testing level, etc., a total of 5 levels, which is the data parameter under the indicator of the number of constituent levels. Domain-specific vocabulary such as "wafer bonding", "doping concentration", "yield rate", etc., the statistical number is 90, which is the data parameter under the indicator of domain-specific vocabulary. Through the above detailed scenario examples, the server can accurately obtain the data parameters of the current industrial simulation request under each preset indicator, providing basic data for the subsequent evaluation of the logical depth and execution status of the request.

[0067] In the embodiment of the present invention, the evaluation of the request logic depth of the current industrial simulation request based on the accumulated value of the data parameter may be implemented through the following example.

[0068] Performing a standardization operation on the accumulated value of the data parameter to obtain a standardized value;

[0069] The standardized value is used as the requested logical depth of the current industrial simulation request.

[0070] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the production quality improvement of a wind power equipment manufacturer. First, the server calculates the data parameter cumulative value according to the previously set multiple preset indicators (data volume indicator, object richness indicator, component level number indicator and domain-specific vocabulary indicator) and the corresponding weighting factors. Assume that the calculated data parameter cumulative value is 500. Next, the server performs a standardization operation on the data parameter cumulative value. The standardization operation usually maps the original data to a specific interval, such as between 0 and 1, for comparison and evaluation. The server will use the following standardization formula: Standardized value = (data parameter cumulative value-minimum value of data parameter cumulative value) / (maximum value of data parameter cumulative value-minimum value of data parameter cumulative value). Assume that in all industrial simulation requests processed in the past, the minimum value of the data parameter cumulative value is 100 and the maximum value is 1000. Then the standardized value = (500-100) / (1000-100) = 400 / 900≈0.44. The server uses this standardized value of 0.44 as the request logic depth of the current industrial simulation request. This means that in the server's evaluation system, the request logic depth of the current industrial simulation request is at a lower-middle level. This may indicate that the description of the request in some aspects is not comprehensive or in-depth enough and needs to be further improved and optimized. For example, in another scenario, the current industrial simulation request is about the improvement of production efficiency of a solar panel manufacturer. The cumulative value of the data parameters calculated by the server is 800. The minimum value of the cumulative value of the past data parameters is 200 and the maximum value is 1200. The standardized value = (800-200) / (1200-200)=600 / 1000=0.6. The server uses 0.6 as the request logic depth of the current industrial simulation request, indicating that its request logic depth is at a medium level. Through such standardized operations and evaluations, the server can make unified and objective comparisons and evaluations of different industrial simulation requests in terms of request logic depth, providing more accurate and valuable basis for subsequent processing and decision-making.

[0071] In an embodiment of the present invention, the number of the industrial control parameter instances is multiple, and the industrial simulation architecture executes the industrial control parameter operation items for each industrial control parameter instance based on the current industrial simulation request to obtain the real simulation operation results corresponding to each industrial control parameter instance;

[0072] The step of evaluating the simulation operation reliability of the current industrial simulation request on the industrial simulation architecture based on the real simulation operation result to obtain a second evaluation result includes:

[0073] Obtaining standard simulation operation results corresponding to each industrial control parameter instance;

[0074] According to the standard simulation operation results and the corresponding real simulation operation results corresponding to the industrial control parameter instances, the evaluation scores of the industrial control parameter instances on the industrial simulation architecture are evaluated;

[0075] Integrate the evaluation scores of multiple industrial control parameter instances on the industrial simulation architecture to obtain the evaluation score of the current industrial simulation request on the industrial simulation architecture;

[0076] Based on the evaluation score of the current industrial simulation request on the industrial simulation architecture, the simulation operation reliability of the current industrial simulation request on the industrial simulation architecture is evaluated to obtain a second evaluation result.

[0077] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of production energy consumption of a cement plant. There are multiple industrial control parameter instances, such as different types of cement kilns, different batches of raw materials, different production time periods, etc. First, the industrial simulation architecture executes industrial control parameter operations for each industrial control parameter instance based on the current industrial simulation request. For example, for a certain type of cement kiln, simulation operation is performed according to the parameter settings in the request (such as temperature, pressure, raw material ratio, etc.); for a certain batch of raw materials, simulation is performed according to specific processing conditions; for a certain production time period, simulation is performed according to a given production rhythm and control strategy. Through these simulation operations, the real simulation operation results corresponding to each industrial control parameter instance are obtained. For example, the actual energy consumption data of a certain type of cement kiln, the quality index of cement produced by a certain batch of raw materials, the output in a certain time period, etc. The server obtains the standard simulation operation results corresponding to each industrial control parameter instance. These standard results can be based on past optimal production data, industry standards, or theoretical models that have been verified many times. Then, the server evaluates the evaluation score of each industrial control parameter instance on the industrial simulation architecture according to the standard simulation operation results and the corresponding real simulation operation results corresponding to each industrial control parameter instance. Assume that for a certain type of cement kiln, the standard unit energy consumption is 100 kWh per ton of cement, while the unit energy consumption in the real simulation operation result is 110 kWh per ton of cement. According to the preset evaluation rules, the server calculates the evaluation score of this instance as 80 points (out of 100 points). For another batch of raw materials, the standard cement strength should reach 50MPa, while the cement strength in the real simulation result is 48MPa, and the server evaluates its score as 85 points. For a certain production period, the standard output should be 100 tons per hour, the real output is 90 tons per hour, and the evaluation score is 75 points. Next, the server integrates the evaluation scores of multiple industrial control parameter instances on the industrial simulation architecture. Assume that there are 10 industrial control parameter instances, and the evaluation scores are 80 points, 85 points, 75 points, 90 points, 88 points, 78 points, 82 points, 92 points, 86 points, and 80 points respectively. The server can obtain the evaluation score of the current industrial simulation request on the industrial simulation architecture by calculating the average value, weighted average value or other integration methods. If the average value is used for calculation, the evaluation score is: (80+85+75+90+88+78+82+92+86+80)÷10=83 points. Based on the evaluation score of 83 points of the current industrial simulation request on the industrial simulation architecture, the server evaluates the reliability of the simulation operation of the current industrial simulation request on the industrial simulation architecture. If the preset reliability standard is 80 points or more for reliability, then the server derives the second evaluation result that the simulation operation of the current industrial simulation request on the industrial simulation architecture is reliable.For example, in another scenario, the current industrial simulation request is about improving the paper quality of a paper mill. Industrial control parameter instances include different types of pulp, different paper machine equipment, and different production environment conditions. The industrial simulation architecture simulates each instance to obtain real simulation operation results, such as paper thickness, flatness, strength, and other indicators. The server obtains the corresponding standard simulation operation results, such as the standard paper thickness should be between 0.1 mm and 0.2 mm, the flatness error should not exceed 0.05 mm, and the strength should reach a certain value. Then the evaluation score of each instance is evaluated separately. For example, for a certain type of pulp, the actual paper thickness is 0.18 mm, the flatness error is 0.04 mm, the strength meets the requirements, and the evaluation score is 90 points. For a certain paper machine equipment, the actual paper flatness is slightly worse, and the evaluation score is 80 points. For a certain production environment condition, the actual paper strength does not meet the standard, and the evaluation score is 70 points. Assume that there are 8 instances in total, and the evaluation scores are 90, 80, 70, 85, 95, 88, 75, and 82 respectively. The evaluation score is calculated using the average value: (90+80+70+85+95+88+75+82)÷8=83.125. Based on this evaluation score of 83.125, the server evaluates the simulation operation reliability of the current industrial simulation request on the industrial simulation architecture and obtains the corresponding second evaluation result.

[0078] In an embodiment of the present invention, the number of the industrial simulation frameworks is multiple, and one industrial simulation framework corresponds to one second evaluation result;

[0079] The generating, based on the first evaluation result and the second evaluation result, a reliability score of the current industrial simulation request includes:

[0080] Based on the reliability of the simulation operation in the second evaluation results corresponding to the multiple industrial simulation architectures, a score evaluation is performed on the simulation request completion of the current industrial simulation request to obtain a simulation request completion score;

[0081] The first evaluation result and the simulation request completion score are used to generate a reliability score for the current industrial simulation request.

[0082] In an embodiment of the present invention, for example, it is assumed that the industrial simulation request currently received by the server is about the improvement of production efficiency of an automobile engine manufacturing plant. There are multiple industrial simulation architectures here, such as architecture A, architecture B and architecture C. First, each industrial simulation architecture will process the current industrial simulation request and generate a corresponding second evaluation result. When processing the request, architecture A simulates the equipment layout, worker operation process, etc. on the production line based on its specific algorithm and model to obtain a second evaluation result. Assume that the evaluation result shows that under this architecture, according to the setting of the current industrial simulation request, the production efficiency is expected to increase by 15%. Architecture B focuses more on the simulation of raw material supply and inventory management, and the second evaluation result obtained is that the production efficiency is expected to increase by 12%. Architecture C comprehensively considers factors such as market demand fluctuations and quality control, and its second evaluation result is that the production efficiency may increase by 10%. Based on the simulation operation reliability in the second evaluation results corresponding to these multiple industrial simulation architectures, the server performs a score evaluation on the simulation request completion of the current industrial simulation request to obtain the simulation request completion score. For example, the server sets a certain scoring standard: if the production efficiency improvement is expected to exceed 15%, the simulation request completion score is above 90 points; between 10% and 15%, the score is 80-90 points; between 5% and 10%, the score is 70-80 points; below 5%, the score is below 70 points. According to this standard, the simulation request completion score of architecture A is 85 points, architecture B is 80 points, and architecture C is 75 points. Next, the server obtains the first evaluation result obtained by the previous evaluation. Assume that in the first evaluation result, the description accuracy of the current industrial simulation request is 85 points, the description accuracy is 75 points, and the request logic depth is 80 points. Then, the server uses the first evaluation result and the simulation request completion score to generate a reliable score for the current industrial simulation request. One possible calculation method is to assign different weights to each item in the first evaluation result and the simulation request completion score. Assume that the description accuracy weight is 20%, the description accuracy weight is 30%, the request logic depth weight is 20%, and the simulation request completion score weight is 30%. The calculation is as follows: description accuracy × weight + description accuracy × weight + request logic depth × weight + simulation request completion score × weight = 85 × 0.2 + 75 × 0.3 + 80 × 0.2 + 80 × 0.3 = 17 + 22.5 + 16 + 24 = 79.5. In the end, the server generates a reliable score of 79.5 for the current industrial simulation request. For another example, the current industrial simulation request is about product quality optimization in an electronic product assembly plant. Industrial simulation architecture D mainly simulates the quality inspection and screening process of parts, and the second evaluation result is that the product defect rate can be reduced by 8%. Industrial simulation architecture E focuses on the improvement of assembly process, and it is expected that the product defect rate can be reduced by 6%.Industrial simulation architecture F takes into account the impact of the production environment on product quality and obtains an evaluation result that the product defect rate can be reduced by 5%. According to the set scoring criteria, the simulation request completion score of architecture D is 80 points, architecture E is 75 points, and architecture F is 70 points. In the first evaluation result, the description accuracy is 78 points, the description accuracy is 72 points, and the request logic depth is 75 points. The same weight calculation method is used: 8×0.2+72×0.3+75×0.2+75×0.3=15.6+21.6+15+22.5=74.7. The reliability score generated by the server is 74.7 points.

[0083] In an embodiment of the present invention, the multiple industrial simulation frameworks include multiple simulation framework models of multiple simulation frameworks, and one simulation framework model is one industrial simulation framework;

[0084] The simulation job reliability in the second evaluation results corresponding to the multiple industrial simulation architectures is used to perform a score evaluation on the simulation request completion of the current industrial simulation request to obtain a simulation request completion score, which can be implemented through the following examples.

[0085] Polling each simulation framework of the plurality of simulation frameworks to determine a target simulation framework currently being polled;

[0086] Performing an averaging operation on the reliability of the simulation operations in the second evaluation results corresponding to each simulation architecture model of the target simulation architecture to obtain an average reliability score;

[0087] Based on the average reliability score, performing a score evaluation on the simulation request completion degree of the current industrial simulation request on the target simulation architecture to obtain an evaluation score of the current industrial simulation request on the target simulation architecture;

[0088] After each of the multiple simulation frameworks is polled, the evaluation scores of the current industrial simulation request on each simulation framework are integrated to obtain a simulation request completion score.

[0089] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of the production process of a steel plant, aiming to reduce costs and improve product quality. There are multiple industrial simulation architectures here, such as architecture A, architecture B and architecture C, and each architecture contains multiple simulation architecture models. First, the server starts to poll each simulation architecture in the multiple simulation architectures. Polling to architecture A, it is determined as the target simulation architecture of the current polling. Architecture A contains three simulation architecture models, namely model A1, model A2 and model A3. The server averages the reliability of the simulation operation in the second evaluation results corresponding to model A1, model A2 and model A3. Assuming that the simulation operation reliability evaluation of model A1 is 85%, model A2 is 90%, and model A3 is 80%, the average reliability score is: (85%+90%+80%)÷3=85. Based on this average reliability score of 85%, the server performs a score evaluation on the simulation request completion of the current industrial simulation request on architecture A. The server may set that if the average reliability score is above 80%, the evaluation score on the architecture is 90 points; between 70%-80%, the evaluation score is 80 points; between 60%-70%, the evaluation score is 70 points; below 60%, the evaluation score is below 60 points. Since the average reliability score is 85%, the evaluation score of the current industrial simulation request on architecture A is 90 points. Next, the server polls architecture B and determines it as the target simulation architecture currently being polled. Architecture B contains two simulation architecture models, namely model B1 and model B2. Assuming that the simulation job reliability evaluation of model B1 is 75% and that of model B2 is 85%, the average reliability score is: (75%+85%)÷2=80%. Based on the 80% average reliability score, according to the score evaluation setting, the evaluation score of the current industrial simulation request on architecture B is 80 points. Then, the server polls architecture C and determines it as the target simulation architecture currently being polled. Architecture C contains four simulation architecture models, namely, Model C1, Model C2, Model C3, and Model C4. Assuming that the simulation job reliability evaluation of Model C1 is 90%, Model C2 is 88%, Model C3 is 85%, and Model C4 is 92%, the average reliability score is: (90%+88%+85%+92%)÷4=89%. Based on the 89% average reliability score, the evaluation score of the current industrial simulation request on Architecture C is 90 points. After each simulation architecture in multiple simulation architectures is polled, the server integrates the evaluation scores of the current industrial simulation request on each simulation architecture. Assume that the integration is performed in a simple average manner, that is, (90+80+90)÷3=86.67 points. In the end, the server obtains a simulation request completion score of approximately 86.67 points. For another example, the current industrial simulation request is about the energy-saving and emission-reduction improvement of a chemical plant. The multiple simulation architectures that exist include Architecture D, Architecture E, and Architecture F.Polling to architecture D, which contains two simulation architecture models. Assuming that the simulation job reliability of model D1 is 70% and that of model D2 is 75%, the average reliability score is 72.5%, and the evaluation score on this architecture is 70 points. Polling to architecture E, which contains three simulation architecture models, assume that the reliability evaluations are 80%, 85%, and 78% respectively, and the average reliability score is 81%, and the evaluation score on this architecture is 85 points. Polling to architecture F, which contains three simulation architecture models, assume that the reliability evaluations are 90%, 88%, and 92% respectively, and the average reliability score is 90%, and the evaluation score on this architecture is 95 points. Integrate these evaluation scores, assuming that the weighted average is used, the weight of architecture D is 0.2, the weight of architecture E is 0.3, and the weight of architecture F is 0.5, then: 70×0.2+85×0.3+95×0.5=87 points. The final simulation request completion score is 87 points.

[0090] In an embodiment of the present invention, based on the average reliability score, a score evaluation is performed on the simulation request completion degree of the current industrial simulation request on the target simulation architecture to obtain the evaluation score of the current industrial simulation request on the target simulation architecture, which can be implemented through the following example.

[0091] Determine the deviation value between the simulation operation reliability in the second evaluation result corresponding to each simulation architecture model of the target simulation architecture and the reliability average score respectively, and obtain a plurality of deviation values;

[0092] Based on the multiple deviation values, determine the reliability confidence of the current industrial simulation request on the target simulation architecture; wherein the reliability confidence is used to characterize the reliability between the simulation job reliability of the current industrial simulation request on each simulation architecture model of the corresponding simulation architecture;

[0093] Based on the reliability confidence of the current industrial simulation request on the target simulation architecture, a score evaluation is performed on the simulation request completion degree of the current industrial simulation request on the target simulation architecture to obtain an evaluation score of the current industrial simulation request on the target simulation architecture.

[0094] In an embodiment of the present invention, illustratively, it is assumed that the current industrial simulation request received by the server is about the optimization of the drug production process of a pharmaceutical factory, aiming to improve the purity and production efficiency of drugs. There is a target simulation architecture, such as architecture X, which includes three simulation architecture models: model X1, model X2 and model X3. First, the server calculates the average reliability score. Assuming that the simulation operation reliability of model X1 is 88%, model X2 is 92%, and model X3 is 85%, the average reliability score is: (88+92+85%)÷3=88.33. Next, the deviation value between the simulation operation reliability and the average reliability score in the second evaluation result corresponding to each simulation architecture model of the target simulation architecture is determined respectively. For model X1, the deviation value is (88%-88.33%=-0.33%) For model X2, the deviation value is (92%-88.33%=3.67%). For model X3, the deviation value is (85%-88.33%=-3.33%). Based on these deviation values, the server determines the reliability confidence of the current industrial simulation request on the target simulation architecture. The server may use an algorithm. For example, if the absolute values ​​of the deviation values ​​are small, it means that the results of each model are relatively close, and the reliability confidence is high; if the deviation is large, it means that the results between the models are very different, and the reliability confidence is low. Assume that based on the above deviation values, the server calculates the reliability confidence to be 80%. This means that the reliability of the simulation job reliability of the current industrial simulation request on each model of the target simulation architecture is 80%. Then, based on the reliability confidence of the current industrial simulation request on the target simulation architecture, the simulation request completion degree of the current industrial simulation request on the target simulation architecture is evaluated by score. The server may set the following rules: if the reliability confidence is higher than 90%, the evaluation score is above 95 points; between 80%-90%, the evaluation score is 85-95 points; between 70%-80%, the evaluation score is 75-85 points; below 70%, the evaluation score is below 75 points. Since the reliability confidence is 80%, the server gives the current industrial simulation request an evaluation score of 85 points on the target simulation architecture. For another example, the current industrial simulation request is about the automation upgrade of a production line in a food processing plant. The target simulation architecture, such as architecture Y, contains four simulation architecture models: model Y1, model Y2, model Y3, and model Y4. Assuming that the simulation operation reliability of model Y1 is 75%, model Y2 is 80%, model Y3 is 78%, and model Y4 is 82%, the average reliability score is: (75%+80%+78%+82%)÷4=78.75%. Deviation value: For model Y1, the deviation value is (75%-78.75%=-3.75%). For model Y2, the deviation value is (80%-78.75%=1.25%).For model Y3, the deviation value is (78%-78.75%=-0.75%). For model Y4, the deviation value is (82%-78.75%=3.25%). Assume that the reliability confidence calculated based on these deviation values ​​is 70%. According to the score evaluation rule, since the reliability confidence is 70%, the server gives the current industrial simulation request an evaluation score of 75 points on the target simulation architecture.

[0095] In an embodiment of the present invention, the first evaluation result includes parameter values: description accuracy, description accuracy, and request logic depth;

[0096] The step of using the first evaluation result and the simulation request completion score to generate a reliability score for the current industrial simulation request includes:

[0097] Obtaining the importance factor of each parameter value in the first evaluation result and the importance factor of the simulation request completion score;

[0098] Determine a comprehensive score of the current industrial simulation request based on each acquired importance factor, each parameter value in the first evaluation result, and the completion score of the simulation request;

[0099] The comprehensive score of the current industrial simulation request is used to construct a reliable score of the current industrial simulation request.

[0100] In the embodiment of the present invention, for example, it is assumed that the current industrial simulation request received by the server is about the improvement of the production process of a textile factory to reduce the defective rate and increase the production speed. First, the first evaluation result obtained by the server includes three parameter values: description accuracy, description accuracy, and request logic depth. In terms of description accuracy, the server evaluates the degree of conformity between the description of the production process in the current industrial simulation request and the actual situation. For example, when describing the spinning speed of the yarn, whether it accurately reflects the actual possible range, the description accuracy value given is 80 points. In terms of description accuracy, check whether there are vague, non-standard or ambiguous expressions in the request, such as whether the description of the maintenance cycle of the textile equipment is clear and unambiguous, and the description accuracy value evaluated is 75 points. In terms of request logic depth, analyze whether the request is comprehensive and in-depth when considering the factors affecting the production process, such as whether the raw material quality, equipment performance, and worker operation level are comprehensively considered, and the request logic depth value given is 85 points. Next, the server obtains the importance factor of each parameter value in the first evaluation result. Assume that the importance factor of description accuracy is 0.2, the importance factor of description accuracy is 0.3, and the importance factor of request logic depth is 0.3. At the same time, the importance factor of the simulation request completion score is obtained, assuming it is 0.2. Then, the server determines the comprehensive score of the current industrial simulation request based on each importance factor obtained, the parameter values ​​in the first evaluation result, and the simulation request completion score. Assume that the simulation request completion score is 80 points. The calculation is as follows: description accuracy × importance factor + description accuracy × importance factor + request logic depth × importance factor + simulation request completion score × importance factor = 80 × 0.2 + 75 × 0.3 + 85 × 0.3 + 80 × 0.2 = 16 + 22.5 + 25.5 + 16 = 80. Finally, the server uses this comprehensive score of 80 points to construct the reliability score of the current industrial simulation request. This means that in the server's evaluation system, the reliability of the current industrial simulation request for improving the production process of a textile factory is rated as 80 points. For example, in another scenario, the current industrial simulation request is about the optimization of the circuit board production process of an electronics factory. The description accuracy of the server evaluation is 78 points, the description accuracy is 82 points, and the request logic depth is 75 points. The importance factors of each parameter value are 0.2, 0.3, and 0.3 respectively. The simulation request completion score is 70 points, and its importance factor is 0.2. Calculate the comprehensive score: 78×0.2+82×0.3+75×0.3+70×0.2=15.6+24.6+22.5+14=76.7. The server constructs the reliability score of the current industrial simulation request with 76.7 points, indicating that the simulation request for the optimization of the circuit board production process has a certain degree of reliability, but may need further improvement and perfection in some aspects.

[0101] In an embodiment of the present invention, the process of evaluating the execution status of the current industrial simulation request includes: determining the feature similarity between the content text data of the current industrial simulation request and the industrial control parameter operation item;

[0102] The method further comprises:

[0103] Based on the numerical value of the parameter value of the feature similarity, evaluating the key point consistency between the description key points of the current industrial simulation request and the description key points of the content text data;

[0104] If the key point consistency is evaluated, the reliability evaluation step of the current industrial simulation request is executed;

[0105] If the key point consistency is not evaluated, the current industrial simulation request is determined to be an illegal industrial simulation request, and the reliability evaluation step of the current industrial simulation request is terminated.

[0106] In an embodiment of the present invention, for example, it is assumed that the current industrial simulation request received by the server is about the optimization of the production line of an automobile parts factory, aiming to improve the accuracy of parts and reduce the scrap rate. First, the server determines the feature similarity between the content text data of the current industrial simulation request and the industrial control parameter operation item. The server will analyze the text description in the current industrial simulation request, such as the accuracy adjustment of processing equipment, the improvement of process flow, the setting of quality inspection standards, etc., and extract key features. At the same time, the detailed text data of the industrial control parameter operation items, including past production records, equipment parameter ranges, quality standard requirements, etc., are also feature extracted. Then the feature similarity between the two is calculated by a specific algorithm. Assume that the calculated feature similarity is 0.75. Based on the numerical value of the parameter value of this feature similarity, the server evaluates the key point consistency between the description key points of the current industrial simulation request and the description key points of the content text data. The server will further carefully compare the specific measures for improving accuracy in the current industrial simulation request, such as the frequency of tool replacement, the adjustment of cutting speed, and other description key points, with the historical experience and standard requirements for accuracy control in the industrial control parameter operation items. If the server finds that these description key points match each other and can be logically and technically consistent, that is, the key point consistency is evaluated. Then the server will continue to perform the reliability evaluation step of the current industrial simulation request. For example, start to obtain valid industrial simulation requests, extract features, evaluate description accuracy, request logic depth, etc. However, if the server finds that the current industrial simulation request proposes to improve the accuracy of parts by significantly increasing the operating speed of production equipment, and the content text data of the industrial control parameter operation items indicates that the equipment will cause the accuracy to decrease if it runs beyond a certain speed, there is an obvious conflict between the description key points of the two, and the key point consistency is not evaluated. At this time, the server will determine the current industrial simulation request as an illegal industrial simulation request and end the reliability evaluation step of the current industrial simulation request. Because such a request does not conform to the actual production experience and technical requirements, it is not worth further evaluating the reliability. For example, in another scenario, the current industrial simulation request is about the optimization of the chemical reaction process of a chemical plant to improve the purity and yield of the product. The feature similarity determined by the server is 0.6. When evaluating the consistency of key points, the server compares the key points described in the current industrial simulation request, such as the adjustment of reaction temperature and pressure, and the use of catalysts, with the key points described in the previous optimal reaction conditions and limiting factors recorded in the industrial control parameter operation items. If it is found that the increase in the amount of a certain catalyst proposed in the request to improve purity and yield is consistent with previous experimental data and theoretical analysis, and the key point consistency is evaluated, the server will continue to conduct reliability assessments, such as calling the industrial simulation architecture for simulation operations.However, if the request proposes to increase the reaction temperature to a level far beyond the equipment's tolerance without changing other conditions, which conflicts with the equipment restrictions in the industrial control parameter operation items, and the key point consistency is not evaluated, the server will determine that the request is illegal and end the reliability evaluation. Through the above detailed scenario examples, the server can accurately evaluate the key point consistency based on feature similarity, thereby making a judgment on the legality of the current industrial simulation request and deciding whether to continue the reliability evaluation.

[0107] In an embodiment of the present invention, the acquisition of task data related to the target industrial simulation task module and determination of the target task category corresponding to the task data based on a preset task feature coding model set may be implemented through the following examples.

[0108] Acquire task data related to a target industrial simulation task module, and extract task features of the task data;

[0109] Extracting time series distribution features according to past task data and past task categories corresponding to the target industrial simulation task module at multiple past time nodes;

[0110] Merging the task feature with the time series distribution feature execution feature to obtain a merged feature;

[0111] The combined feature is converted to a feature domain where the pending task category feature is located, and a feature association operation between the combined feature and the task category feature is performed on the feature domain to obtain an association confidence between the combined feature and the pending task category feature;

[0112] A target task category feature is determined according to the association confidence, and a task category corresponding to the target task category feature is determined as the target task category corresponding to the task data.

[0113] In an embodiment of the present invention, exemplarily, first, the server obtains task data related to the target industrial simulation task module. These data may include detailed information such as the operation time of each workstation on the assembly line, the interval of parts supply, and the proficiency of the workers. At the same time, the server extracts the task features of these task data, such as extracting the distribution range of the operation time, the frequency characteristics of the parts supply, the level characteristics of the workers' skill level, etc. Then, the server obtains the past task data and past task categories corresponding to the target industrial simulation task module at multiple past time nodes. For example, data related to similar assembly line efficiency at different time nodes such as the past week, month, and quarter and the task categories determined at that time. By analyzing these past data, the server extracts their time series distribution features, such as the trend of task difficulty changing over time, the stage characteristics of efficiency improvement, etc. Next, the server merges the currently extracted task features with the above-obtained time series distribution feature execution features. Assuming that the current operation time distribution in the task feature is more concentrated in a certain interval, and the time series distribution feature shows that the operation time has a trend of gradually shortening in the past period of time, a more comprehensive feature representation is formed after merging, which includes the current specific features and historical change trends. The server converts this merged feature to the feature domain where the pending task category feature is located. For example, the pending task category feature may be divided based on different efficiency improvement strategies, such as optimizing workstation layout, improving parts supply management, strengthening employee training, etc. In this feature domain, the server performs feature association operations between the merged feature and the task category feature. Assuming that the result of the association operation shows that the merged feature has a high correlation with the task category feature of "optimizing workstation layout", the server calculates the correlation confidence between the two. If this correlation confidence exceeds the preset threshold, such as 80%, the server determines that the task category feature of "optimizing workstation layout" is the target task category feature. Finally, the server determines the task category of "optimizing workstation layout" as the target task category corresponding to the current task data, thereby clarifying the direction for subsequent simulation and optimization work. For another example, the target industrial simulation task module is about the optimization of the distribution path of a logistics distribution center. The task data obtained by the server includes the quantity of goods at each distribution point in different time periods, the driving speed of vehicles, traffic congestion, etc. The extracted task features can be the peak and trough time periods of the quantity of goods, the range of changes in the average driving speed, etc. The acquired past task data and categories cover the delivery conditions and corresponding optimization strategy categories during different seasons and promotional activities. The extracted time series distribution features may reflect the increased delivery pressure during holidays, the changes in the traffic conditions of certain roads in specific seasons, etc. After merging the current task features and time series distribution features, a comprehensive feature containing current real-time features and historical rules is obtained.Convert it to the pending task category feature domain, such as "optimize vehicle scheduling algorithm", "increase the number of delivery vehicles", "adjust delivery route planning", etc. Through feature association operations and confidence calculations, if it is found that the association confidence with "adjust delivery route planning" is high, it will be determined as the target task category feature, and the corresponding task category "adjust delivery route planning" will become the target task category corresponding to the current task data. Through the above detailed scenario examples, the server can effectively determine the target task category corresponding to the task data, providing accurate guidance for industrial simulation and optimization.

[0114] In an embodiment of the present invention, extracting time series distribution features according to past task data and past task categories corresponding to the target industrial simulation task module at multiple past time nodes includes:

[0115] Extracting task features of each past task data, and extracting task category features of each past task category;

[0116] Merging the task feature corresponding to the same past time node with the task category feature execution feature to obtain the past merged features corresponding to each of the multiple past time nodes;

[0117] According to the time series relationship, a sequence arrangement operation is performed on the past merged features corresponding to the multiple past time nodes, and a feature encoding operation is performed on the feature array obtained by the sequence arrangement operation to obtain the time series distribution feature.

[0118] In an embodiment of the present invention, for example, it is assumed that the target industrial simulation task module received by the server is about the production process optimization of a garment factory, aiming to improve production efficiency and product quality. First, the server extracts the task features of each past task data, and extracts the task category features of each past task category. For past task data, for example, at a certain time node in the past, the task data obtained by the server includes the number of orders for different styles of clothing, the types and quantities of required fabrics, the operating status of production equipment, the working hours of workers, etc. The server analyzes these data and extracts task features, such as the distribution characteristics of the order quantity, the main types and quantity ratios of fabric requirements, the stability indicators of equipment operation, and the concentrated time periods of workers' working hours. At the same time, for the corresponding past task categories, such as "optimizing cutting process", "improving sewing technology", "improving quality inspection efficiency", etc., the server also extracts the task category features. For example, the "optimizing cutting process" category may have the characteristics of reducing fabric waste and increasing cutting speed, etc.; the "improving sewing technology" category may have the characteristics of reducing sewing defect rate and increasing sewing thread strength, etc.; the "improving quality inspection efficiency" category may have the characteristics of shortening quality inspection time and increasing defective product detection rate, etc. Next, the server merges the task features corresponding to the same past time node with the task category feature execution features to obtain the past merged features corresponding to multiple past time nodes. Assume that at the time node one month ago, the task feature was that the number of orders soared, the equipment had a slight malfunction, and the workers worked overtime more; the task category feature was "optimizing cutting process", which aims to increase the cutting speed to cope with the increase in orders. The server merges these features to form the past merged features of the time node, which may be described as "due to the surge in orders, equipment failures, and workers working overtime, the speed is improved by optimizing the cutting process." Then, the server performs a sequence arrangement operation on the past merged features corresponding to multiple past time nodes according to the time series relationship, and performs a feature encoding operation on the feature array obtained by the sequence arrangement operation to obtain the time series distribution features. For example, there are different task features and task category features merged in the past three months. The server arranges these past merged features into a sequence in chronological order. Then, a specific encoding method is used to encode this sequence, such as converting different features into digital encoding or vector representation. The result obtained after encoding is the time series distribution characteristics of the production situation and task processing of the garment factory in the past three months, which may show information such as the changing trend of production demand and the shift of task processing focus. For another example, the target industrial simulation task module is about the optimization of the new product R&D process of a food processing plant. For past task data, for example, at a certain time point two months ago, the task data obtained included the market demand research results of new products, the supply of raw materials, the staffing of the R&D team, the frequency of use of experimental equipment, etc.The extracted task features can be the strong market demand for a certain flavor, the shortage of certain raw materials, the distribution of professional backgrounds of R&D personnel, the high-load operation of experimental equipment, etc. For past task categories, such as "improving formula design", "optimizing experimental process", "strengthening quality inspection", etc., the extracted task category features can be reducing costs and improving taste under the category of "improving formula design"; shortening the experimental cycle and improving the success rate under the category of "optimizing experimental process"; strict standards and improving detection accuracy under the category of "strengthening quality inspection". Merge the task features and task category features at the same time node. For example, at the time node one and a half months ago, the merged feature can be "tight raw material supply, inexperienced R&D personnel, and reducing costs and adapting to market demand by improving formula design". Arrange these merged features according to the time sequence relationship, such as the order from one month ago to now. Encode the arranged feature array to obtain time series distribution features that can reflect various changes and trends in the new product development process of food processing plants, such as the solution process of raw material supply problems and the shift of R&D focus from formula to experimental process.

[0119] In the embodiment of the present invention, the extraction of task features of each past task data can be performed and implemented through the following examples.

[0120] Determining the data categories included in the past task data;

[0121] Selecting at least one feature extraction model corresponding to the data category from the preset task feature encoding model set, and loading the past task data into the at least one feature extraction model to obtain a feature representation of the corresponding data category generated by each feature extraction model;

[0122] The task features obtained by integrating the feature representations generated by each feature extraction model are determined as the past task features of the past task data.

[0123] In an embodiment of the present invention, for example, it is assumed that the server receives a target industrial simulation task module for improving the efficiency of a filling production line of a beverage production plant. First, the server determines the data categories contained in the past task data. For example, the past task data contains the data categories of the performance parameters of the filling equipment, covering the filling speed, failure rate and other information of the equipment; it also contains the data categories of the beverage formula, such as the proportion of various ingredients, sweetness, etc.; and the data categories of the production personnel's operation process, including operation steps, average time, etc. Next, the server selects the feature extraction model corresponding to these data categories from the preset task feature encoding model set. For the performance parameters of the filling equipment, select a model that can extract features such as equipment operation stability and speed change law; for the beverage formula, select a model that can analyze the influence of the ingredient ratio on the taste and production difficulty; for the production personnel's operation process, select a model that can obtain features such as the degree of operation standardization and efficiency influencing factors. Then, the server loads the past task data into the selected feature extraction model. Input the performance parameter data of the filling equipment into the corresponding model to obtain feature representations such as the average filling speed of the equipment is 50 bottles per minute and the failure rate is about 5%; input the beverage formula data into the corresponding model to obtain feature representations such as the increase in the proportion of a certain ingredient will increase the sweetness but increase the difficulty of production; input the production personnel operation process data into the model to obtain feature representations such as the average time for each operation step is 10 seconds and the degree of operation standardization is 80%. Finally, the server integrates the feature representations generated by each feature extraction model. The performance characteristics of the filling equipment, the characteristics of the beverage formula, and the characteristics of the production personnel operation process are combined to determine the past task characteristics of the past task data. For example, the integrated past task characteristics may indicate that under the current filling equipment performance and beverage formula conditions, according to the specific production personnel operation process, the efficiency of the filling production line is at a certain level. For another example, the server receives a target industrial simulation task module for the optimization of wood processing technology for a furniture manufacturer. The server determines the data categories contained in the past task data, such as the type and quality data category of wood, the parameter data category of processing tools, and the data category of worker skill level. From the preset task feature coding model set, select a model that can extract features such as wood hardness and texture characteristics for the type and quality of wood; select a model that can extract features such as tool sharpness and equipment power for the parameters of processing tools; select a model that can obtain features such as operating proficiency and years of experience for the skill level of workers. Load past task data into these models. For example, input the wood type and quality data into the model to obtain feature representations such as wood hardness of a certain value and texture uniformity; input the processing tool parameter data into the model to obtain feature representations such as tool sharpness retention time and equipment maximum power; input the worker skill level data into the model to obtain feature representations such as the average worker operating proficiency of intermediate and the years of experience of about 3 years.The server integrates these feature representations and determines them as past task features of past task data. The integrated features may show the current status and existing problems of wood processing technology under specific wood and processing tool conditions and with the current worker's skill level.

[0124] In the embodiment of the present invention, the task feature corresponding to the same past time node is merged with the task category feature execution feature to obtain the past merged features corresponding to each of the multiple past time nodes, which can be implemented through the following examples.

[0125] Determine the data category as an index vector, and determine the feature representation corresponding to the data category as a response vector, and generate a plurality of index response vectors corresponding to the task feature;

[0126] The task features and task category features corresponding to the same past time node are merged through a feature reconstruction model to obtain the past merged features, wherein the feature reconstruction model includes a plurality of cascaded feature reconstruction sub-models, each feature reconstruction sub-model includes a computing unit for determining the cross-source interaction mechanism of different data categories according to a plurality of index response vectors and task category features.

[0127] In the embodiment of the present invention, for example, it is assumed that the server receives a target industrial simulation task module for stamping process improvement of an automobile parts factory. First, the server determines the data category as an index vector, and determines the feature representation corresponding to the data category as a response vector, and generates multiple index response vectors corresponding to the task feature. For example, the data category can be the type of stamping equipment, the material of the mold, the characteristics of the raw material, etc. The type of stamping equipment, such as "hydraulic stamping machine", is determined as an index vector, and its corresponding feature representation, such as "maximum pressure 100 tons, stroke speed 5 times per second", is determined as a response vector, and so on, to generate multiple such index response vectors. Next, the server performs a merge of the task features and task category features corresponding to the same past time node through a feature reconstruction model to obtain a past merged feature. The feature reconstruction model includes multiple cascaded feature reconstruction sub-models. Assuming that at a past time node one month ago, the task features include the characteristics of the stamping equipment used at that time being a certain model of hydraulic press, its stroke speed and pressure; the mold material is high-strength steel, its wear resistance and pressure resistance; the raw material is a certain alloy, its ductility and hardness characteristics, etc. The task category feature is "optimize stamping parameters to improve part accuracy". After the first feature reconstruction sub-model receives these index response vectors and task category features, its calculation unit will determine the cross-source interaction mechanism between the equipment type and accuracy optimization based on the type and performance of the stamping equipment and the accuracy requirements in the task category features, such as calculating the limitations and potential of the current equipment for accuracy improvement under the existing parameters. The second feature reconstruction sub-model will further determine the cross-source interaction mechanism between the mold material and accuracy optimization based on the previous sub-model and the characteristics of the mold material, such as analyzing the impact of the mold's wear resistance on the surface quality of the part, and how to adjust the mold maintenance cycle according to the accuracy requirements. The third feature reconstruction sub-model combines the characteristics of the raw materials and the cross-source interaction mechanism determined earlier to finally derive the past merged features of this past time node. It can include the specific adjustments to the stamping parameters that need to be made to improve the accuracy of the part under the conditions of the equipment, mold and raw materials at the time, as well as the challenges and limitations that may be faced. For another example, suppose it is a target industrial simulation task module for optimizing the circuit board printing process of an electronics factory. The server determines data categories, such as "printing equipment model", "ink type", "circuit board material", etc. as index vectors, and corresponding feature representations, such as "resolution of a certain type of printing equipment", "drying time of a certain ink", "flatness of a certain material circuit board", etc. as response vectors, and generates multiple index response vectors. At the past time point two weeks ago, the task features included the accuracy and speed characteristics of the specific printing equipment used at the time, the viscosity and adhesion characteristics of the ink, the material and circuit layout characteristics of the circuit board, etc., and the task category feature was "improving the clarity of the printed circuit." Each sub-model in the feature reconstruction model works in sequence.The first sub-model determines the cross-source interaction mechanism between device performance and line clarity based on the characteristics of the printing equipment and the task category characteristics, such as determining whether the device resolution meets the clarity requirements. The second sub-model further analyzes based on the characteristics of the ink, such as the impact of ink viscosity on the uniformity of line thickness. The third sub-model comprehensively considers factors such as the material of the circuit board, and ultimately obtains the past merged features of this past time node. For example, in order to improve line clarity, specific adjustment suggestions are required for printing equipment parameters, ink selection, and circuit board preprocessing. Through the above detailed scenario examples, the server can use the feature reconstruction model to effectively merge the task features and task category features of the same past time node, providing valuable past merged features for subsequent analysis and optimization.

[0128] In the embodiment of the present invention, the merging of the task features and the task category features corresponding to the same past time node by the feature reconstruction model to obtain the past merged features can be implemented through the following examples.

[0129] In each feature reconstruction sub-model, the data categories included in the task feature are polled, and a corresponding cross-source interaction mechanism is determined for the currently polled data category;

[0130] Optimizing the task category feature according to the determined cross-source interaction mechanism, and determining the cross-source interaction mechanism corresponding to the next round of data category polling through the optimized task category feature;

[0131] The task category is optimized according to the cross-source interaction mechanism corresponding to the last data category polled, and the optimized task category is sent to the secondary feature reconstruction sub-model, so that the secondary feature reconstruction sub-model executes the corresponding cross-source interaction mechanism according to the input task category feature, and determines the output result of the last feature reconstruction sub-model as the past merged feature.

[0132] In the embodiment of the present invention, for example, it is assumed that the server receives a target industrial simulation task module for optimizing a pharmaceutical packaging production line of a pharmaceutical factory. First, in each feature reconstruction sub-model of the feature reconstruction model, the data categories contained in the task features are polled. For example, the data categories contained in the task features include the model of the packaging equipment, the type of packaging materials, the production speed, etc. The server first polls the data category of the model of the packaging equipment, and determines the corresponding cross-source interaction mechanism for this data category. Assume that the current packaging equipment model is "ModelX", which is characterized by being able to package 100 medicines per minute, but is prone to the problem of loose sealing when running at high speed. The cross-source interaction mechanism determined by the server may be that this equipment model needs to pay special attention to the parameter adjustment of the sealing process while improving the production speed to ensure the packaging quality. The task category feature is optimized according to the determined cross-source interaction mechanism. For example, the original task category feature is "improving packaging efficiency", which becomes "improving packaging efficiency while ensuring packaging quality" after optimization. Then, the cross-source interaction mechanism corresponding to the second round of data category polling is determined by the optimized task category feature. Next, the server polls the data category of the type of packaging material. Assume that the packaging material currently used is a certain plastic film, which has a certain degree of flexibility but slightly higher air permeability. The determined cross-source interaction mechanism can be that when using this packaging material, it is necessary to pay attention to adjusting the sealing method and storage conditions of the packaging to prevent the medicine from getting damp. The task category features are further optimized according to the newly determined cross-source interaction mechanism, for example, from "improving packaging efficiency while ensuring packaging quality" to "improving packaging efficiency while ensuring packaging quality and drug storage requirements". Continue polling, when the server polls the data category of production speed. Assume that the current production speed is 80 drugs per minute, and there is still room for improvement. The determined cross-source interaction mechanism can be that the production speed should be improved by comprehensively considering factors such as equipment performance, packaging materials, and personnel operation proficiency. According to this cross-source interaction mechanism, the task category features are optimized again, for example, optimized to "comprehensively consider all factors, and increase the production speed to 120 per minute while ensuring packaging quality and meeting drug storage requirements". After optimizing the task category according to the cross-source interaction mechanism corresponding to the last data category (production speed) polled, the optimized task category is sent to the secondary feature reconstruction sub-model. The secondary feature reconstruction sub-model performs the corresponding cross-source interaction mechanism according to the input optimized task category features. For example, the secondary feature reconstruction sub-model will further analyze how to achieve the improvement of production speed, the guarantee of packaging quality and the satisfaction of drug storage requirements in specific operations, and finally determine the output result of the last feature reconstruction sub-model as the past merged feature. This past merged feature can include detailed information such as specific equipment parameter adjustment suggestions, packaging material selection criteria, personnel training points, and expected production speed and packaging quality indicators.

[0133] In an embodiment of the present invention, there are multiple task category features corresponding to the same past time node; the merging of the task features corresponding to the same past time node and the task category features through the feature reconstruction model to obtain the past merged features can be implemented through the following examples.

[0134] Loading a plurality of merged features generated by the feature reconstruction model into the feature dimension reduction model respectively, wherein the plurality of merged features are obtained by merging the task feature with each task category feature execution feature respectively;

[0135] In the feature dimensionality reduction model, dimensionality reduction processing is performed on the input multiple merged features to obtain the past merged features.

[0136] In the embodiment of the present invention, for example, it is assumed that the server receives a target industrial simulation task module about the optimization of the production process of a garment factory. At the same past time node, there are multiple corresponding task category features. For example, the task category features include "improving production efficiency", "reducing defective rate", "optimizing fabric utilization", etc. The server performs the merging of the task features corresponding to the same past time node and these task category features through the feature reconstruction model. It is assumed that the task features include the operating status of the production equipment, the skill level of the workers, the type and quality of the fabric, etc. First, the server loads the multiple merged features generated by the feature reconstruction model into the feature dimension reduction model respectively. For example, the features generated by merging the task features with the task category feature of "improving production efficiency" include the proportion of full-load operation time of the equipment, the output per unit time of the workers, etc.; the features generated by merging the task features with the task category feature of "reducing defective rate" cover the impact of the equipment's precision on product quality, the error rate of workers, etc.; the features generated by merging the task features with the task category feature of "optimizing fabric utilization" include the rationality of the fabric cutting plan, the reuse of the remaining fabric, etc. Then, in the feature dimensionality reduction model, dimensionality reduction processing is performed on these multiple merged features input. The server uses dimensionality reduction algorithms such as principal component analysis (PCA) or linear discriminant analysis (LDA) to process these complex high-dimensional features. Taking principal component analysis as an example, the server calculates the covariance matrix of these features, finds the direction with the largest variance (i.e., the principal component), and projects the high-dimensional features onto these principal components to achieve dimensionality reduction. After dimensionality reduction processing, the originally complex multiple merged features are compressed into a simpler past merged feature that still contains key information. This past merged feature may comprehensively reflect the comprehensive performance and key issues of the garment factory in terms of production efficiency, defective rate, and fabric utilization at that past time node. For example, the past merged feature after dimensionality reduction may indicate that the operating status of the production equipment is the key factor affecting production efficiency and defective rate, while the type and quality of fabric have a greater impact on fabric utilization. For another example, suppose that the server processes the target industrial simulation task module of assembly line optimization of an electronic equipment factory. The task category features at the same past time node include "shortening assembly time", "improving product stability", "reducing costs", etc. The server merges the task features (such as the timeliness of parts supply, the assembly speed of workers, the performance of assembly tools, etc.) with these task category features to generate multiple merged features. These merged features are then loaded into the feature dimensionality reduction model. For example, the features merged with "shortening assembly time" include the average operation time of each workstation, the waiting time in the assembly process, etc.; the features merged with "improving product stability" include the compatibility test results of parts, the error rate in the assembly process, etc.; the features merged with "reducing costs" cover the purchase price of parts, the cost increase caused by scrap rate, etc.In the feature dimensionality reduction model, these features are processed by a dimensionality reduction algorithm. The final past merged features may indicate that the timeliness of parts supply has an important impact on assembly time and cost, and the assembly speed of workers is one of the key factors to improve product stability.

[0137] In the embodiment of the present invention, the task feature includes feature representations of multiple data categories; the task feature is combined with the time series distribution feature to obtain a combined feature, which can be implemented through the following example.

[0138] Determine the data category as an index, and determine the feature representation corresponding to the data category as a response, and generate a plurality of index response arrays corresponding to the task feature;

[0139] The merging of the time series distribution feature and the task feature is performed through a feature reconstruction model to obtain the merged feature, wherein the feature reconstruction model includes a plurality of cascaded feature reconstruction sub-models, each feature reconstruction sub-model includes a computing unit for determining the cross-source interaction mechanism of different data categories according to a plurality of index response vectors and the time series distribution feature.

[0140] In the embodiment of the present invention, for example, it is assumed that the server receives a target industrial simulation task module for CNC machining process optimization of a mechanical processing plant. The task feature includes feature representations of multiple data categories, such as material properties of processed parts (hardness, toughness, etc.), parameters of CNC equipment (precision, speed, etc.), type and wear of tools, etc. First, the server determines the data category as an index, and determines the feature representation corresponding to the data category as a response, and generates multiple index response arrays corresponding to the task feature. For example, "part material properties" is determined as index 1, and its corresponding feature representations such as "hardness is X, toughness is Y" are determined as responses; "CNC equipment parameters" are determined as index 2, and the corresponding "precision is Z, speed is W" are determined as responses, and so on. Next, the time series distribution feature and the task feature are merged through the feature reconstruction model to obtain the merged feature. The feature reconstruction model includes multiple cascaded feature reconstruction sub-models. Assuming that in the past period of time, the production of the mechanical processing plant has changed at different stages, and the time series distribution feature shows that in some time periods, the hardness of the part material fluctuates due to raw material supply problems; in other time periods, the CNC equipment has been upgraded, and the accuracy and speed have been improved. After receiving these index response arrays and time series distribution features, the calculation unit of the first feature reconstruction sub-model will determine the cross-source interaction mechanism between the change of material hardness and the processing technology according to the index response vector of "part material characteristics" and the fluctuation of material hardness in the time series distribution features, such as the need to reduce the tool feed speed when the material hardness is high. The second feature reconstruction sub-model will combine the index response vector of "CNC equipment parameters" and the equipment upgrade in the time series distribution features to determine the cross-source interaction mechanism between equipment upgrade and processing efficiency, such as the appropriate reduction of processing allowance after the equipment accuracy is improved. The third feature reconstruction sub-model comprehensively considers the index response vector and time series distribution features of other data categories and finally obtains the merged feature. This merged feature can include the optimal CNC machining process parameter settings, tool selection strategies, and quality control points for different time periods and different conditions. For another example, suppose that the server processes a target industrial simulation task module for the optimization of the chemical reaction process of a chemical plant. Task features include the concentration and purity of reactants, the temperature and pressure control of the reactor, the type and amount of catalysts, and other feature representations of multiple data categories. The server determines "concentration and purity of reactants" as index 1, and the corresponding specific concentration and purity values ​​as responses; determines "reactor temperature and pressure" as index 2, and the corresponding temperature and pressure setting values ​​as responses, and so on to generate multiple index response arrays. In past production, the time series distribution characteristics reflected the impact of ambient temperature in different seasons on the heat dissipation of the reactor, and the changes in reactant purity caused by fluctuations in the raw material market.The first feature reconstruction sub-model determines the cross-source interaction mechanism between the change in reactant purity and the reaction conversion rate based on the index response vector of "reactant concentration and purity" and the change in purity in the time-series distribution characteristics. For example, when the purity decreases, the reaction time needs to be appropriately extended. The second feature reconstruction sub-model determines the cross-source interaction mechanism between the ambient temperature and the reactor temperature control strategy based on the index response vector of "reactor temperature and pressure" and the influence of the ambient temperature in the time-series distribution characteristics. For example, the cooling measures for the reactor need to be strengthened in summer. After being processed by multiple feature reconstruction sub-models, the final merged features may cover the optimal combination of reaction conditions under different conditions, catalyst adjustment strategies, and safety production precautions.

[0141] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned industrial simulation optimization method based on artificial intelligence. Figure 2 As shown, Figure 2 The block diagram of the computer device 100 provided in the embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112 and a communication unit 113. To achieve data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are directly or indirectly electrically connected to each other. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0142] For illustrative purposes, the foregoing description is made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise form disclosed. Numerous modifications and variations are possible in accordance with the above teachings. These embodiments are selected and described in order to best illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can best utilize the present disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. An industrial simulation optimization method based on artificial intelligence, characterized in that: include: Acquire task data related to the target industrial simulation task module, and determine the target task category corresponding to the task data based on a preset task feature coding model set; According to the target task category, determining a current industrial simulation request matching the target task category from a preset industrial simulation request template library; On the basis that the current industrial simulation request passes the reliability evaluation, providing the current industrial simulation request to the simulation interface where the target industrial simulation task module is located; In response to a user calling the target industrial simulation task module and a simulation instruction of a current industrial simulation request through the simulation interface, executing an industrial simulation request for the task data; The step of acquiring task data related to the target industrial simulation task module and determining the target task category corresponding to the task data based on a preset task feature coding model set includes: Acquire task data related to the target industrial simulation task module, and extract task features of the task data; the task features include feature representations of multiple data categories; Determine the data categories included in the past task data; Selecting at least one feature extraction model corresponding to the data category from the preset task feature encoding model set, and loading the past task data into the at least one feature extraction model to obtain a feature representation of the corresponding data category generated by each feature extraction model; Integrating the feature representations generated by each feature extraction model to obtain task features, determining the past task features of the past task data, and extracting task category features of each past task category; Determine the data category as an index vector, and determine the feature representation corresponding to the data category as a response vector, and generate a plurality of index response vectors corresponding to the task feature; In each feature reconstruction sub-model, the data categories included in the task feature are polled, and a corresponding cross-source interaction mechanism is determined for the currently polled data category; Optimizing the task category feature according to the determined cross-source interaction mechanism, and determining the cross-source interaction mechanism corresponding to the next round of data category polling through the optimized task category feature; Optimize the task category according to the cross-source interaction mechanism corresponding to the last data category polled, and send the optimized task category to the secondary feature reconstruction sub-model, so that the secondary feature reconstruction sub-model performs the corresponding cross-source interaction mechanism determination according to the input task category feature, and determines the output result of the last feature reconstruction sub-model as the past merged feature, wherein the feature reconstruction model includes a plurality of cascaded feature reconstruction sub-models, and each feature reconstruction sub-model includes a calculation unit for determining the cross-source interaction mechanism of different data categories according to a plurality of index response vectors and task category features; According to the time series relationship, a sequence arrangement operation is performed on the past merged features corresponding to multiple past time nodes, and a feature encoding operation is performed on the feature array obtained by the sequence arrangement operation to obtain a time series distribution feature; Determine the data category as an index, and determine the feature representation corresponding to the data category as a response, and generate a plurality of index response arrays corresponding to the task feature; The time series distribution feature and the task feature are merged by a feature reconstruction model to obtain the merged feature, wherein the feature reconstruction model includes a plurality of cascaded feature reconstruction sub-models, each of which includes a computing unit for determining a cross-source interaction mechanism of different data categories according to a plurality of index response vectors and the time series distribution feature; The combined feature is converted to a feature domain where the pending task category feature is located, and a feature association operation between the combined feature and the task category feature is performed on the feature domain to obtain an association confidence between the combined feature and the pending task category feature; A target task category feature is determined according to the association confidence, and a task category corresponding to the target task category feature is determined as the target task category corresponding to the task data.

2. The method according to claim 1, characterized in that The reliability of the current industrial simulation request is evaluated by: Obtaining request content of a current industrial simulation request to be evaluated, wherein the current industrial simulation request is obtained based on an industrial control parameter operation item; According to the request content of the current industrial simulation request, the execution status of the current industrial simulation request is evaluated to obtain a first evaluation result; wherein the execution status is used to characterize the reliability of the industrial simulation architecture in executing the industrial control parameter operation item based on the current industrial simulation request; The industrial simulation architecture is called to execute the industrial control parameter operation item on the industrial control parameter instance based on the current industrial simulation request to obtain a real simulation operation result; the number of the industrial control parameter instances is multiple, and the industrial simulation architecture executes the industrial control parameter operation item on each industrial control parameter instance based on the current industrial simulation request to obtain a real simulation operation result corresponding to each industrial control parameter instance; Obtaining standard simulation operation results corresponding to each industrial control parameter instance; According to the standard simulation operation results and the corresponding real simulation operation results corresponding to the industrial control parameter instances, the evaluation scores of the industrial control parameter instances on the industrial simulation architecture are evaluated; Integrate the evaluation scores of multiple industrial control parameter instances on the industrial simulation architecture to obtain the evaluation score of the current industrial simulation request on the industrial simulation architecture; Based on the evaluation score of the current industrial simulation request on the industrial simulation architecture, the reliability of the simulation operation of the current industrial simulation request on the industrial simulation architecture is evaluated to obtain a second evaluation result; the number of the industrial simulation architectures is multiple, and one industrial simulation architecture corresponds to one second evaluation result; the multiple industrial simulation architectures include multiple simulation architecture models of multiple simulation architectures, and one simulation architecture model is one industrial simulation architecture; Polling each simulation framework of the plurality of simulation frameworks to determine a target simulation framework currently being polled; Performing an averaging operation on the reliability of the simulation operations in the second evaluation results corresponding to each simulation architecture model of the target simulation architecture to obtain an average reliability score; Based on the average reliability score, performing a score evaluation on the simulation request completion degree of the current industrial simulation request on the target simulation architecture to obtain an evaluation score of the current industrial simulation request on the target simulation architecture; After each simulation framework in the plurality of simulation frameworks is polled, the evaluation scores of the current industrial simulation request on each simulation framework are integrated to obtain a completion score of the simulation request; The first evaluation result and the simulation request completion score are used to generate a reliability score for the current industrial simulation request.

3. The method according to claim 2, characterized in that The request content of the current industrial simulation request includes: a first content description feature obtained by performing a feature extraction operation on the current industrial simulation request; the execution status of the current industrial simulation request includes the description accuracy of the current industrial simulation request; The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result includes: Acquire content text data of the industrial control parameter operation item, and perform a feature extraction operation on the content text data to obtain a second content description feature; Determining feature similarity between the current industrial simulation request and the content text data based on the first content description feature and the second content description feature; The description accuracy of the current industrial simulation request is evaluated according to the determined feature similarity, and the evaluated description accuracy is added to the first evaluation result.

4. The method according to claim 2, characterized in that: The request content of the current industrial simulation request includes: a first content description feature obtained by performing a feature extraction operation on the current industrial simulation request; the execution status of the current industrial simulation request includes the description accuracy of the current industrial simulation request; The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result includes: Obtaining a valid industrial simulation request under a description reliability attribute, wherein the valid industrial simulation request is an industrial simulation request obtained based on the industrial control parameter operation item and meeting corresponding attribute requirements; Performing a feature extraction operation on the valid industrial simulation request to obtain a third content description feature; Determine, based on the first content description feature and the third content description feature, a request overlap degree between the current industrial simulation request and the valid industrial simulation request; Polling each sub-attribute in the attribute describing reliability to determine a target sub-attribute currently being polled; According to the request overlap between the current industrial simulation request and the valid industrial simulation request under the target sub-attribute, performing score evaluation on the current industrial simulation request under the target sub-attribute to obtain an evaluation score of the current industrial simulation request under the target sub-attribute; After each sub-attribute in the description reliability attribute is polled, the lowest evaluation score is selected from the evaluation scores of the current industrial simulation request under each sub-attribute as the description accuracy of the current industrial simulation request, and the evaluated description accuracy is added to the first evaluation result; The description reliability attribute includes sub-attributes: spelling correctness sub-attribute, description standardization sub-attribute, temporal accuracy sub-attribute and description habit adaptability sub-attribute; Among them, the attribute requirements of the spelling correctness sub-attribute include: there are no spelling errors in the industrial simulation request; the attribute requirements of the description standardization sub-attribute include: there is no non-standard description in the industrial simulation request; the attribute requirements of the timing accuracy sub-attribute include: there is no inaccurate timing in the industrial simulation request; the attribute requirements of the description convention adaptability sub-attribute include: the expression of the industrial simulation request satisfies the description convention corresponding to the corresponding industrial simulation request; a valid industrial simulation request is configured under each sub-attribute in the description reliability attribute; the valid industrial simulation request under each sub-attribute is subjected to feature extraction operations respectively to obtain the third content description features corresponding to each sub-attribute; and the third content description features corresponding to each sub-attribute are respectively used to determine the request overlap between the current industrial simulation request and the valid industrial simulation request under the corresponding sub-attribute.

5. The method according to claim 2, characterized in that: The request content of the current industrial simulation request includes data parameters of the current industrial simulation request under multiple preset indicators; the execution status of the current industrial simulation request includes the request logic depth of the current industrial simulation request; The step of evaluating the execution status of the current industrial simulation request according to the request content of the current industrial simulation request to obtain a first evaluation result includes: Obtaining a weighting factor for each of the plurality of preset indicators; Using the weighting factors of the preset indicators, linearly superimposing the data parameters of the current industrial simulation request under the multiple preset indicators to obtain the accumulated value of the data parameters; Performing a standardization operation on the accumulated value of the data parameter to obtain a standardized value; Using the standardized value as the request logic depth of the current industrial simulation request, and adding the evaluated request logic depth to the first evaluation result; The multiple preset indicators include: a data volume indicator, an object richness indicator, a component level number indicator and a domain-specific vocabulary indicator; wherein: the data parameter of the current industrial simulation request under the data volume indicator is the total amount of request data of the current industrial simulation request, and the total amount of request data is the number of text elements in the current industrial simulation request; the data parameter of the current industrial simulation request under the object richness indicator is the object richness in the current industrial simulation request, and the object is an attribute used to divide different operations and different operating systems; the data parameter of the current industrial simulation request under the component level number indicator is the number of component levels in the current industrial simulation request, and the component level is the data obtained by dividing the current industrial simulation request into request levels; the data parameter of the current industrial simulation request under the domain-specific vocabulary indicator is the domain-specific vocabulary in the current industrial simulation request.

6. The method according to claim 2, characterized in that The step of performing score evaluation on the simulation request completion degree of the current industrial simulation request on the target simulation architecture based on the average reliability score to obtain the evaluation score of the current industrial simulation request on the target simulation architecture includes: Determine the deviation value between the simulation operation reliability in the second evaluation result corresponding to each simulation architecture model of the target simulation architecture and the reliability average score respectively, and obtain multiple deviation values; Based on the multiple deviation values, determine the reliability confidence of the current industrial simulation request on the target simulation architecture; wherein the reliability confidence is used to characterize the reliability between the simulation job reliability of the current industrial simulation request on each simulation architecture model of the corresponding simulation architecture; Based on the reliability confidence of the current industrial simulation request on the target simulation architecture, a score evaluation is performed on the simulation request completion degree of the current industrial simulation request on the target simulation architecture to obtain an evaluation score of the current industrial simulation request on the target simulation architecture.

7. The method according to claim 2, characterized in that The first evaluation result includes parameter values: description accuracy, description accuracy, and request logic depth; The step of using the first evaluation result and the simulation request completion score to generate a reliability score for the current industrial simulation request includes: Obtaining the importance factor of each parameter value in the first evaluation result and the importance factor of the simulation request completion score; Determine a comprehensive score of the current industrial simulation request based on each acquired importance factor, each parameter value in the first evaluation result, and the completion score of the simulation request; The comprehensive score of the current industrial simulation request is used to construct a reliable score of the current industrial simulation request.

8. The method according to claim 2, characterized in that: The process of evaluating the execution status of the current industrial simulation request includes: determining the feature similarity between the content text data of the current industrial simulation request and the industrial control parameter operation item; The method further comprises: Based on the numerical value of the parameter value of the feature similarity, evaluating the key point consistency between the description key points of the current industrial simulation request and the description key points of the content text data; If the key point consistency is evaluated, the reliability evaluation step of the current industrial simulation request is executed; If the key point consistency is not evaluated, the current industrial simulation request is determined to be an illegal industrial simulation request, and the reliability evaluation step of the current industrial simulation request is terminated.

9. A server system, characterized in that: The method comprises a server, wherein the server is used to execute the method described in any one of claims 1 to 8.

Images