Concrete pole manufacturing process control method and system based on digital twinning

Abstract

This invention relates to the field of concrete pole manufacturing technology, and discloses a method and system for controlling the manufacturing process of concrete poles based on digital twins. The method includes: S1, real-time data acquisition and synchronization; S2, synchronous simulation and closed-loop control; S3, batch quality deviation calculation; S4, model parameter attribution correction; S5, model parameter evolution trend analysis; and S6, model parameter pre-compensation adjustment. This invention can effectively eliminate cognitive biases between the virtual model and the physical entity, avoid erroneous decisions caused by model inaccuracies, thereby ensuring the stability of pole product quality while improving production efficiency and energy utilization, and preventing human-machine control conflicts.

Description

Technical Field

[0001] This invention relates to the field of concrete pole manufacturing technology, and in particular to a method and system for controlling the manufacturing process of concrete poles based on digital twins. Background Technology

[0002] In modern industrial production, especially in the manufacturing of concrete utility poles, advanced control systems are increasingly incorporating digital twin technology to improve product quality and production efficiency. Digital twin technology constructs a virtual model in a computer that perfectly corresponds to the physical production line, used to simulate and predict the production process.

[0003] However, even if a virtual model is initially precisely calibrated based on standard conditions, it will deviate from the actual physical process over time, a phenomenon known as "model drift." There are two main reasons for this drift: first, while the sources of raw materials may meet national macro-standards, their microscopic characteristics contain subtle differences that are difficult for conventional sensors to detect; second, equipment undergoes imperceptible performance changes during continuous use. Once these small but persistent deviations accumulate, they cause the virtual model to become inaccurate. If the control system makes decisions based on this inaccurate model, it will not only affect product quality but may even lead to human-machine interface conflicts.

[0004] For example, when the virtual model incorrectly determines that the concrete strength has reached the demolding requirement ahead of schedule, the system will issue an instruction to prematurely end the steam curing process. The resulting poles, when tested for mechanical properties, often exhibit compressive strength and crack resistance lower than standard requirements, and may even show internal micro-cracks, posing a serious safety hazard. Conversely, if the hydration reaction rate of the actual concrete material slows down, and the system operates according to standard procedures, it will also result in energy waste and reduced production efficiency. The root of the problem lies in the fact that the once incredibly accurate digital twin model has gradually become "disconnected" from its physical entity due to various subtle changes in the real world. This virtual "twin" has failed to grow and change along with its real-world counterpart; its knowledge base has not been updated, leading to a lag and bias in its understanding of the physical world. Summary of the Invention

[0005] This invention provides a method and system for controlling the manufacturing process of concrete poles based on digital twins. It aims to solve the technical problem that existing digital twin control methods and systems, during long-term operation, cause deviations between the virtual model and the real physical process, i.e., "model drift," due to subtle changes in the microscopic properties of raw materials or the performance of equipment, which in turn affects product quality and may even lead to human-machine control conflicts.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: In a first aspect, the present invention proposes a method for controlling the manufacturing process of concrete poles based on digital twins, comprising the following steps: S1. Real-time data acquisition and synchronization: Real-time acquisition of process data at each key node of the production line, acquisition of multi-source process parameters, and real-time synchronization of the multi-source process parameters to the pre-constructed digital twin model; S2. Synchronous simulation and closed-loop control: The digital twin model synchronously simulates the state of concrete materials under the current working conditions based on the real-time input of the multi-source process parameters, and predicts the finished product quality indicators in real time based on the simulation results. Based on the finished product quality indicators, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. S3. Batch Quality Deviation Calculation: After a production batch is completed, the quality inspection indicators of the current batch of finished products are obtained, and these indicators are compared with the final finished product quality prediction indicators calculated by the digital twin model based on all process data and adjusted control parameters of the current batch. The total quality deviation is then calculated. ; S4. Model parameter attribution correction: When the total mass deviation is... When the parameters exceed a preset tolerance range, the key model parameters in the digital twin model are corrected according to a preset attribution logic, and a parameter adjustment log is recorded. The preset attribution logic calculates the impact coefficient affecting quality. and quality contribution Total mass deviation Attributable to different influencing factors ; S5. Model Parameter Evolution Trend Analysis: Periodically analyze the parameter adjustment logs. For each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. S6. Model Parameter Pre-compensation Adjustment: When a certain key model parameter is identified as having an upward or downward evolution trend, before the start of the next production batch, the corresponding key model parameter in the digital twin model is pre-compensated and adjusted according to the pre-compensation adjustment amount obtained from the average adjustment magnitude, or the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method, so as to update the initial state of the digital twin model before the start of the next production batch.

[0007] In a preferred embodiment of the present invention, step S4, the method for correcting the key model parameters in the digital twin model according to a preset attribution logic, specifically includes: S41, Benchmark Simulation: This simulates all preset influencing factors from the start of the current production batch. initial parameter values The data is input into the digital twin model for benchmark simulation to obtain benchmark simulation quality indicators. ; S42. Sensitivity Analysis: Targeting multiple pre-defined influencing factors. Each target impact factor Perform the following operations: The digital twin model corresponding to the target influence factor The parameter values, from the initial parameter values Adjusted to the first parameter value, the first parameter value and the initial parameter value The difference is a preset small step size ; Maintain the digital twin model excluding the target influencing factor Other influencing factors The parameter values ​​are consistent with the parameter values ​​during the benchmark simulation, and the digital twin model is run based on the first parameter values ​​to obtain the first disturbance simulation quality index. Based on the first disturbance simulation quality index and the benchmark simulation quality index First mass deviation between and the tiny step size Calculate the target impact factor Influence coefficient on finished product quality indicators That is, partial derivatives The influence coefficient The first quality deviation and the tiny step size The following conditions must be met: = ; S43. Obtain the actual deviation of the impact factor: Obtain the actual deviation of each target impact factor. Actual parameter values ​​in the current production batch And calculate relative to the initial parameter values. actual deviation = - ; S44. Calculate contribution and adjustment: Based on each target impact factor Influence coefficient and actual deviation Calculate the impact factor for each target. Mass contribution caused by individual action = ; For all target impact factors Quality contribution The absolute values ​​are normalized to obtain the target influence factors. Contribution weight = , For the j-th impact factor The quality contribution, among which Target Impact Factor The total number; Based on each target impact factor Contribution weight With respect to the total mass deviation Calculate the impact factor for each target. The amount of quality deviation to be borne = * ; Based on each target impact factor The amount of quality deviation to be borne Calculate the impact factor for each target. Parameter adjustment amount ; S45. Correct model parameters and record log: Adjust the amount according to the parameters. The parameters in the digital twin model are corrected, and a parameter adjustment log is recorded. The parameter adjustment log includes at least the parameter name, the value before adjustment, the value after adjustment, the adjustment range, the adjustment direction, the adjustment timestamp, and the batch identifier.

[0008] In a preferred embodiment of the present invention, in step S43, the actual parameter value is obtained. The methods specifically include: For influencing factors that can be directly measured Actual measured values ​​are obtained through online sensors or offline detection as actual parameter values. ; For influencing factors that cannot be directly measured The reverse attribution algorithm is used based on the total mass deviation. and the influence coefficient Solving for actual parameter values .

[0009] In a preferred embodiment of the present invention, the reverse attribution algorithm includes a fitting algorithm based on the least squares method, specifically: Constructing a linear regression model = ,in, For residuals, Target Impact Factor The total number; The least squares method with regularization is used to solve for the influencing factors that cannot be directly measured. actual deviation This minimizes the sum of the residuals' sum of squared values ​​and the regularization penalty term, thus yielding the actual parameter values. The estimated value.

[0010] In a preferred embodiment of the present invention, the method for determining the evolutionary tendency in step S5 is specifically as follows: If continuous M If the average adjustment magnitude calculated in each iteration is greater than the preset positive threshold, it is judged as an upward trend. If continuous M If the average adjustment magnitude calculated in each iteration is less than the preset negative threshold, it is judged as a downward trend. If continuous M If the average adjustment magnitude calculated in each iteration is neither greater than a preset positive threshold nor less than a preset negative threshold, it is judged as having no obvious tendency. M ≥3.

[0011] In a preferred embodiment of the present invention, the specific calculation of the pre-compensation adjustment amount in step S6 is as follows: The pre-compensation adjustment amount obtained based on the average adjustment range is specifically as follows: take the most recent... The average of the adjustment magnitudes is used as the pre-compensation adjustment amount, where... ; The pre-compensation adjustment amount obtained based on trend analysis is specifically: based on the most recent Historical data on the adjustment magnitude were used to fit a trend line using linear regression. The pre-compensation adjustment amount was obtained by multiplying the slope of the trend line by the time step. ; The pre-compensation adjustment amount predicted based on the exponential smoothing method is as follows: The prediction formula is: ,in, For the first Adjusted parameter values ​​for each batch For the first The predicted value for each batch, For the first The predicted value for each batch, The smoothing coefficient satisfies Pre-compensation adjustment amount = - Impact Factor The initial parameter values.

[0012] In a preferred embodiment of the present invention, the method further includes a sensitivity curve establishment step, specifically: Based on finite element analysis software or concrete hydration and mechanical property simulation software, by traversing each influencing factor Multiple preset values ​​were used and multiple simulations were performed to calculate the mass contribution corresponding to each preset value. Thus, the various influencing factors are established. The sensitivity curve of the impact on the quality of the finished product is used to verify the impact coefficient calculated in step S42. The linear approximation validity or the actual deviation When the value is outside the linear range, it can be used directly to query the quality contribution. .

[0013] As a preferred embodiment of the present invention, the influencing factor It includes at least one or more of the following: component activity parameters of raw materials, physical morphology parameters of raw materials, admixture performance parameters, process parameters, and environmental parameters. The finished product quality indicators include at least one or more of the following: compressive strength, crack resistance, and durability indicators of concrete poles.

[0014] As a preferred embodiment of the present invention, the process data of each key node of the production line includes at least one of raw material temperature, raw material humidity, concrete slump, centrifugal vibration data, and curing environment temperature and humidity. The closed-loop control includes at least one of adjusting the water-cement ratio of the mixer, optimizing the centrifuge speed-up curve, and changing the steam curing heating rate.

[0015] Secondly, this invention proposes a digital twin-based control system for the manufacturing process of concrete poles, the system comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the following modules: The data acquisition module is used to collect process data from key nodes of the production line in real time, obtain multi-source process parameters, and synchronize the multi-source process parameters to a pre-built digital twin model in real time. The simulation closed-loop control module is used to synchronously simulate the state of concrete materials under the current working conditions based on the real-time input multi-source process parameters of the digital twin model, and predict the finished product quality index in real time based on the simulation results. Based on the finished product quality index, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. The quality deviation calculation module is used to obtain the quality inspection indicators of the finished products of the current batch after the completion of a production batch, and compare the quality inspection indicators with the final finished product quality prediction indicators calculated by the digital twin model based on all process data of the current batch and the control parameter adjustments already performed, to calculate the total quality deviation. ; Attribution correction module, used when the total mass deviation When the preset allowable range is exceeded, the key model parameters in the digital twin model are corrected according to the preset attribution logic, and the parameter adjustment log is recorded. The evolutionary trend analysis module is used to periodically analyze the parameter adjustment logs, and for each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. The pre-compensation adjustment module is used to pre-compensate and adjust the key model parameters in the digital twin model before the start of the next production batch when a certain key model parameter is identified to have an upward or downward evolution trend. This adjustment is based on the pre-compensation adjustment amount obtained from the average adjustment magnitude, the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method. The result is used to update the initial state of the digital twin model before the start of the next production batch.

[0016] The beneficial effects of this invention are: This invention acquires multi-source process parameters by real-time collection of process data from key nodes of the production line and synchronizes them to a digital twin model. This enables synchronous simulation of the concrete material state and real-time prediction of finished product quality indicators, allowing for dynamic fine-tuning of production process control parameters and achieving closed-loop control of the production process. After a production batch is completed, this invention calculates the total quality deviation. Furthermore, when the deviation exceeds a preset range, the key model parameters in the digital twin model are corrected according to a preset attribution logic, and a parameter adjustment log is recorded. In addition, the present invention periodically analyzes the parameter adjustment log to determine the evolution trend of key model parameters, and when an evolution trend is identified, pre-compensation adjustments are made to the key model parameters in the digital twin model before the start of the next production batch to update the initial state of the model.

[0017] This invention overcomes the limitation of the model being "disconnected" from the physical entity by introducing batch quality deviation calculation, model parameter attribution correction, model parameter evolution trend analysis, and model parameter pre-compensation adjustment mechanisms. This enables the digital twin model to adaptively learn and correct itself based on actual production data. This dynamic correction and pre-compensation mechanism ensures that the virtual model maintains a high degree of consistency with the physical production process, avoiding erroneous control decisions caused by model inaccuracies. Therefore, this invention can significantly improve the control accuracy and product quality stability of the concrete pole manufacturing process, reduce the scrap rate, increase production efficiency, and effectively avoid control conflicts between humans and machines, providing reliable technical support for intelligent manufacturing. Attached Figure Description

[0018] Figure 1 The present invention provides a flowchart of a process control method for manufacturing concrete poles based on digital twins. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention proposes a method for controlling the manufacturing process of concrete poles based on digital twins, comprising the following steps: S1. Real-time data acquisition and synchronization: Real-time acquisition of process data at each key node of the production line, acquisition of multi-source process parameters, and real-time synchronization of the multi-source process parameters to the pre-constructed digital twin model; S2. Synchronous simulation and closed-loop control: The digital twin model synchronously simulates the state of concrete materials under the current working conditions based on the real-time input of the multi-source process parameters, and predicts the finished product quality indicators in real time based on the simulation results. Based on the finished product quality indicators, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. S3. Batch Quality Deviation Calculation: After a production batch is completed, the quality inspection indicators of the current batch of finished products are obtained, and these indicators are compared with the final finished product quality prediction indicators calculated by the digital twin model based on all process data and adjusted control parameters of the current batch. The total quality deviation is then calculated. ; S4. Model parameter attribution correction: When the total mass deviation is... When the parameters exceed a preset tolerance range, the key model parameters in the digital twin model are corrected according to a preset attribution logic, and a parameter adjustment log is recorded. The preset attribution logic calculates the impact coefficient affecting quality. and quality contribution Total mass deviation Attributable to different influencing factors ; S5. Model Parameter Evolution Trend Analysis: Periodically analyze the parameter adjustment logs. For each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. S6. Model Parameter Pre-compensation Adjustment: When a certain key model parameter is identified as having an upward or downward evolution trend, before the start of the next production batch, the corresponding key model parameter in the digital twin model is pre-compensated and adjusted according to the pre-compensation adjustment amount obtained from the average adjustment magnitude, or the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method, so as to update the initial state of the digital twin model before the start of the next production batch.

[0021] In its specific implementation, the present invention can be carried out according to the following steps: S1. Real-time data acquisition and synchronization: This step aims to provide real-time and comprehensive production line data for the digital twin model. Specifically, this can be achieved by deploying various sensors and detection devices, such as temperature sensors, humidity sensors, and vibration sensors, at key nodes in the concrete pole production line. These sensors can collect real-time process data, including the temperature and humidity of raw materials, the slump of concrete, centrifugal vibration data, and the temperature and humidity of the curing environment. This data is considered as multi-source process parameters.

[0022] S2. Synchronous Simulation and Closed-Loop Control: Upon receiving real-time, synchronized multi-source process parameters, the digital twin model synchronously simulates the state of concrete materials under the current operating conditions based on these parameters. For example, the model can simulate the hydration process, temperature field distribution, stress-strain state, and microstructural evolution within the concrete. Based on the simulation results, the model can predict finished product quality indicators in real time, such as the compressive strength, crack resistance, and durability of concrete poles. Subsequently, based on the deviation between the predicted finished product quality indicators and the preset target quality indicators, the control system dynamically fine-tunes the control parameters of the production process. For example, it can adjust the water-cement ratio of the mixer, optimize the centrifuge acceleration curve, and change the steam curing heating rate, thereby achieving closed-loop control of the production process. This dynamic fine-tuning can be implemented using a conventional proportional-integral-derivative (PID) controller.

[0023] S3. Batch quality deviation calculation: After a production batch is completed, the finished products of that batch need to undergo quality inspection to obtain actual quality indicators. These indicators can be obtained through laboratory testing, non-destructive testing, or on-site sampling inspection. Simultaneously, the digital twin model adjusts based on all process data and executed control parameters during the batch's production process to recalculate the final finished product quality prediction indicators. Subsequently, the actual quality inspection indicators are compared with the final finished product quality prediction indicators calculated by the model to calculate the total quality deviation. This bias reflects the overall accuracy of the digital twin model in predicting the current batch.

[0024] S4. Model parameter attribution correction: When the total mass deviation When the parameters exceed the preset tolerance range, it indicates that the digital twin model may exhibit "model drift," requiring correction of its key model parameters. This correction process is performed based on a preset attribution logic. The preset attribution logic calculates the impact coefficients affecting quality. and quality contribution Total mass deviation Attributable to different influencing factors For example, sensitivity analysis can be used to determine the impact of each influencing factor on the overall quality deviation. The degree of contribution. The corrected parameters will be recorded in the parameter tuning log for subsequent analysis and traceability.

[0025] S5. Model parameter evolution trend analysis: To proactively address model drift, the parameter adjustment log needs to be analyzed periodically. For each key model parameter, the system extracts the adjustment magnitude of its most recent T adjustments and calculates the average adjustment magnitude. Subsequently, based on the sign and magnitude of the average adjustment magnitude, the evolutionary trend of each key model parameter is determined. For example, if the average adjustment magnitude of a parameter is consistently positive and exceeds a certain threshold, it may indicate an upward evolutionary trend for that parameter; conversely, it may indicate a downward evolutionary trend. This analysis helps predict future trends in model parameters.

[0026] S6. Model parameter pre-compensation adjustment: When a critical model parameter is identified as exhibiting an upward or downward trend, pre-compensation adjustments need to be made to the critical model parameter in the digital twin model before the start of the next production batch to avoid significant quality deviations. The pre-compensation adjustment amount can be obtained in various ways. For example, the average of the most recent T adjustment magnitudes can be used directly; alternatively, a trend line can be fitted using methods such as linear regression based on historical data of the most recent T adjustment magnitudes, and the pre-compensation adjustment amount can be obtained by multiplying the slope of the trend line by the time step; or, a predictive model such as exponential smoothing can be used to predict the parameter values ​​for the next batch, and the difference between the predicted and initial parameter values ​​can be used as the pre-compensation adjustment amount. Through pre-compensation adjustments, the initial state of the digital twin model before the start of the next production batch can be updated, making it closer to the actual physical process, thereby improving the control accuracy and product quality of subsequent production batches.

[0027] The overall working principle of this invention lies in constructing an adaptive, self-learning digital twin control closed loop. First, through real-time data acquisition and synchronization, the real-time state of the physical production line is mapped onto the digital twin model, ensuring synchronization between the virtual and real worlds. Next, the digital twin model performs synchronous simulation and closed-loop control based on real-time data, enabling refined management of the production process, real-time prediction of finished product quality, and dynamic fine-tuning of control parameters to achieve optimal production conditions. Model deviation is quantified by comparing actual quality inspection indicators with model-predicted quality indicators. When the deviation exceeds the allowable range, the model parameter attribution correction step is triggered, using preset attribution logic to adjust the total quality deviation. The system accurately attributes problems to specific influencing factors and adjusts key model parameters accordingly. This adjustment process not only corrects issues in the current batch but, more importantly, provides a valuable data foundation for subsequent analysis of model parameter evolution trends through parameter adjustment logs. Analysis of historical parameter adjustment logs identifies long-term trends in key model parameters, enabling the prediction of future drift directions and magnitudes. Finally, using these trend analysis results, key model parameters in the digital twin model are proactively adjusted before the start of a new production batch. This pre-compensation adjustment mechanism allows the digital twin model to actively adapt to long-term changes in the physical production process, rather than passively correcting deviations after they occur.

[0028] Through the close coordination of the above steps, this invention forms a continuously optimizing and self-evolving control process. It can not only respond in real time to short-term fluctuations in the production process, but also effectively suppress model drift by analyzing and pre-compensating for the long-term evolution of model parameters, ensuring that the digital twin model always maintains a high degree of consistency with the physical entity.

[0029] Specifically, the attribution correction mechanism of this invention can accurately locate the key influencing factors causing model deviations, avoiding blind adjustments to model parameters and improving the efficiency and accuracy of correction. The model parameter evolution trend analysis and pre-compensation adjustment mechanism give this invention a forward-looking capability, enabling predictive adjustments before significant drift in model parameters occurs, thus elevating control from passive response to proactive prevention. For example, in the production of concrete poles, if the activity parameter of a certain raw material shows a long-term downward trend, this invention can identify this trend in advance and pre-compensate the corresponding parameters in the digital twin model before the next batch of production, thereby avoiding insufficient concrete strength due to decreased raw material activity. This proactive adjustment method is significantly superior to the traditional strategy of passively correcting only after product quality problems occur, greatly improving the stability of the production process and the consistency of product quality. Therefore, this invention demonstrates significant technological progress and practical value in solving the "model drift" problem in the long-term operation of digital twin models.

[0030] Step S4, the method for correcting the key model parameters in the digital twin model according to the preset attribution logic, specifically includes: S41, Benchmark Simulation: This simulates all preset influencing factors from the start of the current production batch. initial parameter values The data is input into the digital twin model for benchmark simulation to obtain benchmark simulation quality indicators. ; S42. Sensitivity Analysis: Targeting multiple pre-defined influencing factors. Each target impact factor Perform the following operations: The digital twin model corresponding to the target influence factor The parameter values, from the initial parameter values Adjusted to the first parameter value, the first parameter value and the initial parameter value The difference is a preset small step size ; Maintain the digital twin model excluding the target influencing factor Other influencing factors The parameter values ​​are consistent with the parameter values ​​during the benchmark simulation, and the digital twin model is run based on the first parameter values ​​to obtain the first disturbance simulation quality index. Based on the first disturbance simulation quality index and the benchmark simulation quality index First mass deviation between and the tiny step size Calculate the target impact factor Influence coefficient on finished product quality indicators That is, partial derivatives The influence coefficient The first quality deviation and the tiny step size The following conditions must be met: = ; S43. Obtain the actual deviation of the impact factor: Obtain the actual deviation of each target impact factor. Actual parameter values ​​in the current production batch And calculate relative to the initial parameter values. actual deviation = - ; S44. Calculate contribution and adjustment: Based on each target impact factor Influence coefficient and actual deviation Calculate the impact factor for each target. Mass contribution caused by individual action = ; For all target impact factors Quality contribution The absolute values ​​are normalized to obtain the target influence factors. Contribution weight = , For the j-th impact factor The quality contribution, among which Target Impact Factor The total number; Based on each target impact factor Contribution weight With respect to the total mass deviation Calculate the impact factor for each target. The amount of quality deviation to be borne = * ; Based on each target impact factor The amount of quality deviation to be borne Calculate the impact factor for each target. Parameter adjustment amount ; S45. Correct model parameters and record log: Adjust the amount according to the parameters. The parameters in the digital twin model are corrected, and a parameter adjustment log is recorded. The parameter adjustment log includes at least the parameter name, the value before adjustment, the value after adjustment, the adjustment range, the adjustment direction, the adjustment timestamp, and the batch identifier.

[0031] This invention introduces a series of structured steps to accurately attribute and correct key model parameters in a digital twin model. This data-driven and model simulation-based attribution correction mechanism significantly improves the fitting accuracy and prediction accuracy of the digital twin model to the actual production process, thereby enhancing the robustness and adaptability of the entire concrete pole manufacturing process control. First, an ideal reference state is established through benchmark simulation, providing a baseline for subsequent deviation analysis. Second, sensitivity analysis quantifies the independent impact of each influencing factor on the finished product quality, providing a foundation for understanding and quantifying the contribution of each factor. Next, by obtaining the actual deviations of the influencing factors, theoretical analysis is combined with actual production data. Finally, the quality contribution degree, contribution weight, and the amount of quality deviation to be borne by each influencing factor are calculated. This allows for precise determination of the adjustment direction and magnitude of each key model parameter. This method, based on influence coefficients... and actual deviation The attribution logic leads to total quality deviation It can be reasonably decomposed into various influencing factors. This allows for precise correction of the digital twin model's parameters. By recording detailed parameter adjustment logs, the transparency and traceability of the correction process are ensured, laying the foundation for subsequent model optimization and trend analysis.

[0032] In step S43, the actual parameter values ​​are obtained. The methods specifically include: For influencing factors that can be directly measured Actual measured values ​​are obtained through online sensors or offline detection as actual parameter values. ; For influencing factors that cannot be directly measured The reverse attribution algorithm is used based on the total mass deviation. and the influence coefficient Solving for actual parameter values .

[0033] Specifically, impact factor Based on their measurability, these factors can be divided into two categories. The first category consists of parameters that can be directly measured in real time using existing technologies, such as the temperature and humidity of raw materials and the slump of concrete. For this type of influencing factor... Its actual parameter value This can be achieved through continuous monitoring using online sensors deployed on the production line or through periodic offline testing (e.g., laboratory sampling). These measurements are directly adopted as the influencing factor. The first category consists of actual parameter values ​​in the current batch. The second category includes parameters whose precise values ​​cannot be obtained directly through physical sensors or conventional testing methods, such as the activity of certain raw material components or the actual effectiveness of additives. For this type of influencing factor... This invention proposes using a reverse attribution algorithm to infer the actual parameter values. This algorithm utilizes the known total mass deviation. and the influence coefficient obtained through sensitivity analysis These influencing factors, which are difficult to measure directly, are derived by reverse engineering using mathematical models. Actual parameter values .

[0034] This invention distinguishes different types of influencing factors The measurability is assessed, and targeted direct measurement or reverse attribution algorithms are employed to obtain actual parameter values. This ensures that all key influencing factors are considered during the model parameter attribution correction step. The actual state can be accurately reflected. Direct measurement ensures the real-time performance and accuracy of measurable parameters, while the reverse attribution algorithm solves the problem of missing data for unmeasurable parameters, enabling effective correction of model parameters even with incomplete information. It is precisely because of the ability to obtain more comprehensive and accurate actual parameter values ​​that... This is what makes the subsequent quality contribution... Calculation and parameter adjustment amount The calculations are grounded in solid data, thus improving the accuracy of attribution correction. Consequently, during the attribution correction of model parameters, the factors causing overall quality deviations can be identified more accurately. The root cause, and calculate a more reasonable parameter adjustment amount. This significantly improves the effectiveness of digital twin model correction and the accuracy of production process control, ultimately helping to steadily improve the finished quality of concrete poles.

[0035] The reverse attribution algorithm includes a least squares-based fitting algorithm, specifically: Constructing a linear regression model = ,in, For residuals, Target Impact Factor The total number; The least squares method with regularization is used to solve for the influencing factors that cannot be directly measured. actual deviation This minimizes the sum of the residuals' sum of squared values ​​and the regularization penalty term, thus yielding the actual parameter values. The estimated value.

[0036] The linear regression model is a statistical model used to establish the dependent variable (in this case, total quality deviation). ) and one or more independent variables (here, the various influencing factors) actual deviation The linear relationship between the residuals. This indicates the difference between the model's predicted values ​​and the actual observed values, reflecting the portion that the model failed to explain. This indicates the target influencing factor considered in the model. The total number. Regularized least squares is an optimization method that aims to minimize the sum of squared residuals while introducing a regularization penalty term to avoid overfitting and improve the model's generalization ability and stability. Constraining the model using norms such as L1 norm (Lasso regression) or L2 norm (Ridge regression) prevents excessively large parameters or overly complex models. Minimizing the sum of squared residuals and the regularization penalty term allows for more robust estimation of influencing factors that cannot be directly measured. actual deviation This allows us to obtain the corresponding actual parameter values. The estimated value.

[0037] This invention transforms the inverse attribution problem into a linear regression problem and solves it using regularized least squares, thereby effectively addressing the total mass deviation. Inversely, we can deduce the various influencing factors that cannot be directly measured. actual deviation Specifically, the linear regression model establishes the total quality deviation. With each influencing factor actual deviation The quantitative relationship between them, including the influence coefficient As a bridge, connecting various influencing factors The changes are correlated with quality deviations. Using regularized least squares not only minimizes the error between model predictions and actual observations, but also effectively suppresses model errors caused by noise or influencing factors in the data by introducing a regularization penalty term. This addresses the overfitting problem that may arise from collinearity, enhancing the model's predictive ability for unknown data and the stability of parameter estimation. Therefore, even with incomplete or uncertain data, it is possible to obtain information on intangible influencing factors. Actual parameter values A reliable estimate.

[0038] In step S5, the method for determining evolutionary tendency is as follows: If continuous M If the average adjustment magnitude calculated in each iteration is greater than the preset positive threshold, it is judged as an upward trend. If continuous M If the average adjustment magnitude calculated in each iteration is less than the preset negative threshold, it is judged as a downward trend. If continuous M If the average adjustment magnitude calculated in each iteration is neither greater than a preset positive threshold nor less than a preset negative threshold, it is judged as having no obvious tendency. M ≥3.

[0039] Among them, "continuous" M "Average adjustment magnitude calculated in this cycle" refers to the average adjustment magnitude extracted from the most recent values ​​of a key model parameter during the analysis of the parameter adjustment log. M batch or M The average adjustment magnitude data over a given time period. Here... M It is an integer, and M A value greater than or equal to 3 is used to ensure the stability and reliability of the judgment. For example, M It can be set to 3, 5 or 10. The specific value can be adjusted according to the fluctuation of historical data and the sensitivity requirements for trend judgment. The positive threshold and negative threshold can be reasonably set according to actual production experience, historical data analysis or expert knowledge to balance sensitivity and false alarm rate.

[0040] This invention sets a continuous M The threshold judgment mechanism for the average adjustment magnitude can effectively filter out occasional or random parameter fluctuations, thereby identifying the true evolution trend of key model parameters. When the average adjustment magnitude is continuous... M When the value exceeds the preset positive threshold multiple times, it indicates a continuous upward trend in the key model parameters. This is then judged as an upward tendency, providing a basis for subsequent pre-compensation adjustments. Conversely, when the average adjustment magnitude continuously... MWhen the value is less than the preset negative threshold, it indicates that the key model parameter has a continuous downward trend, which is judged as a downward tendency. This judgment method based on multiple consecutive observations enhances the robustness of trend judgment, can more accurately identify the long-term evolution trend of key model parameters, and avoids erroneous judgments caused by single or a few abnormal fluctuations. This makes the pre-compensation adjustment of model parameters more accurate and timely, thereby further improving the prediction accuracy and control effect of the digital twin model on the actual production process, and ultimately helping to stabilize and optimize the finished quality of concrete poles.

[0041] In response, this invention further proposes various methods for calculating the pre-compensation adjustment amount, which can accurately calculate the pre-compensation adjustment amount to improve the accuracy and robustness of the pre-compensation adjustment and effectively cope with the long-term evolution trend of model parameters.

[0042] In step S6, the specific calculation of the pre-compensation adjustment amount is as follows: The pre-compensation adjustment amount obtained based on the average adjustment range is specifically as follows: take the most recent... The average of the adjustment magnitudes is used as the pre-compensation adjustment amount, where... ; The pre-compensation adjustment amount obtained based on trend analysis is specifically: based on the most recent Historical data on the adjustment magnitude were used to fit a trend line using linear regression. The pre-compensation adjustment amount was obtained by multiplying the slope of the trend line by the time step. ; The pre-compensation adjustment amount predicted based on the exponential smoothing method is as follows: The prediction formula is: ,in, For the first Adjusted parameter values ​​for each batch For the first The predicted value for each batch, For the first The predicted value for each batch, The smoothing coefficient satisfies Pre-compensation adjustment amount = - Impact Factor The initial parameter values.

[0043] Specifically, when using the average adjustment magnitude to determine the pre-compensation adjustment amount, the core idea is that the future trend of model parameters is correlated with the average level of their recent historical adjustments. This is achieved by analyzing the recent... A simple averaging of these adjustments yields a relatively stable adjustment amount, used to offset persistent parameter drift. The value of is not less than 4, which is intended to ensure that there is enough historical data to calculate a representative average and to avoid bias caused by a few abnormal adjustments.

[0044] Among them, the pre-compensation adjustment amount obtained based on trend analysis aims to identify the long-term evolution trend of parameters by conducting deeper analysis of historical adjustment data. Specifically, by analyzing recent... By performing linear regression fitting on historical data of each adjustment magnitude, a trend line can be obtained. The slope of this trend line reflects the rate and direction of parameter change over time. Multiplying this slope by a time step allows prediction of the potential parameter adjustment before the start of the next production batch. This method better captures gradual changes in parameters and provides more proactive pre-compensation. Similarly, The value of is not less than 4 to ensure the effectiveness and stability of trend fitting.

[0045] In practical applications, the pre-compensation adjustment based on exponential smoothing is a more dynamic and adaptive forecasting method. Exponential smoothing assigns higher weights to recent data, allowing the forecast results to respond more quickly to the latest changes in parameters. (Forecasting formula) middle, Indicates the first Adjusted parameter values ​​for each batch Indicates the first The predicted value for each batch, Indicates the first +1 batch of predicted values, This is the smoothing coefficient, and its value ranges from 0 to 1. The degree to which new data affects the prediction results is determined. The larger the value, the greater the influence of recent data, and the faster the prediction results respond to the latest changes. In this way, the prediction parameter values ​​for the next batch can be obtained and compared with the influence factor. initial parameter values The comparisons are then made to calculate the pre-compensation adjustment. This method can provide relatively accurate predictions when parameter changes have a certain degree of randomness but also exhibit a potential trend.

[0046] This invention effectively solves the problems of unclear calculation methods and insufficient accuracy in pre-compensation adjustment by providing three different methods for calculating pre-compensation adjustment. Specifically, when a tendency for key model parameters to rise or fall is identified, adjustments are no longer made simply, but rather quantitative predictions are made using statistical or time series analysis methods based on historical adjustment data of the parameters. The average adjustment magnitude method provides a robust baseline adjustment by smoothing historical fluctuations; the trend analysis method captures the long-term drift direction and rate of parameters through linear regression, achieving predictive compensation for future changes; and the exponential smoothing method assigns higher weight to recent data, enabling pre-compensation adjustments to respond more sensitively to the latest dynamics of the parameters. These methods all aim to accurately pre-adjust key model parameters in the digital twin model before the start of the next production batch based on the evolution trend of the parameters, so that the model is in an optimized state that is closer to actual working conditions when the new production batch begins.

[0047] The above technical solutions can significantly improve the accuracy of the digital twin model's initial state before the start of the next production batch. By accurately calculating the pre-compensation adjustment, long-term drift or systematic deviations in model parameters can be effectively offset, thereby reducing quality deviations caused by model inaccuracies during production. This not only improves the prediction accuracy and simulation realism of the digital twin model for the actual production process, but also makes subsequent synchronous simulation and closed-loop control more stable and efficient. Ultimately, this helps improve the finished quality of concrete poles, reduce scrap rates and rework costs, and achieve more refined production process management.

[0048] In some preferred embodiments, specific examples are given below. Assume a key model parameter, such as the influencing factor of the water-cement ratio in concrete. In the most recent five production batches, the model parameters were adjusted by +0.01, +0.008, +0.012, +0.009, and +0.011, respectively.

[0049] If the average adjustment range method is used, and If the value is 4, then the average of the four most recent adjustment magnitudes is taken as the pre-compensation adjustment amount. For example, the average of the four most recent adjustment magnitudes (+0.008, +0.012, +0.009, +0.011) is (0.008+0.012+0.009+0.011) / 4 = 0.01. Therefore, before the next production batch begins, this influencing factor... The model parameters will be pre-compensated and adjusted by +0.01.

[0050] If a trend analysis-based method is used, and If the trend line has a slope of 0.0005 (meaning an average increase of 0.0005 per adjustment) and a time step of 1 (representing the transition from the current batch to the next), then the pre-compensation adjustment can be calculated as the slope multiplied by the time step: 0.0005 * 1 = 0.0005. Furthermore, the intercept of the trend line or the last data point needs to be considered to obtain a more complete prediction. For example, if the trend line predicts an adjustment of 0.0115 for the next point, then the pre-compensation adjustment is 0.0115.

[0051] If exponential smoothing is used for forecasting, assuming a smoothing coefficient... =0.3. If the first The adjusted parameter values ​​for each batch are , No. The predicted value for each batch is So, what is the predicted value for the next batch? According to the formula Perform the calculations. Assume the influence factor. initial parameter values The value is 0.45, which is obtained through exponential smoothing. The value is 0.462. Therefore, the pre-compensation adjustment amount = - = 0.462 0.45 = 0.012.

[0052] The pre-compensation adjustment amount calculated by any of the above methods will be used to update the initial state of the digital twin model before the start of the next production batch, thereby improving the model's prediction accuracy.

[0053] In some of the above embodiments, the model parameter attribution correction step S42 calculates the influence coefficient. To quantify each influencing factor The impact on finished product quality. However, this is based on small step sizes. The partial derivative approximation method is essentially a linear approximation. When the influence factor... When the relationship between the product quality indicators and the actual deviation exhibits a significant non-linear characteristic, or when the actual deviation is significant... When the value is too large and exceeds the effective range of linear approximation, then... = Calculated quality contribution Significant errors may exist, affecting the accuracy of model parameter correction. Failure to address these issues could lead to inaccurate model parameter correction, or even introduce new biases, reducing the digital twin model's ability to predict and control actual production processes.

[0054] In response, this invention further proposes a sensitivity curve establishment step to improve the accuracy and applicability of model parameter attribution correction.

[0055] The method also includes a sensitivity curve establishment step, specifically: Based on finite element analysis software or concrete hydration and mechanical property simulation software, by traversing each influencing factor Multiple preset values ​​were used and multiple simulations were performed to calculate the mass contribution corresponding to each preset value. Thus, the various influencing factors are established. The sensitivity curve of the impact on the quality of the finished product is used to verify the impact coefficient calculated in step S42. The linear approximation validity or the actual deviation When the value is outside the linear range, it can be used directly to query the quality contribution. .

[0056] Specifically, the sensitivity curve establishment step refers to an offline analysis process performed before actual production or periodically. This process utilizes specialized simulation tools, such as finite element analysis software or simulation software specifically designed for concrete material properties (e.g., hydration processes, mechanical property development), to conduct in-depth parametric sensitivity analysis on the digital twin model. This includes iterating through each influencing factor... Multiple preset values ​​and multiple simulations refer to the process of setting multiple preset values ​​for each key influencing factor and performing multiple simulations. For example, the component activity parameters of raw materials, the efficacy parameters of admixtures, etc., while maintaining other influencing factors Without changing the influencing factor, systematically change it. The range of values ​​for the influencing factors is determined, and digital twin model simulations are performed for each value. Through these simulations, the results can be obtained under different influencing factors. The changes in finished product quality indicators under various preset values ​​are analyzed, and the quality contribution corresponding to each preset value is calculated. Based on the simulation results, each influencing factor is quantified. The actual impact of specific values ​​on the quality indicators of the finished product is determined by these data points (influence factor values, quality contribution values), which are used to plot or fit the values ​​of each influence factor. The sensitivity curve for the impact on finished product quality, in practical applications, can be understood as describing a specific influencing factor. The function or graph representing the relationship between changes in the value of a factor and changes in the quality indicators of the finished product aims to provide a non-linear and more comprehensive mapping of the influence factor-quality relationship. Furthermore, the sensitivity curve serves two main purposes: firstly, it can be used to verify the influence coefficients calculated in step S42 above. The linear approximation effectiveness, by... With sensitivity curve at initial parameter values By comparing the slopes of nearby slopes, the accuracy of the linear approximation can be assessed. On the other hand, when the actual deviation... When it exceeds the linear range, that is, when the influence factor Actual value and initial parameter value When the differences are significant enough that a linear approximation is no longer applicable, a more accurate quality contribution can be obtained by directly querying the sensitivity curve. This avoids the errors caused by linear approximation.

[0057] This invention effectively addresses the limitations of traditional linear approximation methods in handling nonlinear relationships or large deviations by introducing a sensitivity curve establishment step. Specifically, by conducting a comprehensive sensitivity analysis of the digital twin model in an offline environment using professional simulation software, the influencing factors can be pre-established. A more precise non-linear mapping relationship between finished product quality indicators and actual production processes. If the influence coefficient calculated through step S42 is... The linear approximation may no longer be effective or when the influence factor... actual deviation When the value is large, a more accurate quality contribution can be obtained by directly using a pre-established sensitivity curve. This approach avoids the errors caused by forcibly using linear approximations in nonlinear regions, thus making the attribution correction of model parameters more accurate and ensuring that the influencing factors are accurately adjusted under various operating conditions. Quality contribution Accurate assessments prevent misadjustment of control parameters due to attribution errors, thereby improving the effectiveness of closed-loop control in the production process and ultimately helping to stabilize and improve the finished quality of concrete poles.

[0058] In some preferred embodiments, specific examples are given below. It is assumed that the hydration heat release rate of cement is a key influencing factor during the manufacture of concrete poles. It has a significant impact on the early strength and crack resistance of concrete.

[0059] First, in the sensitivity curve establishment step, a series of simulations can be performed on the digital twin model using concrete hydration and mechanical property simulation software. For example, starting with a baseline value (e.g., 20 J / g / h), the cement hydration heat release rate can be adjusted in increments (e.g., ±2 J / g / h), and the changes in early-age strength and crack resistance of concrete under different values ​​can be simulated. Through multiple simulations, a series of data points can be obtained. Based on these data, a sensitivity curve of the effect of cement hydration heat release rate on early-age strength and crack resistance can be plotted. This curve may show approximately linearity within a certain range, but exhibit nonlinearity at extremely high or low values.

[0060] If a total quality deviation is found after a production batch is completed. The influence coefficient of the cement hydration heat release rate calculated in step S42 is relatively large. This shows its deviation from the actual value. The product of (linear approximation of quality contribution) and the total quality deviation Inconsistent, or actual deviation If the actual hydration heat release rate exceeds the effective range of the linear approximation (e.g., the difference between the actual hydration heat release rate and the benchmark value exceeds 10 J / g / h), then a pre-established sensitivity curve can be directly consulted. For example, if the actual hydration heat release rate is 28 J / g / h, the actual contribution of this value to early strength and crack resistance can be directly found from the sensitivity curve. Instead of simply using = An estimation is performed. In this way, the impact of the cement hydration heat release rate on the total mass deviation can be determined more accurately. This provides a practical contribution, thereby guiding the precise correction of corresponding parameters in the digital twin model.

[0061] The influencing factors It includes at least one or more of the following: component activity parameters of raw materials, physical morphology parameters of raw materials, admixture performance parameters, process parameters, and environmental parameters. The finished product quality indicators include at least one or more of the following: compressive strength, crack resistance, and durability indicators of concrete poles.

[0062] The process data for each key node of the production line includes at least one of the following: raw material temperature, raw material humidity, concrete slump, centrifugal vibration data, and temperature and humidity of the curing environment. The closed-loop control includes at least one of the following: adjusting the water-cement ratio of the mixer, optimizing the centrifuge speed-up curve, and changing the steam curing heating rate.

[0063] This invention provides a digital twin-based control system for the manufacturing process of concrete poles, the system comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the following modules: The data acquisition module is used to collect process data from key nodes of the production line in real time, obtain multi-source process parameters, and synchronize the multi-source process parameters to a pre-built digital twin model in real time. The simulation closed-loop control module is used to synchronously simulate the state of concrete materials under the current working conditions based on the real-time input multi-source process parameters of the digital twin model, and predict the finished product quality index in real time based on the simulation results. Based on the finished product quality index, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. The quality deviation calculation module is used to obtain the quality inspection indicators of the finished products of the current batch after the completion of a production batch, and compare the quality inspection indicators with the final finished product quality prediction indicators calculated by the digital twin model based on all process data of the current batch and the control parameter adjustments already performed, to calculate the total quality deviation. ; Attribution correction module, used when the total mass deviation When the preset allowable range is exceeded, the key model parameters in the digital twin model are corrected according to the preset attribution logic, and the parameter adjustment log is recorded. The evolutionary trend analysis module is used to periodically analyze the parameter adjustment logs, and for each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. The pre-compensation adjustment module is used to pre-compensate and adjust the key model parameters in the digital twin model before the start of the next production batch when a certain key model parameter is identified to have an upward or downward evolution trend. This adjustment is based on the pre-compensation adjustment amount obtained from the average adjustment magnitude, the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method. The result is used to update the initial state of the digital twin model before the start of the next production batch.

[0064] The above description is only a part or preferred embodiment of the present invention. Neither the text nor the drawings should limit the scope of protection of the present invention. All equivalent structural transformations made using the content of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect applications in other related technical fields, are included within the scope of protection of the present invention.

Claims

1. A method for controlling the manufacturing process of concrete utility poles based on digital twins, characterized in that, Includes the following steps: S1. Real-time data acquisition and synchronization: Real-time acquisition of process data at each key node of the production line, acquisition of multi-source process parameters, and real-time synchronization of the multi-source process parameters to the pre-constructed digital twin model; S2. Synchronous simulation and closed-loop control: The digital twin model synchronously simulates the state of concrete materials under the current working conditions based on the real-time input of the multi-source process parameters, and predicts the finished product quality indicators in real time based on the simulation results. Based on the finished product quality indicators, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. S3. Batch Quality Deviation Calculation: After a production batch is completed, the quality inspection indicators of the current batch of finished products are obtained, and these indicators are compared with the final finished product quality prediction indicators calculated by the digital twin model based on all process data and adjusted control parameters of the current batch. The total quality deviation is then calculated. ; S4. Model parameter attribution correction: When the total mass deviation is... When the parameters exceed a preset tolerance range, the key model parameters in the digital twin model are corrected according to a preset attribution logic, and a parameter adjustment log is recorded. The preset attribution logic calculates the impact coefficient affecting quality. and quality contribution Total mass deviation Attributable to different influencing factors ; S5. Model Parameter Evolution Trend Analysis: Periodically analyze the parameter adjustment logs. For each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. S6. Model Parameter Pre-compensation Adjustment: When a key model parameter is identified as having an upward or downward evolution trend, before the start of the next production batch, the corresponding key model parameter in the digital twin model is pre-compensated and adjusted according to the pre-compensation adjustment amount obtained from the average adjustment magnitude, or the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method, so as to update the initial state of the digital twin model before the start of the next production batch.

2. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 1, characterized in that, Step S4, the method for correcting the key model parameters in the digital twin model according to the preset attribution logic, specifically includes: S41, Benchmark Simulation: This simulates all preset influencing factors from the start of the current production batch. initial parameter values The data is input into the digital twin model for benchmark simulation to obtain benchmark simulation quality indicators. ; S42. Sensitivity Analysis: Targeting multiple pre-defined influencing factors. Each target impact factor Perform the following operations: The digital twin model corresponding to the target influence factor The parameter values, from the initial parameter values Adjusted to the first parameter value, the first parameter value and the initial parameter value The difference is a preset small step size ; Maintain the digital twin model excluding the target influencing factor Other influencing factors The parameter values ​​are consistent with the parameter values ​​during the benchmark simulation, and the digital twin model is run based on the first parameter values ​​to obtain the first disturbance simulation quality index. Based on the first disturbance simulation quality index and the benchmark simulation quality index First mass deviation between and the tiny step size Calculate the target impact factor Influence coefficient on finished product quality indicators That is, partial derivatives The influence coefficient The first quality deviation and the tiny step size The following conditions must be met: = ; S43. Obtain the actual deviation of the impact factor: Obtain the actual deviation of each target impact factor. Actual parameter values ​​in the current production batch And calculate relative to the initial parameter values. actual deviation = - ; S44. Calculate contribution and adjustment: Based on each target impact factor Influence coefficient and actual deviation Calculate the impact factor for each target. Mass contribution caused by individual action = ; For all target impact factors Quality contribution The absolute values ​​are normalized to obtain the target influence factors. Contribution weight = , For the j-th impact factor The quality contribution, among which Target Impact Factor The total number; Based on each target impact factor Contribution weight With respect to the total mass deviation Calculate the impact factor for each target. The amount of quality deviation to be borne = * ; Based on each target impact factor The amount of quality deviation to be borne Calculate the impact factor for each target. Parameter adjustment amount ; S45. Correct model parameters and record log: Adjust the amount according to the parameters. The parameters in the digital twin model are corrected, and a parameter adjustment log is recorded. The parameter adjustment log includes at least the parameter name, the value before adjustment, the value after adjustment, the adjustment range, the adjustment direction, the adjustment timestamp, and the batch identifier.

3. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 2, characterized in that, In step S43, the actual parameter values ​​are obtained. The methods specifically include: For influencing factors that can be directly measured Actual measured values ​​are obtained through online sensors or offline detection as actual parameter values. ; For influencing factors that cannot be directly measured The reverse attribution algorithm is used based on the total mass deviation. and the influence coefficient Solving for actual parameter values .

4. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 3, characterized in that, The reverse attribution algorithm includes a least squares-based fitting algorithm, specifically: Constructing a linear regression model = ,in, For residuals, Target Impact Factor The total number; The least squares method with regularization is used to solve for the influencing factors that cannot be directly measured. actual deviation This minimizes the sum of the residuals' sum of squared values ​​and the regularization penalty term, thus yielding the actual parameter values. The estimated value.

5. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 1, characterized in that, In step S5, the method for determining evolutionary tendency is as follows: If continuous M If the average adjustment magnitude calculated in each iteration is greater than the preset positive threshold, it is judged as an upward trend. If continuous M If the average adjustment magnitude calculated in each iteration is less than the preset negative threshold, it is judged as a downward trend. If continuous M If the average adjustment magnitude calculated in each iteration is neither greater than a preset positive threshold nor less than a preset negative threshold, it is judged as having no obvious tendency. M ≥3.

6. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 1, characterized in that, In step S6, the specific calculation of the pre-compensation adjustment amount is as follows: The pre-compensation adjustment amount obtained based on the average adjustment range is specifically as follows: take the most recent... The average of the adjustment magnitudes is used as the pre-compensation adjustment amount, where... ; The pre-compensation adjustment amount obtained based on trend analysis is specifically: based on the most recent Historical data on the adjustment magnitude were used to fit a trend line using linear regression. The pre-compensation adjustment amount was obtained by multiplying the slope of the trend line by the time step. ; The pre-compensation adjustment amount predicted based on the exponential smoothing method is as follows: The prediction formula is: ,in, For the first Adjusted parameter values ​​for each batch For the first The predicted value for each batch, For the first The predicted value for each batch, The smoothing coefficient satisfies Pre-compensation adjustment amount = - Impact Factor The initial parameter values.

7. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 2, characterized in that, The method also includes a sensitivity curve establishment step, specifically: Based on finite element analysis software or concrete hydration and mechanical property simulation software, by traversing each influencing factor Multiple preset values ​​were used and multiple simulations were performed to calculate the mass contribution corresponding to each preset value. Thus, the various influencing factors are established. The sensitivity curve of the impact on the quality of the finished product is used to verify the impact coefficient calculated in step S42. The linear approximation validity or the actual deviation When the value is outside the linear range, it can be used directly to query the quality contribution. .

8. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 1, characterized in that, The influencing factors It includes at least one or more of the following: component activity parameters of raw materials, physical morphology parameters of raw materials, admixture performance parameters, process parameters, and environmental parameters. The finished product quality indicators include at least one or more of the following: compressive strength, crack resistance, and durability indicators of concrete poles.

9. The method for controlling the manufacturing process of concrete poles based on digital twins according to claim 1, characterized in that, The process data for each key node of the production line includes at least one of the following: raw material temperature, raw material humidity, concrete slump, centrifugal vibration data, and temperature and humidity of the curing environment. The closed-loop control includes at least one of the following: adjusting the water-cement ratio of the mixer, optimizing the centrifuge speed-up curve, and changing the steam curing heating rate.

10. A digital twin-based control system for the manufacturing process of concrete poles, characterized in that, The system includes: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the following modules: The data acquisition module is used to collect process data from key nodes of the production line in real time, obtain multi-source process parameters, and synchronize the multi-source process parameters to a pre-built digital twin model in real time. The simulation closed-loop control module is used to synchronously simulate the state of concrete materials under the current working conditions based on the real-time input multi-source process parameters of the digital twin model, and predict the finished product quality index in real time based on the simulation results. Based on the finished product quality index, the control parameters of the production process are dynamically fine-tuned to realize the closed-loop control of the production process. The quality deviation calculation module is used to obtain the quality inspection indicators of the finished products of the current batch after the completion of a production batch, and compare the quality inspection indicators with the final finished product quality prediction indicators calculated by the digital twin model based on all process data of the current batch and the control parameter adjustments already performed, to calculate the total quality deviation. ; Attribution correction module, used when the total mass deviation When the preset allowable range is exceeded, the key model parameters in the digital twin model are corrected according to the preset attribution logic, and the parameter adjustment log is recorded. The evolutionary trend analysis module is used to periodically analyze the parameter adjustment logs, and for each key model parameter, extract the most recent... The adjustment magnitude of each adjustment is calculated, the average adjustment magnitude is determined, and the evolution tendency of each key model parameter is judged based on the sign and magnitude of the average adjustment magnitude. The pre-compensation adjustment module is used to pre-compensate and adjust the key model parameters in the digital twin model before the start of the next production batch when a certain key model parameter is identified to have an upward or downward evolution trend. This adjustment is based on the pre-compensation adjustment amount obtained from the average adjustment magnitude, the pre-compensation adjustment amount obtained from trend analysis, or the pre-compensation adjustment amount predicted by the exponential smoothing method. The result is used to update the initial state of the digital twin model before the start of the next production batch.

Images