Intelligent factory production online monitoring and analysis system based on digital twinning

Abstract

The invention relates to the technical field of factory management, and particularly discloses a digital twinning-based smart factory production online monitoring and analysis system, which comprises a prediction verification module and a processing evaluation module, and is characterized in that the prediction verification module carries out chain prediction on processed data according to data before processing of a processing flow, and the processing evaluation module carries out processing evaluation on the processed data; and then the prediction result is compared with actual data after processing, the comparison result comprises qualification and disqualification, the processing evaluation module constructs an evaluation coefficient according to the chain length and the chain number of chain prediction and the qualified and disqualified number in the verification result, and the production state of the current product is evaluated according to the evaluation coefficient. According to the method, the uncredible simulation result caused by data missing and data noise can be distinguished, meanwhile, it is guaranteed that the normal simulation result is kept unchanged, and the production stability of the smart factory is improved.

Description

Technical Field

[0001] This invention relates to the field of factory management technology, specifically to an online monitoring and analysis system for smart factory production based on digital twins. Background Technology

[0002] Smart factories are one of the development goals of unmanned and intelligent factories. By integrating data within the factory through digital twins, online simulation of the production process can be achieved, enabling more refined management of the factory's production status. This monitoring and management method can promptly detect equipment failures, product quality issues, and other problems. However, the accuracy of the simulation results is highly dependent on the quality of real-time data. Small data issues, such as missing data and noise, can easily cause some simulation results to differ from reality. This is especially true in the factory's quality inspection process, where relying solely on digital factory simulation results can lead to a significant outflow of defective products.

[0003] In view of this, the present invention proposes an online monitoring and analysis system for smart factory production based on digital twins, which improves the quality pass rate of the final product by analyzing the data generated in the digital twin process. Summary of the Invention

[0004] The purpose of this invention is to provide a smart factory production online monitoring and analysis system based on digital twins, solving the following technical problems: How can we improve the quality pass rate of the final product by analyzing the data generated in the digital twin process?

[0005] The objective of this invention can be achieved through the following technical solutions: A smart factory production online monitoring and analysis system based on digital twins includes: The product manufacturing module includes several processing steps, and several processing steps belong to a production chain. For products with multiple production chains, they are analyzed separately. The data acquisition module is used to collect relevant data before and after processing in the factory's processing flow; The digital twin module uses collected data to simulate product status and outputs simulation results. Based on real-time collected multi-source heterogeneous data, including equipment sensor data, process parameters, and environmental monitoring data, the digital twin module constructs a virtual product model and uses a physics engine and machine learning algorithms to achieve dynamic simulation of the product's entire lifecycle. This module simulates the product's state changes at different production stages, including key indicators such as structural deformation, material performance evolution, and assembly tolerance accumulation. It also combines historical data and real-time feedback for self-optimization and iteration, continuously improving simulation accuracy. Simulation results are output in the form of visual charts, 3D dynamic models, and numerical reports, specifically including multi-dimensional product status assessment data such as stress distribution cloud maps, fatigue life prediction curves, and failure probability heat maps. By comparing the simulation results with actual production line data, the system can automatically identify abnormal operating conditions where deviations exceed thresholds. The prediction and verification module performs chain prediction on the data after processing based on the data before processing in a processing flow. Then, it compares the prediction results with the actual data after processing. The comparison results include qualified and unqualified. The prediction and verification module constructs a chain prediction model to perform numerical prediction on the input data before processing, such as raw material parameters, process configuration, and environmental indicators, to generate theoretical prediction values ​​of the corresponding parameters after processing. It then automatically compares these values ​​with the actual output data and outputs a binary judgment result of qualified or unqualified. The processing evaluation module constructs evaluation coefficients based on the chain length and number of chains predicted by the chain and the number of qualified and unqualified products in the verification results, and evaluates the current production status of the product based on the evaluation coefficients. The production status includes stable and abnormal.

[0006] The above technical solution provides a process for prediction and verification using data from the digital twin process. The prediction and verification of this invention is based on data from the digital twin process. If data is missing, the digital twin module will add an average value for simulation in order to stabilize itself. At this time, whether the actual missing data in the chain prediction process is greater than or less than the average value, the deviation will accumulate in the chain prediction process, making it easier to distinguish. Daily data noise will increase the number of unqualified results in the prediction and verification comparison results, which will also be quickly distinguished. Therefore, this invention can distinguish between unreliable simulation results caused by data missing data and data noise, while ensuring that normal simulation results remain unchanged, thereby improving the stability of smart factory production.

[0007] As a further technical solution of the present invention, the process of performing chain prediction includes: A single match: Obtain several product parameters before processing in a process flow, and select at least three sets of parameters with the highest relevance from several product parameters after processing in the same process flow as the first data set; Data prediction: Build a prediction model for each first data set. The input of the prediction model is the parameters of the target data set before processing, and the output is the parameters of the first data set after processing. Constructing a second data set: Select the parameter with the highest correlation between the parameter output by the data prediction and the parameter after the next processing flow is completed, and use it as the second data set. Repeat the data prediction for the second data set to construct multiple second data sets until the correlation between the parameters before and after processing in the second data set is less than 0.4. Constructing a prediction chain: Construct a prediction chain using the data prediction process of the first data set and multiple second data sets. The length of the prediction chain is equal to the number of data predictions.

[0008] As a further technical solution of the present invention: the process of obtaining the relevance includes: Obtain n sets of historical data for parameters before and after processing, and then use the formula: Get relevance Where n is the total number of historical data obtained, and i is a non-zero natural number not greater than n. It is the i-th parameter before processing. It is the average value of the parameters before processing. It is the i-th post-processing parameter. It is the average value of the parameters after processing.

[0009] As a further technical solution of the present invention: the process of comparing the prediction result with the processed actual data includes: Obtain the ratio of the predicted value to the actual value of the corresponding parameter after processing; like If so, the output comparison result is qualified; Otherwise, the output comparison results will be unqualified.

[0010] The above technical solution provides a chain-based prediction process. In constructing the prediction chain, the number of chains serves as the base of the evaluation coefficient, reflecting the correlation strength between various stages in the product production process. A higher number of chains indicates weaker correlation between stages. By quantifying the negative relationship between chain number and correlation, the system can automatically identify differences in production characteristics of different products, avoiding data redundancy and duplicate verification. Furthermore, this mechanism can be extended to dynamic optimization scenarios: when an abnormal increase in the number of prediction chains for a product is detected, a stage coupling warning is triggered, prompting a recalibration of production parameters or adjustment of the supply chain configuration, thereby achieving a functional upgrade from data differentiation to process optimization. The chain-based prediction of this invention uses data monitored by the digital twin process as a benchmark, enabling the accumulation of monitoring data errors for easier verification calculations. In addition, the prediction chain of this invention is constructed based on the correlation between parameters; that is, a higher number of prediction chains indicates weaker correlation between various stages in the product's production process. The number of prediction chains for a given product is fixed, meaning the number of chains serves as the base for obtaining the evaluation coefficient, enabling the differentiation of different products and thus achieving data differentiation during the verification process, avoiding a large amount of duplicate data.

[0011] As a further technical solution of the present invention, the process of constructing the evaluation coefficients includes: Through the formula: Obtain evaluation coefficients ,in , and These are the preset fitting parameters. It represents the number of products that meet the criteria in the chain prediction process. It represents the number of products that fail the comparison results during the chain prediction process. It is the chain length of the j-th predicted chain. It is the number of chains in the chain prediction, and j is not greater than Non-zero natural numbers.

[0012] As a further technical solution of the present invention: the process of obtaining the predicted chain number includes: If the correlation between the parameters before and after processing in the second data set is less than 0.4, then a new matching process will be performed based on the parameters after processing. After a single match, repeat the data prediction, construct the second data set, and construct the prediction chain; The number of times the prediction chain is constructed is the chain number.

[0013] As a further technical solution of the present invention: the process of evaluating the current production status of the product based on the evaluation coefficient includes: Set a critical threshold. If the evaluation coefficient is not less than the critical threshold, the output of the current product's production status is stable. If the evaluation coefficient is greater than the critical threshold, the current production status of the product will be output as abnormal.

[0014] The above technical solution provides a process for constructing evaluation coefficients and evaluating the current production status of the product based on these coefficients. The evaluation coefficients of this invention are obtained by fitting a quadratic polynomial, and the variables used in the fitting are... In the chain prediction process, the decrease in the number of qualified products or the shortening of the prediction chain will lead to a smaller evaluation coefficient. This helps to distinguish whether there is data loss or data noise during the simulation process, and to verify the reliability of the simulation results.

[0015] As a further technical solution of the present invention: the number of processing steps in the product production module is at least four.

[0016] As a further technical solution of the present invention: the process of obtaining the critical threshold includes: Obtain the range of values ​​for the evaluation coefficients, and iterate within that range; Calculate the midpoint of the range of values ; like If the actual state of the product is stable under the given conditions, then update... Otherwise update ; Repeat the iteration until If the value is less than the preset tolerance, the critical threshold is recorded as follows. .

[0017] The beneficial effects of this invention are: (1) The prediction verification of the present invention is based on the data in the digital twin process. If data is missing, the digital twin module will add an average data for simulation in order to stabilize itself. At this time, the actual data missing in the chain prediction process, whether greater than or less than the average, will accumulate deviation in the chain prediction process, which makes it easier to distinguish. Daily data noise will increase the number of unqualified results in the prediction verification comparison results, which will also be quickly distinguished. Therefore, the present invention can distinguish the unreliability of simulation results caused by data missing and data noise, while ensuring that the normal simulation results remain unchanged, thus improving the stability of smart factory production.

[0018] (2) The prediction chain of the present invention is constructed based on the correlation between parameters. That is, the more prediction chains there are, the worse the correlation between each link in the production process of the product. The number of prediction chains for a certain product is fixed. That is, the number of chains is used as the base for obtaining the evaluation coefficient, which can distinguish different products, thereby realizing the differentiation of data in the verification process and avoiding a large amount of duplicate data.

[0019] (3) The evaluation coefficients of this invention are obtained by fitting based on a quadratic polynomial, and the variables used in the fitting are: In the chain prediction process, the decrease in the number of qualified products or the shortening of the prediction chain will lead to a smaller evaluation coefficient. This helps to distinguish whether there is data loss or data noise during the simulation process, and to verify the reliability of the simulation results. Attached Figure Description

[0020] The invention will now be further described with reference to the accompanying drawings.

[0021] Figure 1 This is a schematic diagram of the module composition of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 As shown, in one embodiment, a smart factory production online monitoring and analysis system based on digital twins is provided, including: The product manufacturing module includes several processing steps, and several processing steps belong to a production chain. For products with multiple production chains, they are analyzed separately. The data acquisition module is used to collect relevant data before and after processing in the factory's processing flow. The digital twin module uses collected data to simulate product status and outputs simulation results. Based on real-time collected multi-source heterogeneous data, including equipment sensor data, process parameters, and environmental monitoring data, the digital twin module constructs a virtual product model and uses a physics engine and machine learning algorithms to achieve dynamic simulation of the product's entire lifecycle. This module simulates the product's state changes at different production stages, including key indicators such as structural deformation, material performance evolution, and assembly tolerance accumulation. It also combines historical data and real-time feedback for self-optimization and iteration, continuously improving simulation accuracy. Simulation results are output in the form of visual charts, 3D dynamic models, and numerical reports, specifically including multi-dimensional product status assessment data such as stress distribution cloud maps, fatigue life prediction curves, and failure probability heat maps. By comparing the simulation results with actual production line data, the system can automatically identify abnormal operating conditions where deviations exceed thresholds. The prediction and verification module performs chain predictions on the data before processing based on the data before processing of a processing flow. Then, it compares the prediction results with the actual data after processing. The comparison results include qualified and unqualified. The prediction and verification module constructs a chain prediction model to perform numerical predictions on the input data before processing, such as raw material parameters, process configuration, and environmental indicators, to generate theoretical prediction values ​​of the corresponding parameters after processing. It then automatically compares these predictions with the actual output data and outputs a binary judgment result of qualified or unqualified. The processing evaluation module constructs evaluation coefficients based on the chain length, chain number, and the number of qualified and unqualified products in the verification results, and evaluates the current production status of the product based on the evaluation coefficients. The production status includes stable and abnormal.

[0024] This embodiment provides a prediction and verification process using data from the digital twin process. The prediction and verification of this invention is based on data from the digital twin process. If data is missing, the digital twin module will add an average value for simulation in order to stabilize itself. At this time, whether the actual missing data in the chain prediction process is greater than or less than the average value, the deviation will accumulate in the chain prediction process, making it easier to distinguish. Daily data noise will increase the number of non-conforming results in the prediction and verification comparison results, which will also be quickly distinguished. Therefore, this invention can distinguish between unreliable simulation results caused by data missing data and data noise, while ensuring that normal simulation results remain unchanged, thereby improving the stability of smart factory production.

[0025] The process of performing chain prediction includes: One-time matching: Obtain several product parameters before processing in a process flow, and select at least three sets of parameters with the highest relevance from several product parameters after processing in the same process flow as the first data set. It should be noted that the highest relevance refers to the highest relevance value. If the highest relevance is less than 0.4, then use several product parameters before processing in the next process flow to replace them for one-time matching. Data prediction: Build a prediction model for each first data set. The input of the prediction model is the parameters of the target data set before processing, and the output is the parameters of the first data set after processing. Constructing a second data set: Select the parameter with the highest correlation between the parameter output by the data prediction and the parameter after the next processing flow is completed, and use it as the second data set. Repeat the data prediction for the second data set to construct multiple second data sets until the correlation between the parameters before and after processing in the second data set is less than 0.4. Constructing a prediction chain: Construct a prediction chain using the data prediction process of the first data set and multiple second data sets. The length of the prediction chain is equal to the number of data predictions.

[0026] In one embodiment, the process of obtaining relevance includes: Obtain n sets of historical data for parameters before and after processing, and then use the formula: Get relevance Where n is the total number of historical data obtained, and i is a non-zero natural number not greater than n. It is the i-th parameter before processing. It is the average value of the parameters before processing. It is the i-th post-processing parameter. It is the average value of the parameters after processing.

[0027] The process of comparing the predicted results with the processed actual data includes: Obtain the ratio of the predicted value to the actual value of the corresponding parameter after processing; like If so, the output comparison result is qualified; Otherwise, the output comparison results will be unqualified.

[0028] This embodiment provides a chain prediction process. The chain prediction of the present invention uses the data monitored by the digital twin process as a benchmark, which can realize the accumulation of monitoring data errors to facilitate verification calculation. In addition, the prediction chain of the present invention is constructed based on the correlation between parameters. That is, the more prediction chains there are, the worse the correlation between each link in the production process of the product. The number of prediction chains for a certain product is fixed. That is, the number of chains is used as the base for obtaining the evaluation coefficient, which can distinguish different products, thereby realizing the differentiation of data in the verification process and avoiding a large amount of duplicate data.

[0029] In constructing the prediction chain, the number of chains serves as the base of the evaluation coefficient, reflecting the correlation strength between various stages in the product manufacturing process. A higher number of chains indicates weaker inter-stage correlation. By quantifying the negative relationship between chain number and correlation, the system can automatically identify differences in production characteristics across different products, avoiding data redundancy and redundant verification. Furthermore, this mechanism can be extended to dynamic optimization scenarios: when an abnormal increase in the number of prediction chains for a product is detected, a stage coupling warning is triggered, indicating the need to recalibrate production parameters or adjust the supply chain configuration, thereby achieving a functional upgrade from data differentiation to process optimization.

[0030] In one embodiment, the process of constructing the evaluation coefficients includes: Through the formula: Obtain evaluation coefficients ,in , and These are the preset fitting parameters. It represents the number of products that meet the criteria in the chain prediction process. It represents the number of products that fail the comparison results during the chain prediction process. It is the chain length of the j-th predicted chain. It is the number of chains in the chain prediction, and j is not greater than Non-zero natural numbers.

[0031] In one embodiment, the process of obtaining the predicted chain number includes: If the correlation between the parameters before and after processing in the second data set is less than 0.4, then a new matching process will be performed based on the parameters after processing. After a single match, repeat the data prediction, construct the second data set, and construct the prediction chain; The number of times the prediction chain is constructed is the chain number.

[0032] The process of evaluating the current production status of a product based on evaluation coefficients includes: Set a critical threshold. If the evaluation coefficient is not less than the critical threshold, the output of the current product's production status is stable. If the evaluation coefficient is greater than the critical threshold, the current production status of the product will be output as abnormal.

[0033] This embodiment provides a process for constructing evaluation coefficients and evaluating the current production status of the product based on these coefficients. The evaluation coefficients of this invention are obtained by fitting a quadratic polynomial, and the variables used in the fitting are... In the chain prediction process, the decrease in the number of qualified products or the shortening of the prediction chain will lead to a smaller evaluation coefficient. This helps to distinguish whether there is data loss or data noise during the simulation process, and to verify the reliability of the simulation results.

[0034] In one embodiment, the number of processing steps in the product manufacturing module is at least four.

[0035] It should be noted that if the number of processing steps in the product manufacturing process is insufficient, the problems caused by missing data and data noise cannot be accumulated sufficiently, and the monitoring method provided in this embodiment cannot effectively distinguish them. Therefore, the number of processing steps in the product manufacturing module must be at least four.

[0036] The process of obtaining the critical threshold includes: Obtain the range of values ​​for the evaluation coefficients, and iterate within that range; Calculate the midpoint of the range of values ; like If the actual state of the product is stable under the given conditions, then update... Otherwise update ; Repeat the iteration until If the value is less than the preset tolerance, the critical threshold is recorded as follows. .

[0037] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A smart factory production online monitoring and analysis system based on digital twins, characterized in that, include: The product manufacturing module includes several processing steps. The data acquisition module is used to collect relevant data before and after processing in the factory's processing flow; A digital twin module, which uses the collected relevant data to simulate the product status and outputs the simulation results of the product status; The prediction and verification module performs chain prediction on the data after processing based on the data before processing in a processing flow, and then compares the prediction results with the actual data after processing. The comparison results include qualified and unqualified. The processing evaluation module constructs evaluation coefficients based on the chain length and number of chains predicted by the chain and the number of qualified and unqualified products in the verification results, and evaluates the current production status of the product based on the evaluation coefficients. The production status includes stable and abnormal.

2. The smart factory production online monitoring and analysis system based on digital twins according to claim 1, characterized in that, The process of performing chain prediction includes: A single match: Obtain several product parameters before processing in a process flow, and select at least three sets of parameters with the highest relevance from several product parameters after processing in the same process flow as the first data set; Data prediction: Build a prediction model for each first data set. The input of the prediction model is the parameters of the target data set before processing, and the output is the parameters of the first data set after processing. Constructing a second data set: Select the parameter with the highest correlation between the parameter output by the data prediction and the parameter after the next processing flow is completed, and use it as the second data set. Repeat the data prediction for the second data set to construct multiple second data sets until the correlation between the parameters before and after processing in the second data set is less than 0.

4. Constructing a prediction chain: Construct a prediction chain using the data prediction process of the first data set and multiple second data sets. The length of the prediction chain is equal to the number of data predictions.

3. The smart factory production online monitoring and analysis system based on digital twins according to claim 1, characterized in that, The process of obtaining relevance includes: Obtain n sets of historical data for parameters before and after processing, and then use the formula: Get relevance Where n is the total number of historical data obtained, and i is a non-zero natural number not greater than n. It is the i-th parameter before processing. It is the average value of the parameters before processing. It is the i-th post-processing parameter. It is the average value of the parameters after processing.

4. The smart factory production online monitoring and analysis system based on digital twins according to claim 1, characterized in that, The process of comparing the predicted results with the processed actual data includes: Obtain the ratio of the predicted value to the actual value of the corresponding parameter after processing; like If so, the output comparison result is qualified; Otherwise, the output comparison results will be unqualified.

5. The smart factory production online monitoring and analysis system based on digital twins according to claim 1, characterized in that, The process of constructing evaluation coefficients includes: Through the formula: Obtain evaluation coefficients ,in , and These are the preset fitting parameters. It represents the number of products that meet the criteria in the chain prediction process. It represents the number of products that fail the comparison results during the chain prediction process. It is the chain length of the j-th predicted chain. It is the number of chains in the chain prediction, and j is not greater than Non-zero natural numbers.

6. The smart factory production online monitoring and analysis system based on digital twins according to claim 2, characterized in that, The process of obtaining the predicted chain count includes: If the correlation between the parameters before and after processing in the second data set is less than 0.4, then a new matching process will be performed based on the parameters after processing. After a single match, repeat the data prediction, construct the second data set, and construct the prediction chain; The number of times the prediction chain is constructed is the chain number.

7. The smart factory production online monitoring and analysis system based on digital twins according to claim 5, characterized in that, The process of evaluating the current production status of a product based on evaluation coefficients includes: Set a critical threshold. If the evaluation coefficient is not less than the critical threshold, the output of the current product's production status is stable. If the evaluation coefficient is greater than the critical threshold, the current production status of the product will be output as abnormal.

8. The smart factory production online monitoring and analysis system based on digital twins according to claim 1, characterized in that, The product manufacturing module has at least four processing steps.

9. The online monitoring and analysis system for smart factory production based on digital twins according to claim 7, characterized in that, The process of obtaining the critical threshold includes: Obtain the range of values ​​for the evaluation coefficients, and iterate within that range; Calculate the midpoint of the range of values ; like If the actual state of the product is stable under the given conditions, then update... Otherwise update ; Repeat the iteration until If the value is less than the preset tolerance, the critical threshold is recorded as follows. .

Images