Corrugated paper production line virtual debugging and operation optimization system based on digital twinning

Abstract

The invention discloses a corrugated paper production line virtual debugging and operation optimization system based on digital twinning. The system comprises a digital twinning building module, a multi-dimensional model building and training module, a dominant variable recognition module and a virtual debugging and optimization module. The digital twin construction module is used for acquiring physical entity data and operation condition data of a production line and constructing a digital twin consistent with the physical production line; and the multi-dimensional model construction and training module is used for synchronously constructing a multi-physical sensitive dimension model in the digital twin for key processes of a production line, and relates to the technical field of corrugated paper production. According to the corrugated paper production line virtual debugging and operation optimization system based on digital twinning, by constructing the digital twinning body and the virtual debugging module, the production process can be debugged and optimized in advance in a virtual environment, the debugging period of a traditional production line is remarkably shortened, and the debugging cost and risk in actual production are reduced.

Description

Virtual commissioning and operation optimization system for corrugated paper production lines based on digital twins Technical Field

[0001] This invention relates to the field of corrugated paper production, and more specifically, to a virtual debugging and operation optimization system for a corrugated paper production line based on digital twins. Background Technology

[0002] Traditional debugging and optimization methods face numerous challenges in corrugated paper production. On one hand, physical production line debugging requires actual input of raw materials and equipment operation, which is not only costly but also time-consuming. Improper parameter settings can lead to substandard product quality and increased equipment wear and tear. On the other hand, corrugated paper production involves multiple complex processes that are interconnected and influenced by various factors, such as raw material properties, environmental parameters, and process parameters. Traditional methods struggle to comprehensively and accurately analyze the impact of these factors on the production process and product quality, and fail to quickly and accurately identify the dominant process variables, thus hindering efficient production line optimization.

[0003] Furthermore, with intensifying market competition and increasingly stringent customer demands for product quality, corrugated paper manufacturers need to continuously improve production efficiency, reduce costs, and ensure stable product quality. However, traditional production methods are inadequate in handling complex and ever-changing production conditions and personalized customization needs, failing to adjust production parameters in a timely manner to adapt to different situations, and hindering the intelligent and refined operation management of production lines. Therefore, a new technological approach is urgently needed to address these issues and achieve efficient virtual commissioning and operational optimization of corrugated paper production lines. Summary of the Invention

[0004] The purpose of this invention is to provide a virtual debugging and operation optimization system for corrugated paper production lines based on digital twins. This system solves the problem that existing production methods are inadequate in dealing with complex and ever-changing production conditions and personalized customization needs. They are unable to adjust production parameters in a timely manner to adapt to different situations, making it difficult to achieve intelligent and refined operation management of the production line and meet user requirements.

[0005] This invention achieves the above objectives through the following technical solution: a virtual debugging and operation optimization system for a corrugated paper production line based on digital twins. The system includes: a digital twin construction module, a multi-dimensional model construction and training module, a dominant variable identification module, and a virtual debugging and optimization module. The digital twin construction module is used to acquire physical entity data and operating condition data of the production line and construct a digital twin that is consistent with the physical production line. The multi-dimensional model construction and training module is used to simultaneously construct multi-physical sensitive dimension models for key processes of the production line within the digital twin, and complete model training and parameter calibration. The dominant variable identification module is used to calculate deviation quantification indicators based on the output state of the trained multi-physical sensitive dimension models, and identify target dominant process variables through the deviation quantification indicators. The virtual debugging and optimization module is used to complete virtual debugging based on the target dominant process variables, and optimize the production line operating parameters according to the debugging results, thereby achieving production line operation optimization.

[0006] Furthermore, the multi-physical sensitivity dimension model includes: a macroscopic process model of the first physical sensitivity dimension and a constraint model of the second physical sensitivity dimension; the macroscopic process model generates a comprehensive response to both the properties of production raw materials and environmental disturbances; the constraint model is highly sensitive only to the key properties and core physical transmission behavior of production raw materials and is low in sensitivity to non-key interference factors.

[0007] Furthermore, the physical entity data acquired by the digital twin construction module includes: geometric parameters of production line equipment, equipment connection relationships, and key process boundary conditions; the acquired operating condition data includes raw material attribute data, real-time process parameters, and environmental parameters; the digital twin construction module is also used to preprocess the physical entity data and operating condition data, the preprocessing including data cleaning, outlier removal, and normalization; based on the preprocessed data, a digital twin is constructed using 3D modeling and physical modeling techniques to ensure that the digital twin achieves geometric, physical, and behavioral consistency with the physical production line.

[0008] Furthermore, the multi-dimensional model construction and training module includes the following steps: When constructing the macroscopic process model of the first physically sensitive dimension, macroscopic process equations for key processes are established based on the laws of conservation of mass, conservation of energy, and fluid mechanics. The inputs of the macroscopic process model include key attribute parameters of raw materials, environmental parameters, and core process parameters, and the outputs include product forming quality parameters and process energy consumption. When constructing the constraint model of the second physically sensitive dimension, a constraint model is established based on the core attribute transmission equations and core physical conduction equations of raw materials, retaining only the core physical mechanisms and shielding the influence of non-critical interference factors.

[0009] Furthermore, the multi-dimensional model construction and training module includes the following steps during model training: data partitioning, initial parameter setting, loss function definition, optimizer configuration, iterative training, and parameter calibration. The data partitioning step divides the preprocessed operating data into training, validation, and test sets according to a preset ratio. The initial parameter setting step sets the initial parameters of the macroscopic process model and constraint model based on theoretical calculations and engineering experience. The loss function definition step uses mean squared error as the training loss function. The optimizer configuration step uses an adaptive optimizer and sets relevant parameters. The iterative training step sets the maximum number of training iterations and adopts an early stopping strategy. The parameter calibration step further optimizes the model parameters based on the trained model using a parameter optimization algorithm to ensure that the model prediction error is less than a preset threshold.

[0010] Furthermore, the dominant variable identification module in the digital twin virtual debugging scenario includes the following steps: synchronously running the trained macroscopic process model and constraint model, obtaining the output state vectors of the two types of models, calculating the deviation quantification index of the output state of the multi-physical sensitivity dimension model, wherein the deviation quantification index is calculated using a normalized distance algorithm to calculate the degree of deviation; setting a deviation threshold, and determining that the target dominant process variable fluctuates under the current operating condition when the deviation quantification index exceeds the deviation threshold; determining the target dominant process variable through sensitivity analysis, calculating the influence coefficient of each process parameter on the deviation quantification index, and selecting a preset number of process parameters with the largest absolute value of the sensitivity coefficient as the target dominant process variable.

[0011] Furthermore, the deviation threshold is determined based on the dual-model output data of the stable operation phase of the production line; the stable operation phase of the production line refers to the operation phase in which there are no quality abnormalities or major fluctuations in operating conditions for a continuously preset period of time. The statistical characteristic parameters of the deviation quantification index in this phase are calculated, and the deviation threshold is determined based on the statistical characteristic parameters.

[0012] Furthermore, the virtual debugging and optimization module constructs a virtual debugging scenario based on a digital twin, including the following steps: setting the debugging range and step size of the target dominant process variable, wherein the debugging range is determined based on the production line process specifications, and the step size does not exceed a preset proportion of the maximum value of the variable; under each debugging condition, running the trained multi-physical sensitivity dimension model and recording the deviation quantification index and product quality index; with the goal of minimizing the deviation quantification index and optimizing the product quality index, using an intelligent optimization algorithm to solve for the optimal value of the target dominant process variable.

[0013] Furthermore, the optimization objective function of the intelligent optimization algorithm is set with weight coefficients, which are based on the priority requirements of the production line. When the production line prioritizes product quality, the weight coefficient related to product quality is higher than that related to process stability. When the production line prioritizes process stability, the weight coefficient related to process stability is higher than that related to product quality. When the priorities of quality and stability are balanced, a balanced weight coefficient is set. The product quality indicators include multiple core quality evaluation dimensions, and each quality evaluation dimension is set with a corresponding weight coefficient and a comprehensive product quality score is calculated.

[0014] Furthermore, the virtual debugging and optimization module is also used to send the optimized process parameters to the physical production line to update the operating parameters; collect the operating data and product quality data of the physical production line in real time and feed them back to the digital twin to perform online correction of the multi-physical sensitive dimension model; set a correction trigger threshold, and when the correction trigger condition is met, trigger the online correction of the model parameters. During the correction, an incremental training method is used to adjust the learning rate and the number of iterations to ensure the sustainability of the optimization effect.

[0015] The beneficial effects of this invention are as follows: 1. By constructing a digital twin and a virtual debugging module, the production process can be debugged and optimized in advance in a virtual environment, which significantly shortens the debugging cycle of traditional production lines and reduces the debugging cost and risk in actual production.

[0016] 2. Through multi-physical sensitivity model and intelligent optimization algorithm, it can accurately identify the dominant variables affecting production line efficiency and product quality, and optimize them in a targeted manner to ensure that various process parameters in the production process are optimally adjusted, thereby improving product quality and production efficiency.

[0017] 3. By optimizing key processes in the production process, this invention can effectively reduce energy consumption and material waste in the production process, improve energy utilization efficiency, and reduce scrap rate, thereby achieving high economic and environmental benefits.

[0018] 4. This system can collect real-time operating data and product quality data from the physical production line and feed them back to the digital twin for online correction, ensuring the stability of the production process during long-term operation and enabling it to quickly adapt to changes in production conditions, achieving dynamic adjustment and continuous optimization.

[0019] 5. Through the virtual debugging and optimization module, combined with intelligent optimization algorithms, this system can automatically calculate the optimal process parameters based on the set production priorities, helping production managers make more scientific and accurate decisions, thereby improving the overall operational efficiency of the production line. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and, together with their descriptions, serve to explain this application and do not constitute an undue limitation thereof. In the drawings: Figure 1 is an overall system block diagram of the present invention; Figure 2 is a flowchart of the digital twin construction process of the present invention; Figure 3 is a flowchart of the multi-dimensional model construction and training process of the present invention; Figure 4 is a flowchart of the virtual debugging and optimization process of the present invention. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0022] Example 1: Please refer to Figures 1-4. This invention provides a technical solution: a virtual debugging and operation optimization system for a corrugated paper production line based on digital twins. The system includes: a digital twin construction module, a two-dimensional model construction and training module, a dominant variable identification module, and a virtual debugging and optimization module. The digital twin construction module is used to acquire physical entity data and operating condition data of the corrugated paper production line to construct a digital twin of the corrugated paper production line. The physical entity data refers to various data generated by the actual physical equipment of the corrugated paper production line during operation, such as static data like the size, weight, and material of the equipment, and dynamic data like temperature, pressure, and speed during equipment operation. The operating condition data reflects the operation of the corrugated paper production line under different production conditions. The system includes status data such as production speed, production batch size, raw material characteristics, ambient temperature, and humidity. A digital twin is a digital mapping of a physical entity in virtual space, constructed by acquiring physical entity data and operational status data. It can reflect the physical entity's status and behavior in real time and accurately. A dual-dimensional model construction and training module is used to simultaneously build dual-physical-sensitive dimension models for key processes in the corrugated paper production line within the digital twin, completing model training and parameter calibration. Key processes include gluing, corrugating, and hot-pressing. These key processes are crucial to product quality and production efficiency during corrugated paper production, specifically including gluing, corrugating, and hot-pressing. The first process involves evenly applying glue to corrugated cardboard. The amount and uniformity of the glue application affect the bonding strength and quality of the cardboard. The second process, corrugating, involves using corrugating rollers to shape the paper into corrugated forms. The shape, height, and density of the corrugations significantly impact the strength and cushioning performance of the cardboard. The third process, hot pressing, involves heating and pressing the glued and corrugated cardboard to set its shape and enhance its bonding strength. The temperature, pressure, and time of the hot pressing affect the quality and performance of the cardboard. The dual-physical-sensitive dimension model is a model built for key processes in a corrugated paper production line. This model considers two dimensions closely related to the physical process, enabling a more comprehensive and accurate simulation of the physical behavior and characteristics of key processes. Model training... Parameter calibration involves training the dual-physical sensitivity dimension model using a large amount of actual production data, adjusting the model's parameters to make the model's output results as consistent as possible with actual production conditions, thereby improving the model's accuracy and reliability. The dominant variable identification module is used to calculate the differential deviation based on the output state of the trained dual-physical sensitivity dimension model, and to identify the moisture-dominant process variable through the differential deviation. Here, the output state is the calculation result of the dual-physical sensitivity dimension model under given input conditions, reflecting the operating status of key processes under different parameter settings; the differential deviation is a value that reflects the degree of state change obtained by calculating the difference between the model's output state at different times or under different conditions, and is used to measure the degree of deviation of the model's output state.Moisture content is the process variable that has the greatest impact on the moisture content of corrugated paper during production. It is identified through analysis of differential deviations. Moisture content is one of the key factors affecting corrugated paper quality. The virtual debugging and optimization module is used to perform virtual debugging based on the moisture content process variable and optimize the production line operating parameters based on the debugging results, thereby achieving production line operation optimization. Virtual debugging simulates the operation of the corrugated paper production line in a virtual environment constructed by a digital twin. Adjustments and tests are performed on various aspects of the production line based on the moisture content process variable to verify and optimize the production line's operating parameters and process flow. Operating parameters are... Various parameters affecting the operating status and product quality of the corrugated paper production line, such as sizing amount, corrugated roll speed, hot pressing temperature and pressure, etc.; production line operation optimization involves adjusting and improving the operating parameters of the production line based on the results of virtual debugging, so that the production line can achieve higher production efficiency, better product quality and lower energy consumption in actual operation; among them, the dual physical sensitivity dimension model includes a macroscopic process model of the first physical sensitivity dimension and a constraint model of the second physical sensitivity dimension. The macroscopic process model produces a comprehensive response to the moisture content of the raw paper, impurities and environmental disturbances, while the constraint model is only highly sensitive to the moisture and heat transfer behavior of the raw paper and low sensitive to impurities and random disturbances.

[0023] It should be noted that, during use, the digital twin construction module acquires physical entities and operational data to construct a digital twin, providing a precise virtual mapping for subsequent operations. The dual-dimensional model construction and training module constructs a macro-process model and a constraint model for key processes. The former comprehensively responds to multiple factors, while the latter focuses on the moisture and heat transfer of the raw paper, balancing comprehensiveness and accuracy to improve the model's ability to simulate complex production conditions. After training and calibration, it can more reliably guide production. The dominant variable identification module calculates the differential deviation based on the model output, accurately identifying the moisture-dominant process variable and grasping the key factors affecting production. The virtual debugging and optimization module performs virtual debugging based on this variable, optimizing operating parameters, avoiding the high costs and risks of actual debugging, quickly finding the best parameter combination, achieving efficient, stable, and high-quality operation of the production line, and improving overall production efficiency and product quality.

[0024] In one embodiment, acquiring physical entity data and operating condition data of a corrugated paper production line to construct a digital twin of the corrugated paper production line includes: acquiring physical entity data of the corrugated paper production line, which includes geometric parameters of the production line equipment, equipment connection relationships, and process boundary conditions of key processes; acquiring operating condition data of the production line, which includes raw paper property data, real-time process parameters, and environmental parameters. Raw paper property data includes raw paper moisture content, impurity content, and fiber density; real-time process parameters include sizing amount, corrugating forming pressure, hot pressing temperature and time; and environmental parameters include ambient temperature, humidity, and air pressure. Preprocessing the physical entity data and operating condition data includes data cleaning, outlier removal, and normalization. The normalization expression is: ,in, For the first One original data sample, The total number of data samples, The data is normalized; based on the preprocessed data, a digital twin is constructed using 3D modeling and physical modeling techniques. The digital twin achieves geometric, physical, and behavioral consistency with the physical production line.

[0025] This design acquires physical entity data such as the geometric parameters of the production line equipment, as well as operational data such as the properties of the raw paper. After data cleaning and preprocessing, a digital twin is constructed using 3D and physical modeling technologies to ensure consistency with the physical production line in many aspects. Comprehensive data acquisition and preprocessing guarantee data quality, laying the foundation for building an accurate digital twin. The combination of 3D and physical modeling technologies can highly replicate the physical production line, achieving geometric, physical, and behavioral consistency. This allows the digital twin to accurately simulate the operation of the physical production line, providing a reliable platform for subsequent virtual debugging and optimization, reducing actual debugging costs and risks, and improving production optimization efficiency.

[0026] In one embodiment, a dual-physical-sensitivity model is simultaneously constructed in the digital twin for key processes of the corrugated paper production line. Model training and parameter calibration are then completed. This includes: constructing a macroscopic process model for the first physical-sensitivity dimension: based on the laws of conservation of mass, conservation of energy, and fluid mechanics, macroscopic process equations for key processes are established. The inputs to the macroscopic process model include the moisture content of the base paper, impurity content, ambient temperature, ambient humidity, sizing amount, forming pressure, and hot-pressing temperature. The outputs include corrugated paper forming quality parameters and process energy consumption. The general expression of the macroscopic process model is:

[0027] in, for The time-matter model state vector includes the corrugated paper's moisture content, temperature, forming thickness, etc. For process input vector, This is the interference vector, containing impurity content and environmental disturbances. The macroscopic process mapping function is used; a constraint model for the second physical sensitivity dimension is constructed: based on the moisture transport equation and heat conduction equation of the original paper, a constraint model is established. The constraint model only retains the physical mechanisms related to moisture and heat transfer, and shields the influence of impurities and random disturbances; the expression of the constraint model is:

[0028] in, for Time and location The moisture content of the raw paper at that location, for Time and location The temperature at that location The water diffusion coefficient is... Where is the thermal diffusivity, For the water migration source term function, The heat conduction source term function is used; Model training steps: Step 1: Data partitioning. The preprocessed operating condition data is divided into training set, validation set, and test set in a ratio of 7:2:1. The training set is used for model parameter learning, the validation set is used for hyperparameter tuning, and the test set is used for model generalization performance evaluation; Step 2: Initial parameter setting. The initial parameters of the macroscopic process model are set based on theoretical calculations and engineering experience: the initial value of the correlation coefficient of the application amount is taken as... The initial value of the molding pressure influence coefficient is taken as follows: The initial value of the hot-pressing temperature coefficient is taken as Initial parameters of the constrained model: moisture diffusion coefficient Initial value thermal diffusivity Initial value Step 3: Define the loss function, using mean squared error (MSE) as the training loss function, expressed as:

[0029] in, These are the model's predicted values. These are measured values. For the sample size; Step 4: Optimizer configuration, using the Adaptive Momentum Estimation (Adam) optimizer, with an initial learning rate of 0.001 and a decay coefficient. , The weight decay coefficient is Batch size set to 32; Step 5: Iterative training, setting the maximum number of training iterations to 1000 rounds, and adopting an early stopping strategy: when the validation set loss does not decrease for 20 consecutive rounds, stop training and save the current optimal model parameters; Step 6: Parameter calibration, based on the trained model, further optimize the model parameters using the least squares method, and set a preset threshold for the model prediction error. threshold The quality fluctuation range of historical qualified products on the production line is determined by extracting the standard deviation of key quality indicators of qualified products within the past three months. Key quality indicators include corrugated strength, bond strength, and moisture content. Ensure that the model prediction error is less than .

[0030] This design constructs two physically sensitive dimension models—a macroscopic process model and a constraint model—within the digital twin for key processes. The construction basis and expressions are explained in detail, along with the model training steps, including data partitioning and parameter settings. This dual-model approach takes into account different physical characteristics. The macroscopic process model comprehensively considers multiple factors, while the constraint model focuses on moisture and heat transfer, improving the model's adaptability to complex operating conditions. The detailed training steps ensure reasonable model parameters, enhancing model accuracy and generalization ability. This provides strong support for accurately identifying moisture-dominant process variables, thereby guaranteeing the optimization effect of the production line.

[0031] In one embodiment, the differential deviation is calculated based on the output state of the trained dual-physical-sensitivity model, and the moisture-dominant process variable is identified through the differential deviation. This includes: in a digital twin virtual debugging scenario, simultaneously running the trained macroscopic process model and constraint model to obtain the output state vectors of the two models, denoted as follows:

[0032] This represents the output of the macroscopic process model.

[0033] This represents the output of the constraint model. The output state dimension includes key indicators such as moisture distribution, temperature distribution, and molding strength; the differential deviation of the output state of the dual-physical-sensitivity dimension model is calculated, and the degree of deviation is calculated using normalized Euclidean distance, expressed as: ,

[0034] in, The difference deviation is... For the macroscopic process model Dimensional output, For the constraint model Dimensional output; set deviation threshold threshold The determination rule is as follows: Based on the dual-model output data of the stable operation phase of the production line, the stable operation phase of the production line refers to 72 consecutive hours without quality abnormalities or significant fluctuations in operating conditions. The mean of the differential deviation during this phase is calculated. and standard deviation ,Pick ,when At that time, it was determined that the moisture-dominant process variable under the current operating conditions fluctuated; the moisture-dominant process variable was identified through sensitivity analysis, and the influence coefficients of each process parameter on the differential deviation were calculated:

[0035] in, For the first The sensitivity coefficient of each process parameter For the first Each process parameter; select the one with the largest absolute value of the sensitivity coefficient. One process parameter is used as the dominant process variable for moisture content. The value is based on the total number of process parameters. Confirmed: When hour, ;when hour, ;when hour, The process parameters include the amount of glue applied, the hot pressing temperature, the hot pressing time, and the paper feed speed.

[0036] This design allows for the use of dual models to acquire output state vectors in a virtual debugging scenario, calculation of differential deviation, setting thresholds to assess fluctuations in moisture-dominant process variables, and sensitivity analysis to identify specific variables. By calculating differential deviation and setting thresholds, fluctuations in moisture-dominant process variables can be detected promptly. Sensitivity analysis can accurately identify key variables, helping to quickly pinpoint critical factors affecting production quality. This provides a clear direction for subsequent virtual debugging and parameter optimization, improving the targeting and effectiveness of debugging and optimization, and ultimately enhancing the quality of production line operation.

[0037] In one embodiment, virtual debugging is performed based on the moisture-dominant process variable, and the production line operating parameters are optimized based on the debugging results to achieve production line operation optimization. This includes: constructing a virtual debugging scenario based on a digital twin, setting the debugging range and step size of the moisture-dominant process variable, wherein the debugging range is determined based on the production line process specifications, and the step size does not exceed a certain percentage of the variable's maximum value. Under each debugging condition, the trained dual-physical-sensitivity model is run to record the differential deviation and the quality indicators of the corrugated paper product, including corrugated strength, adhesive strength, and moisture content pass rate. With the goal of minimizing the differential deviation and optimizing the product quality indicators, the particle swarm optimization algorithm is used to solve for the optimal value of the moisture-dominated process variable. The optimization objective function is: ,

[0038] in, For the moisture-dominated process variable vector, , The weighting coefficients and The weighting coefficients are determined as follows: The analytic hierarchy process (AHP) is used in conjunction with the production line's priority requirements. When the production line prioritizes product quality, the core priority of product quality refers to the weighting of the defect rate. ,Pick , When the production line prioritizes process stability, the core priority of process stability refers to the weight given to continuous operation without fluctuations. ,Pick , When quality and stability are given equal priority, take... , ; The formula for calculating the overall product quality score is as follows:

[0039] in, Assigning weights to each quality indicator. Corrugated strength weight Adhesion strength weight Moisture content pass rate weight , The normalized scores for each quality indicator, and the range of normalized scores. Particle swarm optimization algorithm parameter settings: population size set to 50, maximum number of iterations to 200, initial inertia weight value of 0.9, linearly decreasing to 0.4 during iteration, learning factor... The optimized process parameters are distributed to the physical production line to update operating parameters; real-time collection of operating data and product quality data from the physical production line is fed back to the digital twin for online correction of the dual-physical-sensitivity model; and correction trigger thresholds are set. When the comprehensive quality score of 5 consecutive groups of products or difference deviation When this occurs, online model parameter correction is triggered, using incremental training with the learning rate adjusted to a fraction of the initial training learning rate. The number of iterations is equal to the number of initial training iterations. This ensures the sustainability of the optimization results.

[0040] This design, based on a digital twin, constructs a virtual debugging scenario, sets the debugging range and step size, runs dual models to record data, uses particle swarm optimization to solve for the optimal process variable values, distributes the optimized parameters to the production line, and provides real-time feedback data to correct the model. The virtual debugging scenario is safe and efficient. The particle swarm optimization algorithm can quickly find the optimal process parameters, improve production efficiency and product quality, and the real-time feedback data to correct the model ensures that the model always adapts to actual production changes, guarantees the continuity of optimization effects, enables the production line to operate stably in the best state for a long time, reduces production costs, and enhances the company's competitiveness.

[0041] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0042] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A virtual commissioning and operation optimization system for a corrugated paper production line based on digital twins, characterized in that, The system includes: a digital twin construction module, a multi-dimensional model construction and training module, a dominant variable identification module, and a virtual debugging and optimization module. The digital twin construction module acquires physical entity data and operating condition data of the production line to construct a digital twin consistent with the physical production line. The multi-dimensional model construction and training module synchronously constructs multi-physical sensitive dimension models for key processes of the production line within the digital twin, completing model training and parameter calibration. The dominant variable identification module calculates deviation quantification indicators based on the output states of the trained multi-physical sensitive dimension models, identifying target dominant process variables through these deviation quantification indicators. The virtual debugging and optimization module performs virtual debugging based on the target dominant process variables and optimizes production line operating parameters according to the debugging results, achieving production line operation optimization.

2. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins as described in claim 1, characterized in that, The multi-physical sensitivity dimension model includes: a macroscopic process model of the first physical sensitivity dimension and a constraint model of the second physical sensitivity dimension; the macroscopic process model generates a comprehensive response to both the properties of production raw materials and environmental disturbances; the constraint model is highly sensitive only to the key properties and core physical transmission behavior of production raw materials and is low in sensitivity to non-key interference factors.

3. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins as described in claim 1, characterized in that, The physical entity data acquired by the digital twin construction module includes: geometric parameters of production line equipment, equipment connection relationships, and key process boundary conditions; the acquired operating condition data includes raw material attribute data, real-time process parameters, and environmental parameters; the digital twin construction module is also used to preprocess the physical entity data and operating condition data, the preprocessing including data cleaning, outlier removal, and normalization; based on the preprocessed data, a digital twin is constructed using 3D modeling and physical modeling techniques to ensure that the digital twin achieves geometric, physical, and behavioral consistency with the physical production line.

4. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 2, characterized in that, The multi-dimensional model construction and training module includes the following steps: When constructing the macroscopic process model of the first physically sensitive dimension, macroscopic process equations for key processes are established based on the laws of conservation of mass, conservation of energy, and fluid mechanics. The inputs of the macroscopic process model include key attribute parameters of raw materials, environmental parameters, and core process parameters, and the outputs include product forming quality parameters and process energy consumption. When constructing the constraint model of the second physically sensitive dimension, a constraint model is established based on the core attribute transmission equation and core physical conduction equation of raw materials, retaining only the core physical mechanism and shielding the influence of non-critical interference factors.

5. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 2, characterized in that, The multi-dimensional model construction and training module includes the following steps during model training: data partitioning, initial parameter setting, loss function definition, optimizer configuration, iterative training, and parameter calibration. The data partitioning step divides the preprocessed operating data into training, validation, and test sets according to a preset ratio. The initial parameter setting step sets the initial parameters of the macroscopic process model and constraint model based on theoretical calculations and engineering experience. The loss function definition step uses mean squared error as the training loss function. The optimizer configuration step uses an adaptive optimizer and sets relevant parameters. The iterative training step sets the maximum number of training iterations and adopts an early stopping strategy. The parameter calibration step further optimizes the model parameters based on the trained model using a parameter optimization algorithm to ensure that the model prediction error is less than a preset threshold.

6. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 2, characterized in that, The dominant variable identification module, in a digital twin virtual debugging scenario, includes the following steps: synchronously running the trained macroscopic process model and constraint model, obtaining the output state vectors of the two models, calculating the deviation quantification index of the output state of the multi-physical sensitivity dimension model, wherein the deviation quantification index is calculated using a normalized distance algorithm to determine the degree of deviation; setting a deviation threshold, and determining that the target dominant process variable fluctuates under the current operating condition when the deviation quantification index exceeds the deviation threshold; determining the target dominant process variable through sensitivity analysis, calculating the influence coefficient of each process parameter on the deviation quantification index, and selecting a preset number of process parameters with the largest absolute value of the sensitivity coefficient as the target dominant process variable.

7. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 6, characterized in that: The deviation threshold is determined based on the dual-model output data during the stable operation phase of the production line; The stable operation phase of the production line refers to the operation phase during which there are no quality abnormalities or major fluctuations in operating conditions for a continuously preset period of time. Statistical characteristic parameters of the deviation from the quantitative indicators in this phase are calculated, and the deviation threshold is determined based on the statistical characteristic parameters.

8. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 2, characterized in that, The virtual debugging and optimization module constructs a virtual debugging scenario based on a digital twin, including the following steps: setting the debugging range and step size of the target dominant process variable, wherein the debugging range is determined based on the production line process specifications, and the step size does not exceed a preset proportion of the maximum value of the variable; under each debugging condition, running the trained multi-physical sensitivity dimension model, recording the deviation quantification index and product quality index; with the goal of minimizing the deviation quantification index and optimizing the product quality index, using an intelligent optimization algorithm to solve for the optimal value of the target dominant process variable.

9. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins as described in claim 8, characterized in that: The optimization objective function of the intelligent optimization algorithm is set with weight coefficients, which are set based on the priority requirements of the production line. When the production line prioritizes product quality, the weight coefficients related to product quality are higher than the weight coefficients related to process stability. When the production line prioritizes process stability, the weight coefficient related to process stability is higher than that related to product quality. When the priorities of quality and stability are balanced, a balanced weight coefficient is set. The product quality indicators include multiple core quality evaluation dimensions, and each quality evaluation dimension is assigned a corresponding weight coefficient to calculate a comprehensive product quality score.

10. The virtual debugging and operation optimization system for corrugated paper production lines based on digital twins according to claim 1, characterized in that: The virtual debugging and optimization module is also used to send the optimized process parameters to the physical production line to update the operating parameters; collect the operating data and product quality data of the physical production line in real time and feed them back to the digital twin to correct the multi-physical sensitive dimension model online; set a correction trigger threshold, and when the correction trigger condition is met, trigger the online correction of the model parameters. During the correction, an incremental training method is used to adjust the learning rate and the number of iterations to ensure the sustainability of the optimization effect.

Images