A Method for Correcting Digital Twin Models of Factory Workshops Based on Data Acquisition

By collecting and processing real-time operational data, and utilizing edge computing and genetic algorithms to optimize the digital twin model, the problems of insufficient real-time performance and automation in existing technologies have been solved, enabling real-time correction of the digital twin model and efficient optimization of the production process.

CN123113756APending Publication Date: 2026-11-17SUZHOU MARS VISION DIGITAL TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611408897.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing digital twin model calibration methods lack real-time performance and automation, making it difficult to quickly respond to changes in the production process and effectively adapt to complex and ever-changing production environments.

Method used

By collecting and preprocessing real-time operational data, optimized real-time operational data is generated. This data is then compared in real-time using edge computing nodes to generate error analysis results. The parameters of the digital twin model are automatically adjusted, and incremental learning and optimization are performed using a genetic algorithm to generate precise control commands and achieve adaptive adjustment.

Benefits of technology

It enables real-time correction and automated adjustment of digital twin models, improving the accuracy and adaptability of the models and ensuring real-time optimization and stability of the production process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113756A_ABST
    Figure CN123113756A_ABST
Patent Text Reader

Abstract

This invention discloses a method for correcting a digital twin model of a factory workshop based on data acquisition, belonging to the field of digital twin technology. The method includes: acquiring real-time operational data and preprocessing it to generate optimized real-time operational data; constructing a digital twin model framework based on the optimized real-time operational data, and performing real-time comparison through edge computing nodes to generate error analysis results, automatically adjusting the digital twin model parameters, and generating a corrected digital twin model; using a genetic algorithm to incrementally learn and continuously optimize the corrected digital twin model to generate an optimized digital twin model; and feeding real-time operational data back to production control based on the optimized digital twin model, automatically adjusting production process operating parameters, and generating precise control commands. This invention achieves precise correction of the digital twin model, enabling the digital twin model to better match actual production conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a method for correcting a digital twin model of a factory workshop based on data acquisition. Background Technology

[0002] In modern manufacturing, digital twin technology is widely used for real-time monitoring and optimization of production processes. Digital twin models virtually reproduce physical equipment and dynamic behavior, providing accurate simulation and prediction capabilities for production processes. Conventional methods typically rely on collecting real-time operational data to build or update digital twin models based on this data to optimize production processes. By acquiring various parameters from the production environment and using these parameters to periodically or regularly update the digital twin model, production efficiency and quality can be improved.

[0003] However, existing methods have certain limitations in the calibration process of digital twin models. First, traditional digital twin model update methods are usually performed periodically, lacking real-time capability and making it difficult to quickly respond to changes that occur during production. Second, although existing methods have made progress in data acquisition and model building, their level of intelligence and automation in digital twin model calibration is insufficient. They typically rely on preset rules for adjustment and cannot effectively adapt to the complex and ever-changing environment during production. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a data acquisition-based digital twin model correction method for factory workshops to address the problems of poor real-time performance and insufficient automated adjustment.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for correcting a digital twin model of a factory workshop based on data acquisition. The method includes: acquiring real-time operational data and preprocessing it to generate optimized real-time operational data; constructing a digital twin model framework based on the optimized real-time operational data, performing real-time comparison through edge computing nodes to generate error analysis results, automatically adjusting digital twin model parameters, and generating a corrected digital twin model; using a genetic algorithm to incrementally learn and continuously optimize the corrected digital twin model to generate an optimized digital twin model; feeding real-time operational data back to production control based on the optimized digital twin model, automatically adjusting production process operation parameters, and generating precise control commands; and continuously monitoring real-time operational data during the production process based on the precise control commands, automatically identifying potential changes using machine learning prediction algorithms, triggering adaptive adjustments, and generating a digital twin model correction strategy.

[0007] As a preferred embodiment of the factory workshop digital twin model correction method based on data acquisition described in this invention, the real-time operating data includes temperature, humidity, pressure, operating status, vibration, and flow rate.

[0008] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for generating optimized real-time operating data are as follows: Preliminary noise removal and data smoothing are performed on the real-time running data to generate smoothed real-time running data; The smoothed real-time running data is imputed for missing values ​​and standardized to generate standardized real-time running data. Anomaly detection is performed on the standardized real-time running data to generate optimized real-time running data.

[0009] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for constructing the digital twin model framework are as follows: Feature extraction is performed on the optimized real-time operating data to identify and select features related to the workshop production process environment and operating status. Information from different data sources is integrated through time series data analysis methods to generate a fused feature dataset. By combining the fused feature dataset with the physical characteristics of workshop production, a digital twin model framework is constructed through physical modeling and data-driven modeling methods.

[0010] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for generating error analysis results are as follows: A preliminary simulation of the digital twin model framework was conducted to simulate the production process and operating conditions in the workshop, generating preliminary simulation results. By using edge computing nodes, the preliminary simulation results are compared with the optimized real-time running data in real time, and an error analysis report is generated. Based on the error analysis report, assess the deviations and deficiencies of the digital twin model framework, identify the parts that need optimization, and generate error analysis results.

[0011] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for generating the corrected digital twin model are as follows: Based on the error analysis results, the parameters of the digital twin model architecture are automatically adjusted using the gradient descent method to generate the adjusted digital twin model parameters. The particle swarm optimization method is used to optimize the architecture of the digital twin model, improve the structure and function of the digital twin model, and generate the optimized digital twin model architecture. The adjusted digital twin model parameters and the optimized digital twin model architecture are integrated and optimized to generate a corrected digital twin model.

[0012] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for generating the optimized digital twin model are as follows: Based on the corrected digital twin model, initialize the population of the genetic algorithm, set the crossover rate, mutation rate and selection strategy, and generate the initial population; Based on the initial population, the fitness of each corrected digital twin model parameter is evaluated to generate the optimal population after evaluation. Crossover operations are performed on the selected population after evaluation to generate new corrected digital twin model parameters. The new corrected digital twin model parameters are then randomly adjusted through mutation operations to generate a new generation of corrected digital twin model parameters. The fitness of the new generation of corrected digital twin model parameters is compared with that of the adjusted digital twin model parameters. The digital twin model parameters with the best fitness are selected to generate the optimized digital twin model.

[0013] As a preferred embodiment of the data acquisition-based digital twin model calibration method for factory workshops described in this invention, the specific steps for generating and evaluating the preferred population are as follows: Based on the corrected digital twin model parameters of the initial population, the fitness evaluation criteria are set using the mean squared error method. The fitness performance of each corrected digital twin model parameter in the corrected digital twin model is calculated and quantified, and the fitness value of the corrected digital twin model parameter is generated. Based on the fitness values ​​of the corrected digital twin model parameters, an elite selection method is used to select the corrected digital twin model parameters with the best fitness, and an optimal population is generated after evaluation.

[0014] As a preferred embodiment of the factory workshop digital twin model correction method based on data acquisition described in this invention, the specific steps for generating precise control commands are as follows: Based on the optimized digital twin model, various data during the workshop production process are collected in real time and fed back to the production scheduling center to generate a production data stream; By analyzing the production data flow through the optimized digital twin model, the production process operation parameters that need to be adjusted are calculated and identified, and the production process operation parameters are adjusted in real time based on the analysis results to generate the adjusted production process operation parameters. The adjusted production process operating parameters are verified and optimized to generate precise control commands.

[0015] As a preferred embodiment of the data acquisition-based digital twin model correction method for factory workshops described in this invention, the specific steps for generating the digital twin model correction strategy are as follows: Based on precise control instructions, the system continuously monitors real-time operational data during the production process and analyzes the data using machine learning prediction algorithms to automatically identify potential changes in the production process. Based on potential changes, an adaptive adjustment process is triggered to further optimize the adjusted production process operating parameters and generate a digital twin model correction strategy.

[0016] The beneficial effects of this invention are as follows: by collecting and preprocessing real-time operating data, optimized real-time operating data is generated, which effectively improves data quality and provides high-quality data support for subsequent digital twin model calibration, ensuring that the digital twin model can more accurately reflect the production process; by performing real-time comparison through edge computing nodes, error analysis results are generated and the parameters of the digital twin model are automatically adjusted, realizing the accurate calibration of the digital twin model, thereby enabling the digital twin model to better match the actual production situation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for correcting a digital twin model of a factory workshop based on data acquisition.

[0019] Figure 2 A flowchart for generating real-time running data after optimization.

[0020] Figure 3 A flowchart for building a digital twin model framework.

[0021] Figure 4 The flowchart for generating the optimized digital twin model. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for correcting a digital twin model of a factory workshop based on data acquisition, comprising the following steps: S1. Collect real-time running data, preprocess it, and generate optimized real-time running data.

[0026] S1.1 Real-time operating data includes temperature, humidity, pressure, operating status, vibration, and flow rate.

[0027] Specifically, temperature is collected by temperature sensors installed in the workshop to monitor temperature changes in the production environment or operating area in real time; humidity is collected by humidity sensors installed in the workshop to measure the water vapor content in the air in real time; pressure is collected by pressure sensors installed in liquid, gas, or steam pipelines to monitor the pressure value in the pipeline in real time; operating status is collected by operating status monitoring devices, which record information such as start-up, shutdown, and load during the production process in real time to reflect the working status of the equipment; vibration is collected by vibration sensors installed on various mechanical parts in the workshop to monitor vibration in real time and to detect possible equipment abnormalities or malfunctions; flow rate is collected by flow sensors installed in pipelines or conveying channels to measure the flow rate of fluids or gases in real time to ensure normal material or fluid transportation during the production process; and optimized real-time operating data is generated from the real-time collected temperature, humidity, pressure, operating status, vibration, and flow rate data.

[0028] S1.2 Perform preliminary noise removal and data smoothing on the real-time running data to generate smoothed real-time running data.

[0029] Specifically, high-frequency noise in real-time running data is removed by median filtering or mean filtering. Median filtering eliminates abnormal fluctuations by replacing extreme values ​​with the middle value in the data window. Mean filtering smooths out short-term fluctuations by calculating the average value of all data in the data window. The moving average method is used to average the real-time running data through a fixed-size data window, thereby removing rapid fluctuations and generating smoothed real-time running data.

[0030] It should also be noted that fixed-size data windows are usually set according to the specific application scenario and the required level of data smoothness; common data window sizes are, for example, 5 to 20 data points, with the specific size depending on the fluctuation frequency of the real-time running data and the characteristics of the noise that needs to be removed.

[0031] S1.3. Perform missing value imputation and standardization on the smoothed real-time running data to generate standardized real-time running data.

[0032] Specifically, missing values ​​are identified by checking the integrity of each data point. For missing values, a linear interpolation method is used to fill in the missing parts by extrapolating the values ​​of the preceding and following data points. After filling in the missing values, standardization is performed. By adjusting the distribution of the smoothed real-time running data, the mean of all smoothed real-time running data is, for example, 0, and the standard deviation is, for example, 1, to ensure that different features have the same scale, thus generating standardized real-time running data.

[0033] S1.4 Perform anomaly detection on the standardized real-time running data and generate optimized real-time running data.

[0034] Specifically, the Z-score method is used to detect outliers in the standardized real-time running data that deviate from the normal range. By calculating the Z-score of each data point, it is determined whether the data point exceeds the normal range, and the data point is regarded as an outlier. Outlier handling methods, such as replacing or deleting outliers, are applied to ensure the stability and accuracy of the standardized real-time running data. The contextual relationship of the outlier standardized real-time running data is further checked to ensure that outliers will not affect subsequent analysis. After anomaly detection, optimized real-time running data is generated.

[0035] It should also be noted that the normal range is usually based on the statistical distribution of the data and actual business needs. When using the Z-score method, the normal range usually refers to the data points being within the mean ± 3 standard deviations (i.e., the Z-score is between -3 and +3). Under normal distribution, approximately 99.7% of the data points will fall within the normal range.

[0036] S2. Based on the optimized real-time running data, construct a digital twin model framework, and perform real-time comparison through edge computing nodes to generate error analysis results, automatically adjust the digital twin model parameters, and generate a corrected digital twin model.

[0037] S2.1. Extract features from the optimized real-time operating data, identify and select features related to the workshop production process environment and operating status, and integrate information from different data sources through time series data analysis methods to generate a fused feature dataset.

[0038] Specifically, the optimized real-time operational data identifies environmental and operational characteristics related to the workshop production process, such as temperature, humidity, pressure, vibration, and flow rate. These environmental and operational characteristics reflect relevant states during production; for example, temperature and pressure characterize the stability of the production environment, humidity affects the quality of material processing, and vibration and flow rate reflect the health of the operation. Indicators related to production efficiency, process control, and product quality are selected from these environmental and operational characteristics. For instance, temperature and pressure stability are crucial for ensuring product consistency, while vibration and flow rate reflect the working condition and directly impact production efficiency. Time-series data analysis methods are used to integrate information from different data sources, aligning them chronologically to ensure the consistency and continuity of the time-series data. This captures the temporal variation characteristics of the production process, generating a fused feature dataset.

[0039] S2.2. Combine the fused feature dataset with the physical characteristics of workshop production, and construct a digital twin model framework through physical modeling and data-driven modeling methods.

[0040] Specifically, when combining the fused feature dataset with the physical characteristics of workshop production, key variables related to the workshop production process are identified, such as temperature, humidity, pressure, vibration, and flow rate. These key variables represent different physical conditions in the workshop production environment and form the basis for constructing the digital twin model framework. Physical modeling methods are used to combine these key variables with mathematical expressions reflecting the physical laws of the production process. For example, temperature changes can be described by the heat conduction equation, while changes in flow rate and pressure can be modeled using fluid flow equations. The physical processes within the workshop, including production, operation, and environmental conditions, provide a realistic physical foundation for the digital twin model framework. Simultaneously, a data-driven modeling approach is employed, utilizing optimized and smoothed real-time operational data to capture dynamic changes in the production process. Through decision tree methods, nonlinear or complex relationships in the production process that are not fully captured by the physical model can be identified. Decision trees are trained on historical data and then used to divide and classify features from real-time operational data (such as temperature, humidity, and pressure) to form a series of decision rules. These rules reveal nonlinear relationships and potential interactions between variables, thus capturing complex patterns that are difficult to describe using traditional physical models. Decision trees recursively divide samples based on the features of real-time operational data (such as temperature, humidity, and pressure), selecting the most meaningful features for splitting, generating a tree diagram that reflects the internal structure of the data. This allows the digital twin model framework to more accurately understand and simulate nonlinear and complex behaviors in the production process, further improving prediction accuracy, especially when facing dynamic changes that cannot be fully expressed by physical models. By effectively combining physical modeling and data-driven modeling results, and through methods such as weighted fusion, the influence of physical laws and historical data is comprehensively considered to construct a digital twin model framework.

[0041] It should be noted that the mathematical expressions for the physical processes, including production processes, operations, and environmental conditions within the workshop, include the heat conduction equation (temperature-dependent), which is expressed as follows: ; in, Indicates temperature Over time The rate of change Indicates temperature. Indicates thermal diffusivity, This represents the rate of temperature diffusion in space. Indicates time; The fluid flow equation (related to flow rate and pressure) is expressed as follows: ; Where u represents the flow velocity vector, The continuity equation for fluids ensures the conservation of mass. Indicates the gradient sign; The ideal gas law (pressure, temperature, and flow rate related) is expressed as follows: ; in, Indicates pressure, Indicates the mass of the gas. Represents the gas constant. Indicates volume; The vibration equation (vibration-related) is expressed as follows: ; in, Indicates the mass of the vibrating body. Indicates the damping coefficient. Represents the spring constant. Indicates displacement. Represents external forces that change over time. Indicates the acceleration of a vibrating body. This indicates the velocity of the vibrating body.

[0042] S2.3. Conduct preliminary simulation of the digital twin model framework to simulate the production process and operating conditions in the workshop and generate preliminary simulation results.

[0043] Specifically, based on the characteristic data such as temperature, humidity, pressure, vibration, and flow rate in the fused feature dataset, a preliminary simulation of the workshop production process is established. During the simulation, the characteristic data such as temperature, humidity, pressure, vibration, and flow rate are combined with the physical characteristics of the workshop, and the operating conditions in the production process, such as production rate and material flow, are considered. Reasonable mathematical relationships and equations are set to simulate the operating status of workshop equipment and production lines. Dynamic simulation, by inputting characteristic data such as temperature, humidity, pressure, vibration, and flow rate, gradually simulates the changes of various variables over time in the production process. For example, temperature changes may affect work efficiency, while pressure and flow rate changes may affect material flow and production rate. Through interaction, the dynamic changes of each variable can be predicted, and its performance under different conditions can be predicted, generating preliminary simulation results.

[0044] S2.4. Through edge computing nodes, the preliminary simulation results are compared with the optimized real-time running data in real time, and an error analysis report is generated.

[0045] Specifically, each production process parameter in the preliminary simulation results is compared with the actual operating data in the optimized real-time operating data. By calculating the error between the simulation results and the optimized real-time operating data, the differences between the simulation results and the optimized real-time operating data are identified, especially in the changing trends of key characteristics such as temperature, humidity, pressure, vibration, and flow rate. Based on the error between the simulation results and the optimized real-time operating data, the performance of the digital twin model under different production conditions is evaluated. By comparing the differences between the simulation data and the actual operating data, the deviations of the digital twin model are identified, and the adaptability under different production environments and operating conditions is analyzed, generating an error analysis report. The error analysis report includes error source analysis, error type classification, error quantification, error distribution analysis, digital twin model optimization suggestions, assessment of the impact of errors on the production process, and error adjustment strategies.

[0046] S2.5. Based on the error analysis report, evaluate the deviations and deficiencies of the digital twin model framework, identify the parts that need to be optimized, and generate error analysis results.

[0047] Specifically, the parameters within the digital twin model framework are evaluated, and the deviations and deficiencies listed in the error analysis report are analyzed, with a focus on errors exceeding ±3%. According to the error analysis report, if the error between the simulation results and the optimized real-time operating data exceeds 3 standard deviations, it indicates a significant deviation requiring further optimization. By comparing the errors between the preliminary simulation results and the optimized real-time operating data, the parts of the digital twin model framework that need optimization are identified, such as model assumptions, parameter settings, or data processing methods. Finally, by combining the error distribution and the impact of errors on the production process listed in the error analysis report, logarithmic... The most critical aspects affecting the accuracy of the digital twin model framework and the optimization of the production process are identified. For example, special attention is paid to parameters with errors exceeding ±5% or deviations exceeding 3 standard deviations under specific production conditions such as production rates exceeding 100 pieces / hour and temperatures exceeding 80°C. According to the error analysis report, if the difference between simulation data and actual operating data exceeds ±5%, it indicates that the parameter error is significant and has a substantial impact on the optimization and accuracy of the production process. Based on the evaluation results, directions and specific measures for optimizing the digital twin model framework are proposed, such as adjusting model assumptions, optimizing parameter settings, or improving data processing methods. Based on the evaluation and optimization suggestions, error analysis results are generated.

[0048] S2.6 Based on the error analysis results, the parameters of the digital twin model architecture are automatically adjusted using the gradient descent method to generate the adjusted digital twin model parameters.

[0049] Specifically, based on the error analysis results, deviations in various parameters within the digital twin model framework are identified, particularly errors in characteristics such as temperature, humidity, pressure, vibration, and flow rate. The gradient descent method is used to automatically adjust the parameters of the digital twin model architecture. The error gradient is calculated, meaning that based on the sensitivity of the error to each parameter provided in the error analysis results, the direction and magnitude of adjustment for each parameter are determined. The parameters of the digital twin model architecture are gradually adjusted to gradually reduce the error until it reaches a minimum value, such as a mean square error of less than 0.01. This is based on industry standards and the precise requirements of the production process, optimizing the performance of the digital twin model architecture, and generating the adjusted digital twin model parameters.

[0050] It should also be noted that gradient descent is an optimization algorithm designed to improve the accuracy of a digital twin model by continuously adjusting its parameters to minimize the loss function. The basic idea is to calculate the gradient of the loss function with respect to each parameter, determine the direction of the fastest change in the loss function, and adjust the parameters along this direction to reduce the value of the loss function. The magnitude of each adjustment is controlled by the learning rate, which determines the step size for each parameter update. By iteratively calculating the gradient and updating the parameters, the loss function eventually converges to a minimum, for example, a mean squared error less than 0.01, thus optimizing the performance of the digital twin model. Gradient descent is widely used in machine learning and deep learning to help digital twin models continuously improve their prediction accuracy during training.

[0051] S2.7 Utilize particle swarm optimization to optimize the architecture of the digital twin model, improve the structure and function of the digital twin model, and generate the optimized digital twin model architecture.

[0052] Specifically, based on the parameters in the digital twin model architecture, each particle in the particle swarm is initialized, with each particle representing a set of parameter combinations within the digital twin model architecture. Each particle iterates through the search space, evaluating its performance based on its current fitness value, which is calculated by assessing the performance of the digital twin model architecture in simulating the production process. Specifically, the fitness value is evaluated by comparing the error between the model's predictions and actual production data, using mean squared error (MSE) as the fitness metric; a smaller MSE indicates a higher fitness. As the particles iterate, each particle updates its position based on its own experience and that of its neighbors, gradually moving closer to the optimal solution. For example, through 100 to 200 iterations, the number of iterations depends on the convergence of the particle swarm optimization algorithm and the accuracy requirements of the digital twin model. The particle swarm continuously optimizes the parameters in the digital twin model architecture, improving its structure and functionality, enabling more accurate simulation of changes during the production process and generating an optimized digital twin model architecture.

[0053] S2.8. Integrate and optimize the adjusted digital twin model parameters and the optimized digital twin model architecture to generate a corrected digital twin model.

[0054] Specifically, the parameters in the adjusted digital twin model are combined with the structural and functional parameters in the optimized digital twin model architecture to ensure coordinated operation and a unified whole. A particle swarm optimization (PSO) method is applied for joint adjustments, making the adjusted digital twin model parameters more compatible with the functionality and structure of the optimized digital twin model architecture, further improving the accuracy and stability of the digital twin model. Through multiple iterative adjustments, such as 20 to 50 iterations, the number of iterations depends on the convergence of the PSO algorithm and the accuracy requirements of the digital twin model. Each iteration optimizes the model based on the principle of minimizing the error between the adjusted and optimized digital twin model parameters and the optimized digital twin model architecture, until the error is reduced to, for example, a mean square error of less than 0.01. Based on the accuracy requirements of the production process, this ensures that the digital twin model reaches a certain level of accuracy and stability, generating the corrected digital twin model.

[0055] It should also be noted that the steps for joint adjustment using the particle swarm optimization method are as follows: Initialize each particle in the particle swarm, with each particle representing a combination of digital twin model parameters and digital twin model architecture; calculate the fitness value (such as mean squared error (MSE)) by comparing the simulation results of the digital twin model in the production process with the actual operating data to evaluate the model's performance and the fit of the structural parameters; update the position of each particle based on its own experience (local optimum) and the experience of all particles (global optimum), gradually adjusting the various parameters of the digital twin model parameters and digital twin model architecture to more accurately reflect changes in the production process; after 100 to 200 iterations, the number of iterations depending on the experience of the particle swarm optimization algorithm and the convergence speed in practical applications, the particle swarm optimization method optimizes the digital twin model architecture and digital twin model parameters, making the function and structure more compatible, and ultimately improving the accuracy and stability of the digital twin model.

[0056] S3. Using a genetic algorithm, incremental learning and continuous optimization are performed on the corrected digital twin model to generate an optimized digital twin model.

[0057] S3.1. Based on the corrected digital twin model, initialize the population of the genetic algorithm, set the crossover rate, mutation rate and selection strategy, and generate the initial population.

[0058] Specifically, based on the corrected digital twin model, the population of the genetic algorithm is initialized. Each individual represents a possible solution to a set of parameters in the digital twin model. The crossover rate, mutation rate, and selection strategy of the genetic algorithm are set. The crossover rate determines how many individuals exchange information to generate new individuals in each selection process. For example, it is set between 0.7 and 0.9 to balance the exploration and convergence speed of the search space while ensuring sufficient diversity, avoiding premature entrapment in local optima. The mutation rate controls the probability of each individual mutating in the genetic algorithm. For example, it is set between 0.01 and 0.05 to ensure sufficient exploration while avoiding excessive mutation that could lead to instability in the search process, ensuring that the genetic algorithm can effectively avoid local optima and maintain the diversity of the search space. The selection strategy determines how to select individuals for reproduction, generally using strategies such as roulette wheel selection or tournament selection. The initial population is generated by setting the crossover rate, mutation rate, and selection strategy.

[0059] S3.2. Based on the initial population, evaluate the fitness of each corrected digital twin model parameter to generate the evaluated optimal population.

[0060] S3.2.1 Based on the corrected digital twin model parameters of the initial population, the fitness evaluation criteria are set using the mean square error method. The fitness performance of each corrected digital twin model parameter in the corrected digital twin model is calculated and quantified, and the fitness value of the corrected digital twin model parameter is generated.

[0061] Specifically, based on the corrected digital twin model parameters of the initial population, the fitness evaluation criterion is set using the mean squared error method. Each corrected digital twin model parameter is substituted into the corrected digital twin model, and the production process is predicted using the corrected digital twin model to obtain predicted values. The error between the predicted value and the actual production data is calculated using the mean squared error. The error is the difference between the predicted value and the actual value for each data point. The errors of all data points are squared and summed. Then, the average of the sum of squared errors is calculated to obtain the mean squared error. The fitness value of the corrected digital twin model parameters is generated based on the magnitude of the mean squared error. The smaller the fitness value, the higher the prediction accuracy of the corrected digital twin model and the better the matching degree between the corrected digital twin model parameters and the actual data.

[0062] S3.2.2 Based on the fitness values ​​of the corrected digital twin model parameters, the elite selection method is used to select the corrected digital twin model parameters with the best fitness, and an optimal population is generated after evaluation.

[0063] Specifically, based on the fitness values ​​of the corrected digital twin model parameters, the fitness of each corrected digital twin model parameter is ranked, and the individual with the best fitness is selected. For example, the individual with the lowest mean squared error fitness value is selected. An elite selection method is used to retain corrected digital twin model parameters with higher fitness values ​​to ensure that they enter the next generation of the population. For example, the top 20% of the best individuals are selected with fitness values ​​lower than the set corrected digital twin model parameters. This ensures that the best solution in each generation of the genetic algorithm is preserved and passed on to the next generation. Through the elite selection method, an evaluated and optimized population is generated.

[0064] It should also be noted that the specific parameters of the corrected digital twin model vary depending on the application scenario and actual needs. The initial value ranges are set as follows: the initial temperature range is set between 20℃ and 100℃, based on the operating specifications of the workshop equipment and the actual conditions of the production environment; for example, the temperature range reflects the normal operating temperature range of the equipment during production, and the temperature usually varies within this range during operation; the humidity range is set between 30% and 80%, which is to ensure that the production process is not affected by excessively high or low air humidity and maintain normal operation; the pressure range is set between 1 and 10 bar, based on the pressure bearing standards of the workshop and pipelines, to avoid equipment damage due to overpressure or underpressure; the vibration range is set between 0 and 5 mm, based on vibration tolerance, to prevent equipment failure due to excessive vibration; the flow rate range is set between 0 and 1000 L / min, which takes into account the actual requirements of material flow during production, ensuring stable flow within this range and preventing production efficiency from being affected by excessively high or low flow rates.

[0065] S3.3 Perform crossover operations on the selected population after evaluation to generate new corrected digital twin model parameters, and randomly adjust the new corrected digital twin model parameters through mutation operations to generate a new generation of corrected digital twin model parameters.

[0066] Specifically, when performing crossover operations on the selected population after evaluation, two corrected digital twin model parameters with high fitness values ​​are selected for crossover, exchanging some parameters, such as temperature, humidity, pressure, vibration, and flow rate, to generate new corrected digital twin model parameters. The purpose of crossover is to combine advantages to generate new solutions that may be more suitable for the production process. The new corrected digital twin model parameters are then randomly adjusted through mutation operations. Mutation operations involve small, random changes to individual parameters, such as ±5%, to ensure sufficient exploration space while avoiding instability caused by excessive mutations. This also ensures that the mutated solution remains within a reasonable physical range, effectively improving population diversity and avoiding getting trapped in local optima. After each mutation, the new corrected digital twin model parameters continue to participate in fitness evaluation, generating a new generation of corrected digital twin model parameters through crossover and mutation operations.

[0067] S3.4 Compare the fitness of the new generation of corrected digital twin model parameters with the adjusted digital twin model parameters, select the digital twin model parameters with the best fitness, and generate the optimized digital twin model.

[0068] Specifically, the fitness values ​​of each new-generation corrected digital twin model parameter and the adjusted digital twin model parameter are calculated. By comparing the errors exhibited by the new-generation corrected digital twin model parameters and the adjusted digital twin model parameters in the production process, the digital twin model parameter with the best fitness, that is, the parameter that performs best in minimizing error, is selected. The optimal digital twin model parameter is used as the new input to generate the optimized digital twin model.

[0069] It should be noted that a genetic algorithm is used to incrementally learn and continuously optimize the corrected digital twin model. As a global optimization method, the genetic algorithm can effectively search in a large number of possible parameter spaces and adaptively adjust the parameters of the digital twin model, thereby continuously improving the prediction accuracy and adaptability of the digital twin model. Compared with traditional methods, traditional digital twin models usually rely on simple optimization methods and lack the ability to learn and evolve on their own. This invention uses a genetic algorithm for incremental learning, which enables the digital twin model to dynamically adjust and optimize according to real-time production data, continuously improving accuracy and robustness.

[0070] S4. Based on the optimized digital twin model, real-time operating data is fed back to production control in real time, automatically adjusting production process operating parameters and generating precise control commands.

[0071] S4.1 Based on the optimized digital twin model, collect various data in the workshop production process in real time and feed them back to the production scheduling center to generate a production data stream.

[0072] Specifically, based on the optimized digital twin model, real-time operational data during the workshop production process, such as temperature, humidity, pressure, vibration, and flow rate, are collected in real time through sensors and monitoring equipment, ensuring the accuracy and timeliness of the real-time operational data. The data such as temperature, humidity, pressure, vibration, and flow rate are standardized and denoised, and synchronized according to timestamps. The processed real-time operational data is fed back to the production scheduling center in real time through the data transmission path, and organized according to the time sequence and stages of the production process. The integrated real-time operational data forms a production data stream, which contains key indicator information of the production process, such as real-time temperature, real-time humidity, real-time pressure, real-time vibration, and real-time flow rate.

[0073] S4.2 Analyze the production data flow through the optimized digital twin model, calculate and identify the production process operation parameters that need to be adjusted, and adjust the production process operation parameters in real time according to the analysis results to generate the adjusted production process operation parameters.

[0074] Specifically, various data points from the production data stream, such as real-time temperature, humidity, pressure, vibration, and flow rate, are input into the optimized digital twin model. The optimized digital twin model simulates the changing trends of the production process based on these data points and calculates key parameters related to production efficiency, product quality, and equipment health. The analysis results of the optimized digital twin model reveal the production process operating parameters that need adjustment, such as control variables like temperature and pressure. Based on the analysis results of the optimized digital twin model, the production process operating parameters are adjusted in real time to ensure the production process reaches its optimal state, generating the adjusted production process operating parameters.

[0075] S4.3 Verify and optimize the adjusted production process operating parameters to generate precise control instructions.

[0076] Specifically, the adjusted production process operating parameters are input into the optimized digital twin model for simulation verification to check whether the adjusted parameters can effectively improve the efficiency, stability, and product quality of the production process. Through simulation, the optimized digital twin model will provide feedback on the impact of the adjusted operating parameters on the production process, verifying whether the production objectives are met. If further improvement is needed, the optimized digital twin model will propose optimization schemes, adjusting the specific values ​​or settings of the production process operating parameters. After verification and optimization, precise control commands are generated, which include precise regulation of various key parameters in the production process, such as temperature, pressure, and vibration, ensuring precise control and efficient operation of the production process.

[0077] It should also be noted that production targets refer to the specific production requirements and expected results set during the workshop production process, including but not limited to improving production efficiency, ensuring stable product quality, reducing resource consumption, improving equipment stability, and reducing the failure rate during the production process. The formulation of production targets is usually based on the workshop's production capacity, market demand, and product standards to ensure that production tasks are completed efficiently within the specified time while meeting product quality requirements.

[0078] S5. Based on precise control instructions, continuously monitor real-time operating data during the production process, use machine learning prediction algorithms to automatically identify potential changes, trigger adaptive adjustments, and generate digital twin model correction strategies.

[0079] S5.1. Based on precise control instructions, continuously monitor real-time operating data during the production process, and analyze the real-time operating data through machine learning prediction algorithms to automatically identify potential changes in the production process.

[0080] Specifically, based on precise control commands, real-time operational data during the production process is continuously monitored, including key indicators such as temperature, humidity, pressure, vibration, and flow rate. Real-time operational data is constantly collected and transmitted to the data processing stage to ensure that changes in the production process are reflected in real time. Machine learning prediction algorithms analyze the real-time operational data, combining historical and real-time data to train and optimize a digital twin model, identifying potential changes in the production process. By analyzing trends, fluctuations, and anomaly patterns in the real-time operational data, machine learning prediction algorithms automatically identify potential problems that may lead to decreased production efficiency or product quality fluctuations, and promptly issue warnings or optimization suggestions.

[0081] S5.2 Based on potential changes, trigger an adaptive adjustment process to further optimize the adjusted production process operation parameters and generate a digital twin model correction strategy.

[0082] Specifically, based on potential changes, an adaptive adjustment process is triggered. This involves adjusting production process operation parameters in real time based on anomalies or changes identified by machine learning prediction algorithms. Based on feedback information provided by the optimized digital twin model, key parameters such as temperature, humidity, pressure, vibration, and flow rate are automatically fine-tuned to adapt to real-time changes in the production process. The adjusted parameters generate a digital twin model correction strategy, which ensures continuous optimization of the production process based on new data and changes, maintaining stable production efficiency and product quality. The digital twin model correction strategy provides specific guidance for precise control and further optimization of the production process.

[0083] It should be noted that by combining real-time operational data monitoring, machine learning prediction, and adaptive adjustment, potential problems in the production process can be dynamically identified based on real-time operational data, and the digital twin model can be automatically adjusted to cope with these changes. Compared with traditional methods that usually employ static adjustment or manual intervention, this invention achieves real-time adaptive intelligent optimization through machine learning prediction algorithms, which greatly improves response speed and optimization accuracy.

[0084] In summary, this invention effectively improves data quality by: collecting and preprocessing real-time operational data to generate optimized real-time operational data, providing high-quality data support for subsequent digital twin model calibration and ensuring that the digital twin model can more accurately reflect the production process; and by performing real-time comparison through edge computing nodes, generating error analysis results, and automatically adjusting the parameters of the digital twin model, it achieves accurate calibration of the digital twin model, thereby enabling the digital twin model to better match the actual production situation.

[0085] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for correcting a digital twin model of a factory workshop based on data acquisition, characterized in that: include, Collect real-time operational data, preprocess it, and generate optimized real-time operational data; Based on the optimized real-time running data, a digital twin model framework is constructed, and real-time comparison is performed through edge computing nodes to generate error analysis results, automatically adjust the parameters of the digital twin model, and generate a corrected digital twin model. Genetic algorithms are used to incrementally learn and continuously optimize the corrected digital twin model to generate an optimized digital twin model. Based on the optimized digital twin model, real-time operating data is fed back to production control in real time, automatically adjusting the operating parameters of the production process and generating precise control instructions. Based on precise control instructions, the system continuously monitors real-time operational data during the production process, utilizes machine learning prediction algorithms to automatically identify potential changes, triggers adaptive adjustments, and generates digital twin model correction strategies.

2. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The real-time operating data includes temperature, humidity, pressure, operating status, vibration, and flow rate.

3. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for generating optimized real-time running data are as follows. Preliminary noise removal and data smoothing are performed on the real-time running data to generate smoothed real-time running data; The smoothed real-time running data is imputed for missing values ​​and standardized to generate standardized real-time running data. Anomaly detection is performed on the standardized real-time running data to generate optimized real-time running data.

4. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for constructing the digital twin model framework are as follows. Feature extraction is performed on the optimized real-time operating data to identify and select features related to the workshop production process environment and operating status. Information from different data sources is integrated through time series data analysis methods to generate a fused feature dataset. By combining the fused feature dataset with the physical characteristics of workshop production, a digital twin model framework is constructed through physical modeling and data-driven modeling methods.

5. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for generating the error analysis results are as follows: A preliminary simulation of the digital twin model framework was conducted to simulate the production process and operating conditions in the workshop, generating preliminary simulation results. By using edge computing nodes, the preliminary simulation results are compared with the optimized real-time running data in real time, and an error analysis report is generated. Based on the error analysis report, assess the deviations and deficiencies of the digital twin model framework, identify the parts that need optimization, and generate error analysis results.

6. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for generating the corrected digital twin model are as follows. Based on the error analysis results, the parameters of the digital twin model architecture are automatically adjusted using the gradient descent method to generate the adjusted digital twin model parameters. The particle swarm optimization method is used to optimize the architecture of the digital twin model, improve the structure and function of the digital twin model, and generate the optimized digital twin model architecture. The adjusted digital twin model parameters and the optimized digital twin model architecture are integrated and optimized to generate a corrected digital twin model.

7. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for generating the optimized digital twin model are as follows. Based on the corrected digital twin model, initialize the population of the genetic algorithm, set the crossover rate, mutation rate and selection strategy, and generate the initial population; Based on the initial population, the fitness of each corrected digital twin model parameter is evaluated to generate the optimal population after evaluation. Crossover operations are performed on the selected population after evaluation to generate new corrected digital twin model parameters. The new corrected digital twin model parameters are then randomly adjusted through mutation operations to generate a new generation of corrected digital twin model parameters. The fitness of the new generation of corrected digital twin model parameters is compared with that of the adjusted digital twin model parameters. The digital twin model parameters with the best fitness are selected to generate the optimized digital twin model.

8. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 7, characterized in that: The specific steps for generating and evaluating the optimal population are as follows: Based on the corrected digital twin model parameters of the initial population, the fitness evaluation criteria are set using the mean squared error method. The fitness performance of each corrected digital twin model parameter in the corrected digital twin model is calculated and quantified, and the fitness value of the corrected digital twin model parameter is generated. Based on the fitness values ​​of the corrected digital twin model parameters, an elite selection method is used to select the corrected digital twin model parameters with the best fitness, and an optimal population is generated after evaluation.

9. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps for generating precise control commands are as follows: Based on the optimized digital twin model, various data during the workshop production process are collected in real time and fed back to the production scheduling center to generate a production data stream; By analyzing the production data flow through the optimized digital twin model, the production process operation parameters that need to be adjusted are calculated and identified, and the production process operation parameters are adjusted in real time based on the analysis results to generate the adjusted production process operation parameters. The adjusted production process operating parameters are verified and optimized to generate precise control commands.

10. The method for correcting a digital twin model of a factory workshop based on data acquisition as described in claim 1, characterized in that: The specific steps of the generated digital twin model correction strategy are as follows. Based on precise control instructions, the system continuously monitors real-time operational data during the production process and analyzes the data using machine learning prediction algorithms to automatically identify potential changes in the production process. Based on potential changes, an adaptive adjustment process is triggered to further optimize the adjusted production process operating parameters and generate a digital twin model correction strategy.