An aluminum material processing order collaborative configuration method based on digital twinning

Abstract

The application relates to the technical field of industrial computing and virtual resource scheduling, and discloses an aluminum material processing order collaborative configuration method based on digital twinning, which comprises the following steps: translating a set of to-be-configured orders into a target state sequence containing process conditions; generating a state drift amount according to a change slope to calibrate a real-time state point and output a reference state point; calculating an impedance evaluation value according to an absolute difference between the reference state point and the process condition and a state recovery coefficient; when the difference between the impedance evaluation values is less than a set resolution threshold, selecting a sequence corresponding to a maximum value of a die change frequency parameter distribution density as a target configuration result and dispatching the target configuration result to a physical production capacity unit data interface. The application converts the physical continuous inertia cost of state switching into scheduling constraints, reduces the stacking delay and state conflict risk in real scheduling, eliminates high-frequency oscillation of the sequence, and ensures the global stability of resource configuration.

Description

Technical Field

[0001] This invention relates to a collaborative configuration method for aluminum processing orders based on digital twins, belonging to the field of industrial computing and virtual resource scheduling technology. Background Technology

[0002] The current virtual resource scheduling framework relies on the state evolution model to construct a model mirror of industrial assets and processing sequences, converting the production capacity status and operating parameters of physical assets into topological nodes in the virtual space. This provides a data basis for conflict verification and orchestration of order sets. Moreover, in typical operating conditions with high continuity requirements, such as heavy-duty non-ferrous metal forming and processing, material flow is accompanied by deep thermodynamic and mechanical deformation processes. This leads to existing resource allocation methods generally relying on discrete state matching logic, directly comparing the current state parameters output by the virtual model with the target material parameters of the order to be configured, and finally issuing a matching trajectory when the two types of parameters fall into the preset static threshold window.

[0003] This type of processing architecture contains an unrealistic idealization, assuming that the asset state parameters in the virtual space have instantaneous switching characteristics. However, in the real continuous manufacturing physical field, the operating characteristics of multiple types of equipment have irreversible physical inertia and time delay. When scheduling instructions issue process change commands across alloy types or specification ranges, physical assets will inevitably undergo a long period of gradual state change ramp-up. Because the configuration model cannot predict and quantify the time and cost of this state ramp-up during the decision-making stage, the issued control sequences will cause scheduling failures and timing conflicts at the production end. Furthermore, attempts to improve the system by adding static buffer timings to the topology periphery or excessively tightening the matching threshold will not only structurally inhibit the overall throughput efficiency of the system, but also more easily trigger frequent order rejection actions, causing the system behavior to frequently fall into a fragile state of high-frequency oscillation. Based on the aforementioned scheduling optimization for hardware limitations such as mechanical deformation and roller life of physical assets, academia and industry are attempting to address the issue from the perspective of control software methods. To further address the conflicts in roll changing and alternating operations under multi-task conditions, the related control methods also have inherent shortcomings. For example, Chinese invention patent application CN119151487A discloses a production line control method, device, and storage medium based on digital twins. It determines the roll type by the size and material of the order sheet and screens the production path by comparing the number and quantity of roll changes in the digital twin model. This mechanism implicitly relies on the static preset premise that the state switching point is at the ideal discrete boundary. It ignores the dynamic zero-point drift of physical parameters caused by frequency conversion high pressure and deep thermodynamic deformation when the non-ferrous metal continuous rolling process flows across alloy grades. It does not quantitatively characterize the state gradual change ramp-up loss in the continuous field. This makes the production scheduling scheme issued by it prone to frequent trajectory rejection and high-frequency configuration oscillation in the dense viscous region when facing high-gradient working condition disturbances due to the accumulation of small truncation errors. It is difficult to maintain the monotonic convergence and global steady state of the full sequence resource scheduling from the bottom control loop.

[0004] Therefore, how to transform the physical continuity inertia cost of state switching of underlying physical assets into multi-dimensional optimization constraints in virtual space, and establish a full-order configuration process that includes state drift compensation and timing lock, thereby eliminating the time and loss deviation caused by state gradual change, has become an essential requirement for efficient configuration of physical assets. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A collaborative configuration method for aluminum material processing orders based on digital twins, comprising the following steps:

[0006] Step S101: Obtain the set of aluminum processing orders to be configured, and translate the alloy grade data and demand output data contained in each aluminum processing order in the set of aluminum processing orders to be configured into a target state sequence containing process constraint parameters.

[0007] Step S102: Extract the historical change slope of the output parameters of the virtual state evolution model according to the preset drift time window to generate a state drift vector. Use the state drift vector to calibrate the position coordinates of the real-time spatial state point in the virtual space and output the reference state point.

[0008] Step S103: Based on the absolute difference between the current parameter in the reference state point and the corresponding process constraint parameter in the target state sequence, the weighted combination of the preset state recovery coefficient in the virtual state evolution model is applied and summed to calculate the single-valued impedance evaluation value between the reference state point and the starting point of each target state sequence.

[0009] Step S104: When the difference between the single-valued impedance evaluation values ​​of two adjacent target state sequences is less than the set resolution threshold, extract the distribution density of the mode switching frequency parameter in each target state sequence, select the target state sequence corresponding to the maximum distribution density as the target configuration result, and assign the target configuration result to the underlying physical production capacity unit data interface of the virtual state evolution model image to update the virtual resource scheduling track.

[0010] Preferably, after step S104, the following steps are also included: Step S105, calculate the time span of the target state sequence on the underlying physical capacity unit based on the target configuration result, and establish a timing isolation lock inside the virtual state evolution model to suspend the calculation process of the single-value impedance evaluation value of subsequent rounds of orders; Step S106, during the operation of the timing isolation lock, continuously monitor the signal flow received by the data interface of the underlying physical capacity unit until the completion status identification data fed back by the underlying physical capacity unit is obtained, then release the timing isolation lock and release the updated reference state point.

[0011] Preferably, step S101 includes the following sub-steps: step S1011, extracting alloy grade data and demand output data from each aluminum processing order; step S1012, converting the alloy grade data and demand output data into multi-dimensional feature vectors as process constraint parameters in the corresponding target state sequence.

[0012] Preferably, in step S102, the output parameters of the virtual state evolution model include the current operating rate data and operating temperature data of the underlying physical production capacity unit, and the operating environment temperature is set to 20°C to 40°C.

[0013] Preferably, step S103 includes the following sub-steps: step S1031, calculating the Euclidean distance between the reference state point and each target state sequence in the virtual topology space; step S1032, obtaining the over-temperature variable delay parameter of the underlying physical power generation unit, and weighting the Euclidean distance and the over-temperature variable delay parameter to generate a single-value impedance evaluation value.

[0014] Preferably, the secondary constraint parameters include the mold change frequency parameter, cleaning time consumption parameter, and energy consumption rate parameter corresponding to each aluminum material processing order.

[0015] Preferably, in step S104, extracting the distribution density of the mode-switching frequency parameter in each target state sequence includes the following sub-steps: Step S1041, constructing a multidimensional parameter topology space with the mode-switching frequency as the dimension axis; Step S1042, when the difference between the single-value impedance evaluation values ​​is less than the set resolution threshold, calculating the local kernel density estimate of each target state sequence in the multidimensional parameter topology space as the distribution density.

[0016] Preferably, step S105 includes the following sub-steps: step S1051, extracting the aluminum material specification parameters, rolling pass constraints and heat treatment isothermal time data corresponding to the target configuration result; step S1052, calculating the processing time per ton by combining the rated power data of the underlying physical capacity unit and the current state attenuation coefficient, and multiplying the processing time per ton by the required output to output the time span.

[0017] Preferably, in step S105, establishing a timing isolation lock within the virtual state evolution model includes the following sub-steps: Step S1053, starting a delay countdown thread within the virtual state evolution model, and setting the initial time of the delay countdown thread to the time span; Step S1054, before the delay countdown thread reaches zero, based on the timing blocking factor... Intercept subsequent computation requests, among which, This is the current closed state constant of the timing isolation lock.

[0018] Compared with the prior art, the beneficial effects of the present invention are:

[0019] 1. By weighting the parameter difference between the baseline state point and the target state sequence using the state recovery coefficient and forming a single-valued impedance assessment value, the state differences that the equipment needs to overcome to switch from the current operating condition to the target operating condition of the order can be incorporated into the order configuration process. Compared with matching only according to static thresholds, this method can more accurately reflect the time and losses brought about by the gradual change of equipment state, making the scheduling results more consistent with the actual operation of the equipment in the continuous manufacturing process, thereby reducing material stacking delays and state conflicts caused by the asynchrony between scheduling instructions and equipment state.

[0020] 2. By extracting the historical change slope of the output parameters of the virtual state evolution model using a preset drift time window, and calibrating the real-time spatial state points through the state drift vector, the state changes between the data sampling time and the order configuration calculation time can be compensated. The resulting benchmark state points can more timely reflect the actual working conditions of the underlying physical capacity units, reduce the virtual-real state deviation caused by on-site disturbances and calculation delays, and enable subsequent single-value impedance assessment and target state sequence screening to be based on a more accurate and stable equipment state, thereby improving the reliability of digital twin resource scheduling results.

[0021] 3. When the single-value impedance evaluation values ​​of adjacent target state sequences are relatively close, the selection based solely on minute numerical differences is stopped by using the impedance resolution threshold. Furthermore, the target configuration result is determined based on the distribution density of secondary constraint parameters. This can reduce the impact of calculation truncation errors and data fluctuations on order selection. This method can form a stable and clear configuration result among multiple target state sequences with similar conditions, reduce the repeated switching of scheduling results and the frequent issuance of configuration instructions, and keep the order collaborative configuration process stable. Attached Figure Description

[0022] Figure 1 This is a flowchart of the collaborative configuration steps for aluminum material processing orders in this invention;

[0023] Figure 2 This is a schematic diagram showing the changes in system performance and load indicators corresponding to the impedance resolution threshold of the present invention.

[0024] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0026] A collaborative configuration method for aluminum processing orders based on digital twins includes the following steps:

[0027] Step S101: Obtain the set of aluminum processing orders to be configured, and translate the alloy grade data and demand output data contained in each aluminum processing order in the set of aluminum processing orders to be configured into a target state sequence containing process constraint parameters.

[0028] Step S102: Extract the historical change slope of the output parameters of the virtual state evolution model according to the preset drift time window to generate a state drift vector. Use the state drift vector to calibrate the position coordinates of the real-time spatial state point in the virtual space and output the reference state point.

[0029] Step S103: Based on the absolute difference between the current parameter in the reference state point and the corresponding process constraint parameter in the target state sequence, the weighted combination of the preset state recovery coefficient in the virtual state evolution model is applied and summed to calculate the single-valued impedance evaluation value between the reference state point and the starting point of each target state sequence.

[0030] Step S104: When the difference between the single-valued impedance evaluation values ​​of two adjacent target state sequences is less than the set resolution threshold, extract the distribution density of the mode switching frequency parameter in each target state sequence, select the target state sequence corresponding to the maximum distribution density as the target configuration result, and assign the target configuration result to the underlying physical production capacity unit data interface of the virtual state evolution model image to update the virtual resource scheduling track.

[0031] Preferably, after step S104, the following steps are also included: Step S105, calculate the time span of the target state sequence on the underlying physical capacity unit based on the target configuration result, and establish a timing isolation lock inside the virtual state evolution model to suspend the calculation process of the single-value impedance evaluation value of subsequent rounds of orders; Step S106, during the operation of the timing isolation lock, continuously monitor the signal flow received by the data interface of the underlying physical capacity unit until the completion status identification data fed back by the underlying physical capacity unit is obtained, then release the timing isolation lock and release the updated reference state point.

[0032] Preferably, step S101 includes the following sub-steps: step S1011, extracting alloy grade data and demand output data from each aluminum processing order; step S1012, converting the alloy grade data and demand output data into multi-dimensional feature vectors as process constraint parameters in the corresponding target state sequence.

[0033] Preferably, in step S102, the output parameters of the virtual state evolution model include the current operating rate data and operating temperature data of the underlying physical production capacity unit, and the operating environment temperature is set to 20°C to 40°C.

[0034] Preferably, step S103 includes the following sub-steps: step S1031, calculating the Euclidean distance between the reference state point and each target state sequence in the virtual topology space; step S1032, obtaining the over-temperature variable delay parameter of the underlying physical power generation unit, and weighting the Euclidean distance and the over-temperature variable delay parameter to generate a single-value impedance evaluation value.

[0035] Preferably, the secondary constraint parameters include the mold change frequency parameter, cleaning time consumption parameter, and energy consumption rate parameter corresponding to each aluminum material processing order.

[0036] Preferably, in step S104, extracting the distribution density of the mode-switching frequency parameter in each target state sequence includes the following sub-steps: Step S1041, constructing a multidimensional parameter topology space with the mode-switching frequency as the dimension axis; Step S1042, when the difference between the single-value impedance evaluation values ​​is less than the set resolution threshold, calculating the local kernel density estimate of each target state sequence in the multidimensional parameter topology space as the distribution density.

[0037] Preferably, step S105 includes the following sub-steps: step S1051, extracting the aluminum material specification parameters, rolling pass constraints and heat treatment isothermal time data corresponding to the target configuration result; step S1052, calculating the processing time per ton by combining the rated power data of the underlying physical capacity unit and the current state attenuation coefficient, and multiplying the processing time per ton by the required output to output the time span.

[0038] Preferably, in step S105, establishing a timing isolation lock within the virtual state evolution model includes the following sub-steps: Step S1053, starting a delay countdown thread within the virtual state evolution model, and setting the initial time of the delay countdown thread to the time span; Step S1054, before the delay countdown thread reaches zero, based on the timing blocking factor... Intercept subsequent computation requests, among which, This is the current closed state constant of the timing isolation lock.

[0039] Example 1: When the underlying physical production unit continuously processes aluminum materials, the material undergoes thermal changes and mechanical deformation. The operating environment temperature remains between 20°C and 40°C. The set of aluminum processing orders to be configured contains different alloy grade data and required output data. When the equipment switches between alloy grades or specification ranges, the operating speed and operating temperature cannot instantly reach a new stable state, but will undergo a gradual state change process. If only static thresholds are used to compare the current state parameters with the order parameters, the configuration command is likely to be issued before the equipment state switch is completed, resulting in scheduling misalignment, material stacking delays, and timing conflicts. The system obtains the set of aluminum processing orders to be configured and extracts the alloy grade data and required output data of each aluminum processing order one by one. For alloy grade data, the system establishes an alloy grade mapping table, classifying alloy grades in text or code form and converting them into uniquely coded state sequences for corresponding dimensions. For demand and production data, the system normalizes the data according to the rated processing quality of a single batch of the underlying physical production capacity unit, converting it into dimensionless values ​​between 0 and 1. Following a preset dimensional order, the system concatenates the uniquely coded state sequences with the normalized demand and production data column-wise to form a multi-dimensional feature vector for the corresponding aluminum processing order. This multi-dimensional feature vector serves as the starting point for the target state sequence. Each aluminum processing order corresponds to one target state sequence, and each target state sequence is saved separately according to the order identifier. The multi-dimensional feature vector in the target state sequence serves as a process constraint parameter. .

[0040] The virtual state evolution model stores the correspondence between the underlying physical production capacity units, real-time operating parameters, target state sequences, and virtual space coordinates. Following the same dimensional order and normalization rules, it maps real-time spatial state points and process constraint parameters to the same virtual topology space. The underlying physical production capacity unit data interface continuously collects current operating rate data and operating temperature data, inputting the results into the virtual state evolution model. The model determines the real-time spatial state point based on the latest sampled value. Simultaneously, within a preset drift time window, it calculates the ratio of the difference between adjacent sampled values ​​of the current operating rate data and operating temperature data to the sampling interval, and summarizes the slopes of each change to obtain the historical slope of each parameter within the drift time window. The model multiplies the historical slope of each parameter by the time difference between the latest sampling time and the current configuration calculation time to obtain the coordinate correction amount for the corresponding parameter. This correction is then used to construct a state drift vector according to the parameter dimensions. Subsequently, following the direction of each slope, the state drift vector is sequentially superimposed onto the real-time spatial state point to obtain the calibrated reference state point. Through the above processing, the reference state point reflects the device state at the time of configuration calculation, rather than the device state at the time of the most recent sampling of the data interface.

[0041] After obtaining the baseline state point, the system calculates the absolute difference between the current parameters at the baseline state point and the corresponding process constraint parameters at the starting point of each target state sequence, according to the dimensional correspondence in the virtual topology space, and combines this with the preset state recovery coefficients within the virtual state evolution model. Weighted combination and summation, single-valued impedance assessment value The weight combination relationship follows the following formula: ,in, This is a single-value impedance assessment value. The state recovery coefficient, For the corresponding parameters in the reference state point, The formula uses the corresponding process constraint parameters in the starting point of the target state sequence. The summation is performed on parameters that can correspond dimensionally between the reference state point and the starting point of the target state sequence. Each parameter is converted to a dimensionless value according to its preset value range before participating in the calculation. In this embodiment, the system further calculates the Euclidean distance between the reference state point and each starting point of the target state sequence in the virtual topology space. This Euclidean distance is the arithmetic square root of the sum of the squares of the differences between their corresponding dimensions, used to represent the spatial deviation between the current equipment state and the order target state. The over-temperature time delay parameter of the underlying physical capacity unit is determined by the response time required for the equipment to change from the current operating temperature to the target operating temperature, in seconds. During calculation, the system divides the temperature-varying delay parameter by a 100-second reference delay to obtain a dimensionless relative delay scalar. Simultaneously, it divides the Euclidean distance by the maximum characteristic reference distance between the allowed state coordinates in the virtual topological space to obtain a dimensionless geometric distance. The dimensionless geometric distance is multiplied by a weighting coefficient of 0.6, and the dimensionless relative delay scalar is multiplied by a weighting coefficient of 0.4. The two are then superimposed to obtain a single-valued impedance assessment value that simultaneously reflects spatial state deviation and temperature-varying delay. The state recovery coefficient is used to correct the influence of the difference between the corresponding parameters on the state recovery process. The Euclidean distance weight and the temperature-varying delay weight are used to determine the proportion of spatial deviation and delay costs in the single-valued impedance assessment value.

[0042] The system arranges the target state sequences in ascending order of single-value impedance assessment values ​​and compares the single-value impedance assessment values ​​of two adjacent target state sequences. When the difference between the two is less than the impedance resolution threshold, the target configuration result is no longer determined solely based on the smaller single-value impedance assessment value. Instead, the distribution density of the mold-changing frequency parameter in the corresponding target state sequence is extracted. The secondary constraint parameters of each aluminum processing order include mold-changing frequency parameters, cleaning time consumption parameters, and energy consumption rate parameters. The system constructs a multi-dimensional parameter topology space containing three orthogonal axes. The first dimension axis is the mold-changing frequency, in times / day; the second dimension axis is the cleaning time consumption, in minutes / time; and the third dimension axis is the energy consumption rate, in kilowatt-hours / ton. The number of secondary constraint parameters for each target state sequence is... The group is mapped to a discrete coordinate point in the multidimensional parameter topology space. When calculating the distribution density of the mode-switching frequency parameters, the system projects each discrete coordinate point onto the mode-switching frequency dimension axis and uses a Gaussian kernel function to perform local kernel density estimation on the mode-switching frequency coordinates corresponding to each target state sequence. The obtained local kernel density estimate is used as the distribution density of the mode-switching frequency parameters. The cleaning time consumption parameter and energy consumption rate parameter are retained in the corresponding multidimensional coordinates to characterize the secondary constraint state of the target state sequence, but do not replace the mode-switching frequency parameters in the calculation of the distribution density. The system selects the target state sequence corresponding to the maximum value of the distribution density of the mode-switching frequency parameters as the target configuration result, thereby forming a unique selection result among multiple target state sequences with similar single-value impedance evaluation values.

[0043] After determining the target configuration result, the system assigns it to the underlying physical production capacity unit data interface mapped by the virtual state evolution model and updates the virtual resource scheduling track. Based on the target configuration result, the system extracts the corresponding aluminum material specification parameters, rolling pass constraints, and heat treatment isothermal time data. The basic deformation energy required to process each ton of material is determined from the aluminum material specification parameters and rolling pass constraints, and the heat energy retention requirement corresponding to the heat treatment isothermal time is included to obtain the basic energy consumption per ton of aluminum material, in kilojoules per ton. The current state decay coefficient ranges from 1.0 to 1.5 and is used to characterize the design... The impact of equipment aging and transmission wear on effective output power is investigated. When the coefficient is 1.0, the underlying physical production capacity unit operates at its rated power. As the coefficient increases, the effective output power decreases accordingly. The system divides the rated power data by the current state attenuation coefficient to obtain the current actual effective processing power. After dividing the basic energy consumption by the actual effective processing power, the processing time is first obtained in seconds / ton, and then converted into processing time per ton in hours / ton. The processing time per ton is multiplied by the required output corresponding to the target configuration result to obtain the time span of the target state sequence on the underlying physical production capacity unit.

[0044] The system starts a delay countdown thread within the virtual state evolution model and sets the time span directly as the initial time of the delay countdown thread to establish a timing isolation lock. The timing lock factor is set to 1.05 when the timing lock is closed, indicating that the current calculation process is in a locked state and does not participate in the calculation of time span and countdown duration. Before the delay countdown thread reaches zero, the central processing unit continuously reads the remaining countdown time. When the remaining time is greater than 0, the enable flag for single-value impedance evaluation calculation is set to logic false (0), intercepting subsequent rounds of order submissions for single-value impedance evaluation calculation requests and returning a waiting-for-response status. During the timing lock operation, the system continuously monitors the signal flow received by the underlying physical capacity unit data interface, and the delay countdown... When a thread reaches zero but has not yet received completion status data, the timing isolation lock remains closed. Only after the countdown reaches zero and the completion status data from the underlying physical production capacity unit is received will the system release the timing isolation lock and reset the allow flag to logical true (1). Subsequently, the system rereads the current operating rate data and operating temperature data of the underlying physical production capacity unit, calculates the historical change slope and state drift vector according to the latest drift time window, calibrates the real-time spatial state point, generates an updated baseline state point, and then releases the baseline state point to the next round of order configuration calculation.

[0045] After calibrating the real-time spatial state points using the state drift vector, the baseline state points used for order configuration calculations correspond to the actual operating state of the underlying physical capacity units. The single-value impedance evaluation value incorporates the spatial changes required for equipment state switching and the over-temperature time delay into the order selection process. When the single-value impedance evaluation values ​​of multiple target state sequences are close, the target configuration result is determined based on the distribution density of the mold changing frequency parameter. After the target configuration result is issued, the timing isolation lock suspends the calculation of subsequent rounds according to the corresponding time span until the current order is actually completed and an updated baseline state point is generated. Thus, the virtual resource scheduling track is updated according to the actual state changes and completion progress of the physical equipment, avoiding the premature allocation of new aluminum processing orders before the equipment has completed heating, cooling, or operating state switching, reducing scheduling misalignment, material stacking delays, timing conflicts, and high-frequency oscillations caused by repeated issuance of configuration instructions.

[0046] Example 2: The resource scheduling verification platform collects current operating rate data and operating temperature data through the underlying physical production capacity unit data interface. The interface has a temperature measurement resolution of 0.1℃ and a data sampling frequency of 50Hz. During the test, the operating environment temperature of the underlying physical production capacity unit is maintained between 20℃ and 40℃. To simulate the interference of the variable frequency high-voltage line in the continuous processing workshop on the collected signal, Gaussian white noise with a signal-to-noise ratio of 20dB and a harmonic component with a frequency of 50Hz are superimposed on the original data stream. The processed data is then input into the virtual state evolution model. The drift time window is determined according to the data processing load and the transient temperature change rate. When the change rate of the operating temperature data increases, the drift time window is shortened accordingly to reduce the aliasing of discrete sampling points on the time axis. When the rate of change is between 0.2℃ / s and 0.5℃ / s, a 30-second reference temperature window is used. The drift time window is determined by dividing the reference temperature window by 1 and adding twice the temperature change rate. When the temperature change rate is 0.5℃ / s, the drift time window is 15 seconds. The working range of the impedance resolution threshold is set to 0.02 to 0.10. The standard deviation of the background Gaussian white noise collected by the data interface is 0.015. 1.33 times this standard deviation is taken as the basis for determining the lower limit, resulting in a lower limit value of 0.02. The minimum physical impedance difference between adjacent alloy grades is 0.12. 83% of this impedance difference is taken as the basis for determining the upper limit, resulting in an upper limit value of 0.10. This range is used to distinguish between the small fluctuations caused by sensor noise and the actual impedance differences between different target state sequences.

[0047] The experiment setup included a control group using the traditional discrete state comparison method and an experimental group using the configuration method described in this embodiment. Based on the original operating condition drift rate of the input data, the continuous production simulation sequence was divided into three scenarios: low-gradient perturbation, medium-gradient perturbation, and high-gradient perturbation. The original operating condition drift rates for the three scenarios were set to 0.05 units / s, 0.15 units / s, and 0.35 units / s, respectively. Both groups received the same operating rate data, operating temperature data, and noise perturbation. The difference was that the control group directly used real-time spatial state points for order matching, while the experimental group extracted the historical data of operating rate and operating temperature within the drift time window. The historical slope is used to generate a state drift vector, and the real-time spatial state point is calibrated accordingly before the reference state point is output. Under high gradient perturbation, Gaussian white noise in the original data stream causes the real-time spatial state point to continuously deviate from the actual operating trajectory of the underlying physical production unit. The control group did not calibrate this deviation, and the coordinate space deviation of its real-time spatial state point relative to the actual trajectory reached 14.3%. The experimental group generated a state drift vector according to the historical slope and translated and calibrated the virtual space coordinates according to the direction of change of each parameter. The coordinate error of the output reference state point relative to the actual position of the physical entity was reduced to below 0.42%.

[0048] To verify the impact of the impedance resolution threshold on the configuration process, boundary state stress tests were conducted with absolute lower limits, normal median, and absolute upper limits of 0.01, 0.05, and 0.15, respectively. When the impedance resolution threshold was further reduced to 0.005, the absolute difference between the baseline state point and the process constraint parameters was continuously disturbed by Gaussian white noise. The difference between the single-value impedance evaluation values ​​of adjacent target state sequences triggered local kernel density estimation calculations for the mold change frequency parameter, cleaning time consumption parameter, and energy consumption rate parameter within 30 seconds, and the processor's computational load rose to 94.2%, indicating computational saturation. The single-value impedance evaluation values ​​were weighted and summed according to the state recovery coefficient based on the absolute difference between the current parameter and the corresponding process constraint parameter in the baseline state point. After completing the single-value impedance assessment, the system compares the difference between the assessment values ​​of two adjacent target state sequences and decides whether to enter the distribution density calculation process of the die-changing frequency parameter based on the impedance resolution threshold. When the impedance resolution threshold increases to 0.20, the difference judgment interval is too wide, and the sequence containing data truncation error enters the subsequent screening process. The local impedance stagnation state fails to stably trigger the distribution density screening according to the set conditions. The workshop scheduling sequence repeatedly switches between similar target state sequences. The frequency of process conflict between die changing and rolling increases from 0.5 times / hour under normal conditions to 3.8 times / hour. When the impedance resolution threshold is maintained between 0.02 and 0.10, the local impedance stagnation under low gradient disturbance, medium gradient disturbance and high gradient disturbance can all enter the corresponding distribution density screening process.

[0049] After determining the target configuration result that meets the rolling constraints, the system assigns the result to the underlying physical capacity unit data interface and updates the virtual resource scheduling track. Subsequently, based on the target configuration result, it extracts the corresponding aluminum material specification parameters, rolling pass constraints, and heat treatment isothermal time data. Combining the rated power data of the underlying physical capacity unit and the current state decay coefficient, it calculates the processing time per ton. Then, it multiplies the processing time per ton by the required output data to obtain the time span of the target state sequence on the underlying physical capacity unit. The system starts a time delay countdown thread inside the virtual state evolution model, sets the time span as the initial time of the countdown thread, and establishes a timing isolation lock accordingly. Before the countdown reaches zero, the timing blocking factor keeps the timing isolation lock closed. The system intercepts single-value impedance evaluation value calculation requests submitted by subsequent rounds of orders. After the countdown reaches zero, if the completion status identification data has not been received, the timing isolation lock continues to remain closed. The system remains closed until the underlying physical production capacity unit feeds back the completion status identification data through the bus. At this point, the system releases the timing isolation lock, re-collects the current operating rate data and operating temperature data, generates a state drift vector, calibrates the real-time spatial state point, and releases the updated reference state point. Under high gradient disturbance conditions, the control group did not include the ramp-up delay and loss parameters during the equipment state transition process in the order configuration. Its order scheduling conflict rate was 22.4%, and the number of order rejection actions caused by control loop scheduling misalignment was 14.8 times / hour. The experimental group reduced the order scheduling conflict rate to below 1.6% and the number of order rejection actions in the continuous order sequence to 0 times / hour by using reference state point calibration, single-value impedance evaluation, mode-changing frequency parameter distribution density screening, and timing isolation lock control. The virtual resource scheduling track is updated according to the state changes of the underlying physical production capacity unit and the actual completion progress. The continuous order configuration process remains monotonically convergent.

[0050] Example 3: When aluminum forming and processing equipment continuously performs mixed-line operations, the underlying physical production capacity unit experiences physical inertia and time delay during the switching process across alloy grades and specification ranges. The operating speed and operating temperature of the equipment need to undergo a gradual change process to reach a new stable state. This example adopts a control loop composed of a central processing unit and dynamic memory. The operating environment temperature is maintained between 20°C and 40°C. The central processing unit reads order data, current operating speed data, and operating temperature data through the underlying physical production capacity unit data interface and executes order configuration accordingly. The central processing unit first extracts the alloy grade data and required output data of each order from the set of aluminum processing orders to be configured. According to the predetermined encoding, normalization, and dimensional splicing order, it converts the two into multi-dimensional feature vectors as process constraint parameters in the corresponding target state sequence. The current operating speed data and operating temperature data are used as output parameters of the virtual state evolution model to determine the real-time spatial state point. The central processing unit extracts the historical change slope of the above output parameters within the drift time window, generates a state drift vector, and calibrates the position coordinates of the real-time spatial state point in the virtual space according to the change direction of each parameter to obtain the reference state point.

[0051] Before calculating the single-valued impedance assessment value, the central processing unit reads the preset state recovery coefficient inside the virtual state evolution model. This coefficient is calibrated through a step test: under the control environment, multiple sets of samples with identical conditions except for the mode switching frequency parameter are selected, and state switching commands are applied to each. The recovery time of the underlying physical power generation unit from the initial state to the stable state is recorded, and a monotonic correspondence between the recovery time and the change in single-valued impedance is established. When the recovery time is between 120 seconds and 300 seconds, the state recovery coefficient is determined in reverse according to this correspondence. Under the calibration condition with a recovery time of 120 seconds, the state recovery coefficient is taken as 0.85. The state recovery coefficient is calculated by subtracting the ratio of the recovery time to the maximum allowable recovery time from the reference constant 1.0 and multiplying it by 0. The result of 375 is determined, in which the maximum allowable recovery time is set to 300 seconds. After substituting the recovery time of 120 seconds, the state recovery coefficient is 0.85. As the recovery time increases, the state recovery coefficient decreases accordingly, so that the coefficient corresponds to the equipment recovery capability. The central processing unit matches the current parameters in the reference state point with the corresponding process constraint parameters in the starting point of each target state sequence dimension by dimension, calculates the absolute difference between the parameters in each dimension, and performs weighted combination according to the state recovery coefficient. Before participating in the combination, each parameter is converted into a dimensionless value according to the same normalization rule to maintain the consistency of dimensions between different parameters. Subsequently, the central processing unit calculates the Euclidean distance between the reference state point and the starting point of the target state sequence in the virtual topology space.

[0052] The over-temperature time delay parameter is obtained through temperature sensors installed at key thermal nodes of the underlying physical production unit. The temperature sensors record the response time required for the equipment to rise from the initial temperature to the target operating temperature, for example, the time required for the equipment to rise from 20°C to 40°C. This response time is used as the over-temperature time delay parameter. The central processing unit (CPU) performs dimensionless processing on the Euclidean distance and the over-temperature time delay parameter, and then superimposes their weights to obtain the single-value impedance evaluation value between the reference state point and the starting point of each target state sequence. When the difference between the single-value impedance evaluation values ​​of two adjacent target state sequences is less than the impedance resolution threshold, the CPU enters the secondary constraint parameter processing process. The CPU reads the mode switching frequency parameter, cleaning time consumption parameter, and energy consumption rate parameter from the corresponding target state sequence and writes the feature array composed of them into the designated cache area of ​​the dynamic memory. Subsequently, a multi-dimensional parameter topology space with the mode switching frequency as the dimension axis is constructed in the dynamic memory, and the mode switching frequency parameter corresponding to each target state sequence is mapped to this dimension axis.

[0053] The central processing unit (CPU) calls the kernel density estimation program to calculate the local kernel density estimate of the mold-changing frequency parameter for each target state sequence. This estimate is then used as the distribution density of the mold-changing frequency parameter. The cleaning time consumption parameter and energy consumption rate parameter are stored in the multi-dimensional parameter topology space along with the corresponding target state sequence to maintain the correspondence between secondary constraint parameters, but they do not replace the mold-changing frequency parameter in the distribution density calculation. The CPU selects the target state sequence corresponding to the maximum distribution density as the target configuration result and assigns this result to the underlying physical production capacity unit data interface. At the same time, it rewrites the scheduling status bit in the virtual resource scheduling track. After determining the target configuration result, the CPU reads the corresponding aluminum material specification parameters, rolling pass constraints, and heat treatment isothermal time data. It combines the rated power data of the underlying physical production capacity unit with the current state decay coefficient to calculate the processing time per ton. Then, it multiplies the processing time per ton by the required output data to obtain the time span of the target state sequence on the underlying physical production capacity unit.

[0054] The central processing unit (CPU) starts a delay countdown thread within the virtual state evolution model, directly setting the time span as the thread's initial time and establishing a timing isolation lock accordingly. Before the countdown reaches zero, the CPU reads the timing blocking factor and sets the allow flag for the single-value impedance evaluation to 0, intercepting subsequent concurrent scheduling requests. The timing blocking factor serves as the closing state constant of the timing isolation lock, characterizing the safety margin during the blocking period, without changing the time span or the initial time of the delay countdown thread. The value of the timing blocking factor is based on the changes in the physical residual output from the preceding hot rolling process. The central processing unit continuously collects the actual plate thickness and temperature of 10 batches of exported materials, calculates the deviation between each batch and the set target value, and then calculates the root mean square residual based on the deviation sequence. When the root mean square residual is less than or equal to 1%, the closure state constant of the timing block factor is set to 1.00. When the root mean square residual is between 1% and 5%, it is linearly adjusted by adding the root mean square residual multiplied by 1.0 to the baseline value of 1.00. When the root mean square residual is 5%, the closure state constant is 1.05, and this value represents the closure state safety margin adopted by the current timing isolation lock.

[0055] Before the countdown thread reaches zero, the timing isolation lock continuously intercepts single-value impedance evaluation calculation requests for subsequent rounds of orders. After the countdown reaches zero, if the underlying physical capacity unit has not yet returned completion status data, the timing isolation lock remains closed. After the underlying physical capacity unit completes the current batch of work and returns the completion status data through the bus interface, the CPU cancels the countdown thread, releases the timing isolation lock, and restores the allow flag for single-value impedance evaluation calculation to 1. After the timing isolation lock is released, the CPU rereads the current operating rate data and operating temperature data of the underlying physical capacity unit, extracts the historical change slope within the new drift time window and generates a state drift vector, calibrates the real-time spatial state point, and obtains the updated baseline state point. Subsequently, the CPU updates the virtual resource scheduling track and releases the baseline state point to the next round of order configuration, so that the single-value impedance evaluation of the subsequent target state sequence starts from the actual state after the underlying physical capacity unit completes the current batch of work.

[0056] Example 4: This example combines Figures 1 to 2 This paper describes a collaborative configuration method for aluminum processing orders based on digital twins, such as... Figure 1As shown, in step S101, a set of aluminum processing orders to be configured is obtained, and the alloy grade data and demand output data of each aluminum processing order in the set are translated into a target state sequence containing process constraint parameters. After the translation is completed, step S102 is entered, and the historical change slope of the output parameters of the virtual state evolution model is extracted according to the preset drift time window to generate a state drift vector. The position coordinates of the real-time spatial state point in the virtual space are calibrated using the state drift vector and the reference state point is output. After the reference state point is output, step S103 is entered, and the single-value impedance evaluation value is calculated by combining the absolute difference between the current parameter in the reference state point and the corresponding process constraint parameter in the target state sequence with the preset state recovery coefficient in the virtual state evolution model and applying weight combination and summation. After the calculation is completed, step S104 is entered, and when the difference between the single-value impedance evaluation values ​​of two adjacent target state sequences is less than the set resolution threshold, the distribution density of the corresponding mode switching frequency parameter is extracted, and the sequence corresponding to the maximum distribution density is selected as the target configuration result and assigned to the bottom interface update schedule.

[0057] like Figure 2 As shown, the dimensionless impedance resolution threshold includes values ​​of 0.00, 0.04, 0.08, 0.12, 0.16, and 0.20, while the system performance and load indicators (%) include values ​​of 0%, 25%, 50%, 75%, and 100%. The processor workload and scheduling conflict frequency change accordingly with the adjustment of the dimensionless impedance resolution threshold. When the dimensionless impedance resolution threshold increases within the range of 0.00 to 0.04, the processor workload decreases rapidly from 100% and remains relatively flat within the range of 0.04 to 0.20. When the dimensionless impedance resolution threshold is within the range of 0.00 to 0.12, the scheduling conflict frequency remains low, around 0%. As the dimensionless impedance resolution threshold further increases from 0.12 to 0.20, the scheduling conflict frequency increases.

[0058] Example 5: In the newly established aluminum forming and processing manufacturing environment, before the system is formally connected to production, the initialization and calibration of the virtual state evolution model are completed. The central processing unit receives the historical operation records of physical equipment authorized by the terminal device user and anonymized, and extracts the initial operating rate, rated power, and mold changing frequency parameters from them as the initial parameters when different batches of physical assets are connected to the model. In order to identify the mass transfer resistance and response hysteresis caused by the assembly gap of physical assets, the central processing unit controls the test bench to apply discrete gradient excitation signals, and synchronously collects the operating temperature and deformation residual response sequence through corresponding sensors. Subsequently, the input of the gradient excitation signal is calculated. The time-domain discrete variance between the rate and the acquisition sequence is used to calibrate the initialization parameters of the virtual state evolution model. After the initialization parameter calibration is completed, the central processing unit maps the initial operating rate data and operating temperature data of the underlying physical production capacity unit to the virtual space to form real-time spatial state points. Within the preset drift time window, the historical change slope of the operating rate data and operating temperature data is extracted to generate a state drift vector. Then, the position coordinates of the real-time spatial state points are calibrated according to the change direction of each parameter, and the initial reference state point before formal access to production is output. This reference state point serves as the starting state for the first calculation of the single-value impedance evaluation value in the online scheduling phase.

[0059] After the control loop switches to the online scheduling stage, each time the central processing unit receives a set of aluminum processing orders to be configured, it extracts alloy grade data and demand output data from each aluminum processing order. The alloy grade data is converted into the corresponding coded state according to a predetermined mapping relationship, and the demand output data is normalized according to the rated processing quality of a single batch of the underlying physical capacity unit. Then, it is combined into a multi-dimensional feature vector according to a predetermined dimensional order, which serves as the process constraint parameters in the corresponding target state sequence. As the physical equipment operates for a long time, asset wear and tear will cause the actual operating state to gradually deviate from the existing parameters in the virtual state evolution model. The central processing unit periodically calibrates the preset state recovery coefficient and time-series blocking factor inside the model based on the continuously collected historical change slope of the operating temperature and the state drift vector. The state recovery coefficient is updated according to the predetermined correspondence between the equipment recovery time and the change in single-value impedance. The time-series blocking factor is updated according to the residual change between the actual state and the target state during continuous batch processing.

[0060] When the coordinate deviation between the continuously sampled state points and the current reference state point exceeds the preset control range, the central processing unit (CPU) starts a tuning thread to update the dimension mapping rules used to construct multidimensional feature vectors. The updated rules still take alloy grade data and demand production data as input, only adjusting the dimension mapping and combination of the two in the target state sequence. After the update is completed, the process constraint parameters corresponding to each aluminum processing order are regenerated, and the real-time spatial state points are calibrated according to the latest state drift vector to obtain new reference state points. The CPU calculates the absolute difference between the current parameters in the new reference state point and the corresponding process constraint parameters in the starting point of each target state sequence, and then combines the updated state recovery coefficients for weighted combination and summation to obtain the single-valued impedance evaluation value corresponding to each target state sequence. Subsequently, the target state sequences are arranged according to the magnitude of the single-valued impedance evaluation value, and the difference between the single-valued impedance evaluation values ​​of two adjacent target state sequences is compared.

[0061] When the difference between the single-valued impedance assessment values ​​of two adjacent target state sequences is less than the impedance resolution threshold, the central processing unit (CPU) extracts the mold-changing frequency parameter from the corresponding target state sequence, calculates the local kernel density estimate in the multidimensional parameter topology space with the mold-changing frequency as the dimension axis, and uses it as the distribution density of the mold-changing frequency parameter. The CPU selects the target state sequence corresponding to the maximum distribution density as the target configuration result, assigns the result to the underlying physical production capacity unit data interface mapped by the virtual state evolution model, and updates the virtual resource scheduling track. After the target configuration result is issued, the CPU calculates the processing time per ton based on the corresponding aluminum material specifications, rolling pass constraints, heat treatment isothermal time data, rated power data, and current state decay coefficient. Then, it multiplies the processing time per ton by the required output to obtain the time span of the target state sequence on the underlying physical production capacity unit. The virtual state evolution model then starts a delay countdown thread and sets the time span as the initial time of the thread to establish a timing isolation lock.

[0062] Before the countdown thread reaches zero, the CPU maintains the closed state of the timing isolation lock based on the timing blocking factor, intercepting the single-value impedance evaluation calculation requests submitted by subsequent orders. After the countdown reaches zero, if the underlying physical production capacity unit has not yet fed back the completion status identification data, the timing isolation lock remains closed. Only after the countdown reaches zero and the completion status identification data is received will the CPU release the timing isolation lock. After the timing isolation lock is released, the CPU rereads the current operating rate data and operating temperature data of the underlying physical production capacity unit, extracts the historical change slope within the new drift time window and generates a state drift vector, calibrates the real-time spatial state point, and releases the updated reference state point. Subsequent orders re-execute the single-value impedance evaluation based on this reference state point, so that the virtual resource scheduling track is continuously updated according to the wear status of physical equipment, parameter drift and actual completion progress, and the configuration sequence remains within the preset control impedance range.

[0063] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A collaborative configuration method for aluminum processing orders based on digital twins, characterized in that, Includes the following steps: Step S101: Obtain the set of aluminum processing orders to be configured, and translate the alloy grade data and demand output data contained in each aluminum processing order in the set of aluminum processing orders to be configured into a target state sequence containing process constraint parameters. Step S102: Extract the historical change slope of the output parameters of the virtual state evolution model according to the preset drift time window to generate a state drift vector. Use the state drift vector to calibrate the position coordinates of the real-time spatial state point in the virtual space and output the reference state point. Step S103: Based on the absolute difference between the current parameter in the reference state point and the corresponding process constraint parameter in the target state sequence, the weighted combination of the preset state recovery coefficient in the virtual state evolution model is applied and summed to calculate the single-valued impedance evaluation value between the reference state point and the starting point of each target state sequence. Step S104: When the difference between the single-valued impedance evaluation values ​​of two adjacent target state sequences is less than the set resolution threshold, extract the distribution density of the mode switching frequency parameter in each target state sequence, select the target state sequence corresponding to the maximum distribution density as the target configuration result, and assign the target configuration result to the underlying physical production capacity unit data interface of the virtual state evolution model image to update the virtual resource scheduling track.

2. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, Following step S104, the following steps are also included: Step S105, calculate the time span of the target state sequence on the underlying physical capacity unit based on the target configuration result, and establish a timing isolation lock within the virtual state evolution model to suspend the calculation process of the single-value impedance evaluation value of subsequent rounds of orders; Step S106, during the operation of the timing isolation lock, continuously monitor the signal flow received by the data interface of the underlying physical capacity unit until the completion status identification data fed back by the underlying physical capacity unit is obtained, then release the timing isolation lock and release the updated reference state point.

3. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, Step S101 includes the following sub-steps: Step S1011, extract alloy grade data and demand output data from each aluminum processing order; Step S1012, convert the alloy grade data and demand output data into multi-dimensional feature vectors as process constraint parameters in the corresponding target state sequence.

4. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, In step S102, the output parameters of the virtual state evolution model include the current operating rate data and operating temperature data of the underlying physical production capacity unit, and the operating environment temperature is set to 20℃ to 40℃.

5. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, Step S103 includes the following sub-steps: Step S1031, calculate the Euclidean distance between the reference state point and each target state sequence in the virtual topology space; Step S1032, obtain the over-temperature variable delay parameter of the underlying physical power generation unit, and superimpose the weights of the Euclidean distance and the over-temperature variable delay parameter to generate a single-value impedance evaluation value.

6. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, Secondary constraint parameters include mold change frequency parameters, cleaning time consumption parameters, and energy consumption rate parameters corresponding to each aluminum processing order.

7. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 1, characterized in that, In step S104, extracting the distribution density of the mode switching frequency parameter in each target state sequence includes the following sub-steps: Step S1041, constructing a multidimensional parameter topology space with the mode switching frequency as the dimension axis; Step S1042, when the difference of the single-value impedance evaluation value is less than the set resolution threshold, calculating the local kernel density estimate of each target state sequence in the multidimensional parameter topology space as the distribution density.

8. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 2, characterized in that, Step S105 includes the following sub-steps: Step S1051, extract the aluminum material specification parameters, rolling pass constraints and heat treatment isothermal time data corresponding to the target configuration result; Step S1052, calculate the processing time per ton by combining the rated power data of the underlying physical capacity unit and the current state attenuation coefficient, and multiply the processing time per ton by the required output to output the time span.

9. The method for collaborative configuration of aluminum processing orders based on digital twins according to claim 2, characterized in that, In step S105, establishing a timing isolation lock within the virtual state evolution model includes the following sub-steps: Step S1053, starting a delay countdown thread within the virtual state evolution model, and setting the initial time of the delay countdown thread to the time span; Step S1054, before the delay countdown thread reaches zero, based on the timing blocking factor... Intercept subsequent computation requests, among which, This is the current closing state constant of the timing isolation lock.

Images