A method for generating production model for process optimization based on the fusion of data flow and control flow

By integrating data flow and control flow in industrial production systems, training production models and dynamically adjusting control information, the problem of inefficient production in traditional systems is solved, and more efficient resource utilization and fault reduction are achieved.

CN118734238BActive Publication Date: 2024-11-08GUANGDONG HENGDIAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411231965.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-11-08
Estimated Expiration
2044-09-04

AI Technical Summary

Technical Problem

In traditional industrial production systems, data flow and control flow are relatively independent, resulting in low production efficiency and resource utilization. The existing technology lacks monitoring data coordination in real-time state adjustment, and the failure rate is high.

Method used

By reading fault data and related data flows and control flows from storage units, training production models, periodically obtaining and preprocessing data flows and control flows, dynamically adjusting production equipment control information, and correcting feedback logic when prediction results are abnormal , establish the corresponding relationship between control information and data flow, and realize the integrated optimization production model of data flow and control flow.

Benefits of technology

Improve production efficiency and reduce the incidence of failures. Through real-time prediction and control flow information cycle feedback, the stability and resource utilization of the production process are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118734238B_ABST
    Figure CN118734238B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of data processing technology, and more specifically to a method for generating a process optimization production model based on the fusion of data flow and control flow, comprising: reading multiple fault data and data flows in a preset time period before the corresponding moment of the fault data, and obtaining M types of data flows and control flows related to the fault data based on N types of data flows, and training the production model; obtaining the acquisition data flow, and also obtaining the correlation of different types of data flows, correcting the acquisition data flow and control flow based on the correlation, optimizing the production model, and adjusting the control information of the production equipment according to the first prediction result, and also correcting the feedback logic of the control flow, obtaining the second prediction result based on the new target data flow and control flow within the set time, and adjusting the control information again; establishing the corresponding relationship between the control information and the target data flow and the target control flow, and adjusting the control information when the jth prediction result is abnormal. The present invention can effectively improve production efficiency and reduce the occurrence of faults.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically to a method for generating a process optimization production model based on the fusion of data flow and control flow. Background Art

[0002] In traditional industrial production systems, data flow and control flow are relatively independent, and data collection, transmission and analysis are usually completed separately from production process control, resulting in low production efficiency and resource utilization. With the development of Industry 4.0, flexible production has become an important trend, requiring production systems to quickly respond to market changes and demand fluctuations, while optimizing process flows to improve quality and efficiency.

[0003] Similar prior art includes a Chinese patent application with publication number CN117590821A, which discloses a production line optimization method based on an industrial large model. First, the target production line data is determined through the production line data selection rules; then the production line detection results are determined using the industrial large model and the target production line data; based on the production line detection results, the target production line equipment is further determined, thereby determining the optimization strategy using the production line data of the target production line equipment and the production line equipment associated with it. On the one hand, the production line detection results are first determined using the industrial large model and the initially screened production line data, and based on the production line detection results, the production line equipment that may have problems can be more accurately mined; on the other hand, by expanding the production line equipment with problems, the optimization range of the production line can be expanded to achieve more comprehensive production line optimization. However, the optimization of the production line is only based on the detection results of the production line, without considering the fusion with the control information, and the optimization efficiency of the production line is low.

[0004] Similar prior art also includes a Chinese patent application with publication number CN114510874A, which discloses a production scheduling and machine maintenance method based on a joint optimization model of production and maintenance. Considering the mutual relationship between production scheduling and machine maintenance and the random degradation of machines, a joint optimization model based on production and maintenance is established to perform production engineering scheduling and generate machine maintenance optimization; including: constructing a degradation model of production machines; establishing a joint optimization model of production and maintenance for the production process that considers the correlation between production and maintenance; designing an adaptive machine maintenance strategy AJMW, and on this basis designing a joint optimization method based on a metaheuristic algorithm and an adaptive machine maintenance strategy AJMW, which is used to solve the joint optimization model of production and maintenance, and realize engineering scheduling and maintenance optimization of a mixed production system. Although the invention performs adaptive maintenance and adjustment on the machine according to the real-time status, it lacks the coordination of monitoring data, and the improvement in production efficiency is small.

[0005] Therefore, the present invention proposes a method for generating a process optimization production model based on the fusion of data flow and control flow, which adjusts the control information in the production process by analyzing and collecting data flow and control flow, optimizes the production model, improves production efficiency and reduces the occurrence of failures. Summary of the invention

[0006] In traditional industrial production systems, data flow and control flow are relatively independent, and data collection, transmission and analysis are usually completed separately from production process control, resulting in low production efficiency and resource utilization. Even if the existing technology performs adaptive maintenance and adjustments on machines according to real-time status, the improvement in production efficiency is very small due to the lack of monitoring data.

[0007] In order to better solve the above problems, the present invention provides a method for generating a process optimization production model based on the fusion of data flow and control flow, which is implemented by executing the following steps:

[0008] Step S1: reading a plurality of fault data and a data stream within a preset time period before a corresponding moment of the fault data from a storage unit, wherein the data stream includes N types of data streams, and obtaining M types of data streams and M types of control streams related to the fault data based on the N types of data streams, and training a production model based on the M types of data streams and the M types of control streams to obtain an optimized production model, wherein N represents a positive integer greater than or equal to 3, and M represents a positive integer greater than or equal to 1 and less than or equal to N;

[0009] Step S2: taking the M types of data streams as acquisition data streams, periodically acquiring the acquisition data streams and control streams corresponding to the acquisition data streams, preprocessing the acquisition data streams and the control streams, acquiring the correlation of different types of data streams in the acquisition data streams, correcting the preprocessed acquisition data streams and the control streams based on the correlations, and acquiring a target data stream and a target control stream corresponding to the target data stream;

[0010] Step S3: input the target data flow and the target control flow into the optimized production model, and obtain a first prediction result, and dynamically adjust the control information of the production equipment according to the first prediction result, and also correct the feedback logic of the target control flow, and obtain a second prediction result based on the new target data flow and the target control flow within a set time after the control information is adjusted, and adjust the control information again according to the second prediction result, and save the control information and the corresponding target data flow and the control flow corresponding to the target data flow to a storage unit;

[0011] Step S4: Establish a correspondence between the control information and the target data flow and the target control flow based on the second prediction result, and when the j-th prediction result is abnormal, adjust the control information based on the correspondence and the target data flow and the target control flow corresponding to the j-th prediction result, wherein j represents a positive integer greater than or equal to 1.

[0012] As a preferred technical solution of the present invention, in step S1, training the production model includes the following steps:

[0013] Step S11: randomly select M types of data flows and control flows corresponding to the M types of data flows from the N types of data flows of the data flow, and obtain training data sets, and using each of the training data sets as first test data to train the production model, to obtain First target production model;

[0014] Step S12: Use the first test data corresponding to the i-th production model to test the i-th production model. When the accuracy of each of the i-th production models is less than the set value, repeat the step S11 and increase the value of M. Otherwise, execute step S13, wherein the i-th production model is any one of the first target production models, and the value range of i is any integer between 1 and the total number of the first target production models;

[0015] Step S13: When there are K first target production models whose accuracy rates are greater than the set value, each input data flow and control flow in the first test data corresponding to each first target production model is randomly crossed with the normal data flow and normal control flow to obtain second test data, input the second test data into the corresponding first target production model, and obtain a second test result, compare the second test result with the first test result corresponding to the first test data, and obtain the second target production model, where K represents a positive integer greater than or equal to 3.

[0016] As a preferred technical solution of the present invention, in step S13, comparing the second test result with the first test result corresponding to the first test data to obtain a second target production model includes:

[0017] When the second test result is consistent with the first test result corresponding to the first test data, the input data flow and the control flow in the second test data that intersect with the normal data flow and the normal control flow are deleted from the corresponding first target production model; when the second test result is inconsistent with the first test result corresponding to the first test data, the input data flow and the control flow in the second test data that intersect with the normal data flow and the normal control flow are retained to obtain the second target production model.

[0018] As a preferred technical solution of the present invention, the step S13 further includes:

[0019] Sort the second target production models according to their accuracy, obtain the t second target production models with the highest accuracy, assign weights to the t second target production models according to the corresponding accuracy, weight the t second target production models, and obtain the final optimized production model, where the value range of t is a positive integer greater than or equal to 2 and less than or equal to 5.

[0020] As a preferred technical solution of the present invention, step S2 includes the following steps:

[0021] Step S21: arranging the collected data stream according to the collection time, obtaining a data group corresponding to each collection time in the collected data stream, and obtaining a processed data stream based on the data group;

[0022] Step S22: Analyze the correlation between each type of data flow in the processed data flow, and correct the values ​​of related data in the processed data flow based on the correlation, change trend and data difference between adjacent data, and use the corrected processed data flow as the target data flow, and the control flow corresponding to the target data flow as the target control flow.

[0023] As a preferred technical solution of the present invention, in step S21, based on the data group, obtaining a processed data stream includes:

[0024] When at least one of the data groups is empty or exceeds the value range, the data group is taken as the first data group, the collected data stream is removed from the first data group to obtain the remaining data stream, the remaining data stream is compared with the historical collected data stream, and a similar data stream with the highest similarity to the remaining data stream is obtained, and the second data group in the similar data stream with the same collection time as the first data group replaces the first data group in the collected data stream, and a processed data stream is obtained.

[0025] As a preferred technical solution of the present invention, in step S22, correcting the value of the relevant data in the processed data stream includes:

[0026] In two types of data streams that are related, at a target position with inconsistent change trends, the first data difference and the second data difference of the two types of data streams with the previous and next data at the target position are calculated respectively, and the second data stream in which the first data difference and the second data difference are both within a set range is used as a reference benchmark to correct the second data stream, the two types of data streams include the first data stream and the second data stream, and the corrected processed data stream is used as the target data stream.

[0027] As a preferred technical solution of the present invention, step S3 further includes the following steps:

[0028] After the control information is adjusted, the new target data flow and the target control flow are obtained at each set time interval, and a second prediction result is obtained based on the new target data flow and the target control flow. When the second prediction result is abnormal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is less than or equal to a preset difference, the adjustment amplitude of the control information is increased. Conversely, when the second prediction result is normal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is greater than the preset difference, the control information is not adjusted.

[0029] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0030] The technical solution of the present invention reads multiple fault data and data streams within a preset time period before the corresponding moment of the fault data from a storage unit, obtains M types of data streams and M types of control streams related to the fault data based on N types of data streams, and trains the production model based on the M types of data streams and M types of control streams. This will greatly reduce the amount of training data, thereby reducing the complexity of calculation and improving the optimization efficiency of the production model.

[0031] The technical solution of the present invention periodically obtains the collected data stream and the control stream corresponding to the collected data stream, pre-processes the collected data stream and the control stream, obtains the correlation of different types of data streams in the collected data stream, corrects the pre-processed collected data stream and control stream based on the correlation, extracts more accurate target data stream and target control stream from multiple types of data streams and control streams, and provides support for further analysis and prediction of production models.

[0032] The technical solution of the present invention optimizes the production model by inputting the target data flow and the target control flow into the production model, and obtains the first prediction result, and adjusts the control information of the production equipment according to the first prediction result, and corrects the feedback logic of the target control flow. After the control information is adjusted, the second prediction result is obtained based on the new target data flow and the target control flow within a set time, and the control information is adjusted again according to the second prediction result, thereby realizing an optimized production model that integrates the real-time prediction of the data flow and the information loop feedback of the control flow in the production process, thereby achieving the purpose of improving production efficiency and reducing the occurrence of failures.

[0033] The technical solution of the present invention can effectively improve the stability and production efficiency of the production process and reduce the occurrence rate of failures by establishing a corresponding relationship between the control information and the target data flow and the target control flow based on the second prediction result, and adjusting the control information based on the corresponding relationship and the target data flow and the target control flow corresponding to the j-th prediction result when the j-th prediction result is abnormal. Through the coordination between the above steps, the production efficiency can be effectively improved and the occurrence of failures can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 The present invention is a flowchart of the steps of the method for generating a process optimization production model based on the fusion of data flow and control flow. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0037] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0038] In traditional industrial production systems, data flow and control flow are relatively independent, and data collection, transmission and analysis are usually completed separately from production process control, resulting in low production efficiency and resource utilization.

[0039] In view of the above technical problems, the present invention proposes Figure 1 The process optimization production model generation method based on the fusion of data flow and control flow is implemented by executing the following steps:

[0040] Step S1: read multiple fault data and data streams within a preset time period before the corresponding moment of the fault data from a storage unit, wherein the data stream includes N types of data streams, and obtain M types of data streams and M types of control streams related to the fault data based on the N types of data streams, and train a production model based on the M types of data streams and M types of control streams to obtain an optimized production model, wherein N represents a positive integer greater than or equal to 3, and M represents a positive integer greater than or equal to 1 and less than or equal to N.

[0041] Specifically, historical fault data and N types of data streams within a preset time period before the fault corresponding to the fault data occurs are read from the storage unit. The data stream consists of N types of data streams, including at least N types of data streams such as ambient temperature data stream, ambient humidity data stream, equipment temperature data stream, equipment current data stream, equipment voltage data stream and equipment power usage data stream in the production process. The control stream corresponds to the data stream one by one, and each type of data stream is controlled by the corresponding type of control stream. Not all of the N types of data streams are related to the fault data. There may be only M types of data streams and control streams related to the fault data. Then, finding these M types of data streams and control streams to optimize the production model will greatly reduce the amount of training data, thereby reducing the complexity of calculation and improving the optimization efficiency of the production model. Finding M types of data streams and control streams will be described in detail below.

[0042] Step S2: Take M types of data streams as acquisition data streams, periodically obtain the acquisition data streams and the control streams corresponding to the acquisition data streams, preprocess the acquisition data streams and the control streams, obtain the correlation of different types of data streams in the acquisition data stream, correct the preprocessed acquisition data streams and control streams based on the correlation, and obtain the target data stream and the target control stream corresponding to the target data stream.

[0043] Specifically, M types of data streams and control streams are periodically collected as new collected data streams and corresponding control streams, and the collected data streams and control streams are cleaned and standardized, and other preprocessing work is performed. Data stream preprocessing belongs to the prior art and will not be repeated here. Since there is often a certain correlation between different types of data streams, for example, in a certain production process, changes in temperature may affect changes in pressure, and changes in pressure may further affect product quality or production efficiency. Therefore, by analyzing the correlation between different types of data streams, the interaction and mutual influence between various types of data streams can be more accurately understood. The analysis of the correlation between different types of data streams will be described in detail below. In the data collection and preprocessing stage, although the collected data streams have been basically preprocessed, these collected data streams may still contain some noise or deviations caused by sensor errors, environmental factors or other unknown factors. Then, by using the known correlation between various types of collected data streams, the preprocessed collected data streams and control streams can be further corrected or adjusted to reduce the impact of noise or deviation, and extract more accurate target data streams and target control streams from multiple types of data streams and control streams, to provide support for further analysis and prediction of production models.

[0044] Step S3: Input the target data flow and the target control flow into the optimized production model, and obtain the first prediction result, and dynamically adjust the control information of the production equipment according to the first prediction result, and correct the feedback logic of the target control flow, and obtain the second prediction result based on the new target data flow and the target control flow within the set time after the control information is adjusted, and adjust the control information again according to the second prediction result, and save the control information and the corresponding target data flow and the control flow corresponding to the target data flow to the storage unit.

[0045] Specifically, traditional production control is often based on fixed parameters or rules, which makes it difficult to adapt to real-time changes in the production process. In the present invention, the target data flow and the target control flow are obtained in real time and input into the optimized production model for prediction, so as to realize the dynamic perception of the production status. The prediction results of the optimized production model are verified by the target data flow and the target control flow, wherein the prediction results can be evaluations of production efficiency, product quality and failure risk. In the case of an abnormal first prediction result, the control information of the production equipment is adjusted, and the feedback logic of the target control flow is also corrected. The first prediction result reflects the future operating state and performance of the system under a given control strategy. By analyzing the first prediction result, possible problems and deficiencies in the current control strategy can be identified, such as insufficient control accuracy, too slow response speed, etc. Based on the analysis of the first prediction result, the correction direction of the feedback logic can be determined. For example, if the prediction result shows that the control accuracy of the system is insufficient, it may be necessary to adjust the comparison threshold or gain coefficient in the feedback logic; if the prediction result shows that the response speed of the system is too slow, it may be necessary to optimize the control algorithm in the feedback logic or increase the bandwidth of the controller. After the control information is adjusted, the new target data flow and target control flow within the set time are re-acquired to verify the optimized production model again. If the second prediction result is also abnormal, the control information and the target control flow and the feedback logic of the target control flow are continuously adjusted until the prediction result reaches the normal production state. The control information and its corresponding target data flow and target control flow are recorded to provide a reference for the prediction of subsequent production models and the control of production equipment. It can also be used to analyze the trends, laws and patterns of the production process, and provide data support for future production optimization and decision-making. Through the above scheme, an optimized production model that integrates the real-time prediction of data flow and the information loop feedback of control flow in the production process is realized, so as to improve production efficiency and reduce the occurrence of failures.

[0046] Step S4: Establish a correspondence between the control information and the target data flow and the target control flow based on the second prediction result, and when the jth prediction result is abnormal, adjust the control information based on the correspondence and the target data flow and the target control flow corresponding to the jth prediction result, where j represents a positive integer greater than or equal to 1.

[0047] Specifically, based on multiple prediction results and corresponding control information, target data flow and target control flow, a corresponding relationship between them is established. When a prediction result (such as the jth prediction result) is abnormal, the problem can be quickly located and the control information can be adjusted based on the previously established corresponding relationship and the target data flow and target control flow corresponding to the prediction result to restore the normal operation of the production process. By continuously collecting new target data flow and target control flow, dynamically adjusting production control information, and combining prediction results with exception handling mechanisms, the stability and production efficiency of the production process can be effectively improved, and the failure rate can be reduced.

[0048] Through the coordination of the above steps, production efficiency can be effectively improved and the occurrence of failures can be reduced.

[0049] Furthermore, in the above step S1, training the production model includes the following steps:

[0050] Step S11: Randomly select M types of data flows and control flows corresponding to the M types of data flows from the N types of data flows, and obtain training data sets, and use each training data set as the first test data to train the production model, and obtain A first target production model.

[0051] Specifically, for example, by randomly selecting 4 types of data flows and a control flow from the 10 types of data flows, we will get 210 training data sets. By using each training data set to independently train the production model, we can get 210 first target production models, which provides a basis for optimizing the production model.

[0052] Step S12: Use the first test data corresponding to the ith production model to test the ith production model. When the accuracy of each ith production model is less than the set value, repeat step S11 and increase the value of M. Otherwise, execute step S13, wherein the ith production model is any production model in the first target production model, and the value range of i is any integer between 1 and the total number of the first target production models.

[0053] Specifically, the first test data corresponding to each i-th production model is used to test the accuracy of the model, wherein the first test data is the training data used to train the i-th production model. If the accuracy of all models is less than the set value, it means that the currently selected M types of data streams may not be sufficient to fully describe the fault characteristics, so it is necessary to increase the value of M. For example, if M was equal to 4 before, this time M can be 5, and step S11 is repeated, that is, the number of data types used to train the production model is increased. If the accuracy of at least one model is higher than the set value, it means that the currently selected M+1 or M+2 types of data streams can describe the fault characteristics, and then proceed to the next step. After analyzing the above steps, K first target production models can be obtained.

[0054] Step S13: When the accuracy of K first target production models is greater than the set value, each input data flow and control flow in the first test data corresponding to each first target production model is randomly crossed with the normal data flow and normal control flow to obtain the second test data, input the second test data into the corresponding first target production model, and obtain the second test result, compare the second test result with the first test result corresponding to the first test data, and obtain the second target production model, where K represents a positive integer greater than or equal to 3.

[0055] Specifically, when the accuracy of K first target production models is greater than a set value, the production model needs to be further optimized to train a more accurate and efficient production model, thereby improving production efficiency and product quality.

[0056] Furthermore, in the above step S13, the second test result is compared with the first test result corresponding to the first test data to obtain the second target production model, including:

[0057] When the second test result is consistent with the first test result corresponding to the first test data, the input data flow and control flow in the second test data that intersect with the normal data flow and the normal control flow are deleted from the corresponding first target production model. When the second test result is inconsistent with the first test result corresponding to the first test data, the input data flow and control flow in the second test data that intersect with the normal data flow and the normal control flow are retained to obtain the second target production model.

[0058] Specifically, the method for further optimizing the production model is to randomly cross each input data stream and control stream in the first test data corresponding to each model with the normal data stream and normal control stream to generate the second test data. Then, the second test data is input into the corresponding first target production model, and the second test result is compared with the first test result. If the two are consistent, it means that the input data stream and control stream have no significant effect on the model prediction result, and it can be deleted from the model; if they are inconsistent, it means that the input data stream and control stream have an effect on the model prediction result, and the input data stream and control stream are retained. The process of further optimizing the production model helps to identify and retain the input data streams that have an important impact on the prediction results of the production model, thereby further optimizing the model.

[0059] Furthermore, the above step S13 also includes:

[0060] Sort the second target production models according to their accuracy, obtain the t second target production models with the highest accuracy, assign weights to the t second target production models according to the corresponding accuracy, weight the t second target production models, and obtain the final optimized production model, where the value range of t is a positive integer greater than or equal to 2 and less than or equal to 5.

[0061] Specifically, in order to further improve the accuracy of the production model, by combining the advantages of multiple models, multiple optimized production models (i.e., second target production models) are further screened and weighted to obtain a final optimized production model. All second target production models are sorted according to their accuracy on the test data, and the t second target production models with the highest accuracy are selected for subsequent processing, where the value range of t is selected based on the consideration of balancing model diversity and computational complexity. According to the accuracy of each selected second target production model, a weight is assigned to it. Generally, the higher the accuracy of the model, the greater the weight assigned. For new input data, each model will give a prediction result, and then these prediction results are weighted averaged according to their respective weights to obtain the final prediction result. Through the above steps, the advantages of multiple production models can be fully utilized to generate a more accurate and efficient optimized production model.

[0062] Furthermore, the above step S2 includes the following steps:

[0063] Step S21: Arrange the collected data stream according to the collection time, obtain the data group corresponding to each collection time in the collected data stream, and obtain the processed data stream based on the data group.

[0064] Step S22: Analyze the correlation between each type of data flow in the processing data flow, and correct the value of the relevant data in the processing data flow based on the correlation, change trend and data difference between adjacent data, and use the corrected processing data flow as the target data flow, and the control flow corresponding to the target data flow as the target control flow.

[0065] Specifically, the collected data streams are sorted according to the collection time to obtain a time-series data stream, and each data group in the time-series data stream is analyzed and processed to improve the accuracy of the data stream, and the processed data stream is obtained. The specific processing method of the data group will be described below. Then the correlation between different types of data in the processed data stream is analyzed to understand the correlation between different data streams, laying the foundation for data stream correction. The correction method of the data stream is detailed below.

[0066] Furthermore, in the above step S21, based on the data group, obtaining the processed data stream includes:

[0067] When at least one of the data groups is empty or exceeds the value range, the data group is taken as the first data group, the collected data stream is removed from the first data group to obtain the remaining data stream, the remaining data stream is compared with the historical collected data stream, and the similar data stream with the highest similarity to the remaining data stream is obtained, and the second data group in the similar data stream with the same collection time as the first data group replaces the first data group in the collected data stream, and the processed data stream is obtained.

[0068] Specifically, the data group at each time point is checked in the sequential data stream. If data is missing (i.e., the data group is empty) or the data exceeds a reasonable range (i.e., exceeds the value range), these data groups are marked as the first data group. The first data group is removed from the collected data stream to obtain the remaining data stream. At this time, the data values ​​of the data groups in the remaining data stream are all within the normal value range. After that, the remaining data stream is compared with the historically collected data stream, and the historical data stream (i.e., the similar data stream) with the highest similarity to the remaining data stream is found in the historically collected data stream. The data group (second data group) corresponding to the collection time of the first data group is found from the similar data stream, and the first data group in the collected data stream is replaced with the second data group, thereby generating a processed data stream to effectively improve the accuracy and reliability of the data stream.

[0069] Furthermore, in the above step S22, the value of the relevant data in the correction processing data stream includes:

[0070] In two types of data streams that are related, at a target position with inconsistent change trends, the first data difference and the second data difference of the two types of data streams at the target position with the previous and next data are calculated respectively, and the second data stream whose first data difference and second data difference are both within the set range is used as a reference benchmark to correct the second data stream, the two types of data streams include the first data stream and the second data stream, and the corrected processed data stream is used as the target data stream.

[0071] Specifically, the rationality of the data stream is further analyzed based on the correlation between the data streams, the change trend, and the data difference between adjacent data (i.e., the difference between the current data and the previous and next data). When the change trends of two related data streams (such as the first data stream and the second data stream) at a certain target position are inconsistent, the data difference (the first data difference and the second data difference) of the two data streams at that position and the previous and next positions are calculated. If both data differences are within the set reasonable range, the data stream (such as the second data stream) with a more stable or more representative change is usually selected as a reference benchmark, and this reference benchmark is used to correct the data value of the other data stream (such as the first data stream) at the target position to ensure the consistency and accuracy of the data. After the data stream is corrected by the above method, the relevant data in the processed data stream is optimized, and the corrected data stream is finally output as the target data stream.

[0072] Furthermore, the above step S3 also includes the following steps:

[0073] After the control information is adjusted, a new target data flow and target control flow are obtained at set intervals, and a second prediction result is obtained based on the new target data flow and target control flow. When the second prediction result is abnormal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is less than or equal to the preset difference, the adjustment range of the control information is increased. Conversely, when the second prediction result is normal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is greater than the preset difference, the control information is not adjusted.

[0074] Specifically, the target data stream (such as parameters such as temperature, pressure, flow rate, etc. in the production process) and the target control stream collected in real time are input into the trained optimization production model. The input target data stream and target control stream are analyzed by using the optimization production model to predict the first prediction result of certain key indicators (such as product quality, production efficiency, etc.) under the current production state. According to the first prediction result, the control information of the production equipment (such as temperature setting value, pressure setting value, flow control, etc.) is adjusted to optimize the production state. After the control information is adjusted, wait for a period of time, which can be a set period of time, so that the production state is adjusted and stabilized according to the new control information. Then, the target data stream and target control stream are obtained again, and the optimization production model is used to predict based on the new target data stream and target control stream to obtain a second prediction result. If the second prediction result is abnormal, or the difference between the new target data stream and target control stream and the target data stream and target control stream corresponding to the first prediction result is less than or equal to the preset difference, it means that the adjustment of the current control information may not be sufficient to achieve the expected effect. At this time, the adjustment range of the control information should be increased to approach the optimal production state more quickly. If the second prediction result is normal, or the difference between the new target data flow and target control flow and the target data flow and target control flow corresponding to the first prediction result is greater than the preset difference, it means that the adjustment of the current control information has achieved a certain effect and the production status is relatively stable. At this time, the current control information can be kept unchanged, or fine-tuned as needed. The control information and its corresponding target data flow and target control flow after each adjustment are saved to the storage unit to facilitate subsequent data analysis, model optimization, and problem tracing. Through the above method, new target data flows and target control flows are continuously obtained, predicted, and the control information is adjusted according to the prediction results to form a closed-loop monitoring and adjustment mechanism to ensure the continuous optimization and stability of the production process.

[0075] To summarize, the present invention reads multiple fault data and data streams within a preset time period before the corresponding moment of the fault data from a storage unit, obtains M types of data streams and M types of control streams related to the fault data based on N types of data streams, and trains the production model based on the M types of data streams and M types of control streams. This will greatly reduce the amount of training data, thereby reducing the complexity of calculations and improving the optimization efficiency of the production model.

[0076] By periodically acquiring the collected data stream and the control stream corresponding to the collected data stream, preprocessing the collected data stream and the control stream, and obtaining the correlation of different types of data streams in the collected data stream, the preprocessed collected data stream and control stream are corrected based on the correlation, and more accurate target data streams and target control streams are extracted from multiple types of data streams and control streams, providing support for further analysis and prediction of production models.

[0077] By inputting the target data flow and the target control flow into the optimized production model, obtaining the first prediction result, adjusting the control information of the production equipment according to the first prediction result, and correcting the feedback logic of the target control flow, after the control information is adjusted, obtaining the second prediction result based on the new target data flow and the target control flow within a set time, and adjusting the control information again according to the second prediction result, an optimized production model is realized that integrates the real-time prediction of the data flow in the production process with the information loop feedback of the control flow, so as to achieve the purpose of improving production efficiency and reducing the occurrence of failures.

[0078] Establishing a correspondence between the control information and the target data flow and the target control flow based on the second prediction result, and adjusting the control information based on the correspondence and the target data flow and the target control flow corresponding to the j-th prediction result when the j-th prediction result is abnormal, can effectively improve the stability and production efficiency of the production process and reduce the occurrence rate of failures. Through the coordination between the above steps, the production efficiency can be effectively improved and the occurrence of failures can be reduced.

[0079] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0080] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0081] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0082] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0083] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating a process optimization production model based on the fusion of data flow and control flow, characterized in that: The method comprises: Step S1: reading multiple fault data and data streams within a preset time period before the corresponding moment of the fault data from a storage unit, wherein the data streams include N types of data streams, and obtaining M types of data streams and L control streams related to the fault data based on the N types of data streams, and training a production model based on the M types of data streams and the L control streams to obtain an optimized production model, wherein N represents a positive integer greater than or equal to 3, and M and L represent positive integers greater than or equal to 1 and less than or equal to N; Step S2: taking the M types of data streams as acquisition data streams, periodically acquiring the acquisition data streams and control streams corresponding to the acquisition data streams, preprocessing the acquisition data streams and the control streams, acquiring the correlation of different types of data streams in the acquisition data streams, correcting the preprocessed acquisition data streams and the control streams based on the correlations, and acquiring a target data stream and a target control stream corresponding to the target data stream; Step S3: input the target data flow and the target control flow into the optimized production model, and obtain a first prediction result, and dynamically adjust the control information of the production equipment according to the first prediction result, and also correct the feedback logic of the target control flow, and obtain a second prediction result based on the new target data flow and the target control flow within a set time after the control information is adjusted, and adjust the control information again according to the second prediction result, and save the control information and the corresponding target data flow and the control flow corresponding to the target data flow to a storage unit; Step S4: establishing a correspondence between the control information and the target data flow and the target control flow based on the second prediction result, and adjusting the control information based on the correspondence and the target data flow and the target control flow corresponding to the j-th prediction result when the j-th prediction result is abnormal, wherein j represents a positive integer greater than or equal to 1; The step S2 comprises the following steps: Step S21: arranging the collected data stream according to the collection time, obtaining a data group corresponding to each collection time in the collected data stream, and obtaining a processed data stream based on the data group; Step S22: analyzing the correlation between each type of data stream in the processed data stream, and correcting the values ​​of the relevant data in the processed data stream based on the correlation, the change trend and the data difference between adjacent data, and using the corrected processed data stream as the target data stream, and the control stream corresponding to the target data stream as the target control stream; In the step S22, correcting the value of the relevant data in the processed data stream includes: In two types of data streams that are related, at a target position with inconsistent change trends, the first data difference and the second data difference of the two types of data streams with the previous and next data at the target position are calculated respectively, and the second data stream in which the first data difference and the second data difference are both within a set range is used as a reference benchmark to correct the second data stream, the two types of data streams include the first data stream and the second data stream, and the corrected processed data stream is used as the target data stream.

2. The method according to claim 1, characterized in that In step S1, training the production model includes the following steps: Step S11: randomly select M types of data flows and L control flows corresponding to the M types of data flows from the N types of data flows of the data flow, and obtain training data sets, and using each of the training data sets and the L control flows as first test data to train the production model, obtaining First target production model; Step S12: Use the first test data corresponding to the i-th production model to test the i-th production model. When the accuracy of each of the i-th production models is less than the set value, repeat the step S11 and increase the value of M. Otherwise, execute step S13, wherein the i-th production model is any one of the first target production models, and the value range of i is any integer between 1 and the total number of the first target production models; Step S13: When there are K first target production models whose accuracy rates are greater than the set value, each input data flow and control flow in the first test data corresponding to each first target production model is randomly crossed with the normal data flow and normal control flow to obtain second test data, input the second test data into the corresponding first target production model, and obtain a second test result, compare the second test result with the first test result corresponding to the first test data, and obtain the second target production model, where K represents a positive integer greater than or equal to 3.

3. The method according to claim 2, characterized in that In the step S13, comparing the second test result with the first test result corresponding to the first test data to obtain a second target production model includes: When the second test result is consistent with the first test result corresponding to the first test data, the input data flow and the control flow in the second test data that intersect with the normal data flow and the normal control flow are deleted from the corresponding first target production model; when the second test result is inconsistent with the first test result corresponding to the first test data, the input data flow and the control flow in the second test data that intersect with the normal data flow and the normal control flow are retained to obtain the second target production model.

4. The method according to claim 2, characterized in that: The step S13 further includes: Sort the second target production models according to their accuracy, obtain the t second target production models with the highest accuracy, assign weights to the t second target production models according to the corresponding accuracy, weight the t second target production models, and obtain the final optimized production model, where the value range of t is a positive integer greater than or equal to 2 and less than or equal to 5.

5. The method according to claim 1, characterized in that In the step S21, based on the data group, obtaining a processed data stream includes: When at least one of the data groups is empty or exceeds the value range, the data group is taken as the first data group, the collected data stream is removed from the first data group to obtain the remaining data stream, the remaining data stream is compared with the historical collected data stream, and a similar data stream with the highest similarity to the remaining data stream is obtained, and the second data group in the similar data stream with the same collection time as the first data group replaces the first data group in the collected data stream, and a processed data stream is obtained.

6. The method according to claim 4, characterized in that The step S3 also includes the following steps: After the control information is adjusted, the new target data flow and the target control flow are obtained at each set time interval, and a second prediction result is obtained based on the new target data flow and the target control flow. When the second prediction result is abnormal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is less than or equal to a preset difference, the adjustment amplitude of the control information is increased. Conversely, when the second prediction result is normal or the difference between the new target data flow and the target data flow corresponding to the first prediction result is greater than the preset difference, the control information is not adjusted.