Smart factory data management system based on digital twinning

Abstract

The invention relates to the technical field of digital twinning, in particular to a smart factory data management system based on digital twinning. The device comprises an abnormal lagging unit, a window changing unit and a window updating unit. According to the method, a window module is defined to obtain a change speed sensitive coefficient through historical data, and then a data change speed value and a digital twin model update frequency are combined with the change speed sensitive coefficient to calculate an update frequency weight coefficient; the size of the sliding window of the digital twin model is defined by using the data change speed value, the update frequency of the digital twin model, the change speed sensitivity coefficient and the update frequency weight coefficient, and the size of the sliding window is adjusted, so that newest data can be quickly included, model parameters can be updated in time, update lag caused by interference of old data is avoided, and the update efficiency is improved. According to the method, the data utilization efficiency is improved, and when the data change speed is high or low, fine changes can be captured in time while the sliding window is properly enlarged, so that model parameters are updated more accurately.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a smart factory data management system based on digital twins. Background Art

[0002] With the development of Industry 4.0 and smart manufacturing, digital twin technology, by building virtual mappings of physical entities, enables real-time status monitoring and dynamic optimization of smart factory equipment, production lines, and entire plant-wide systems. This enables millisecond-level synchronization of data exchange between virtual models and physical entities, providing a high-fidelity simulation environment for production decision-making. However, smart factory data management systems face a core challenge: multi-source heterogeneous data (such as equipment vibration, energy consumption, and quality inspection images) exhibits significantly differentiated characteristics, with rates of change spanning 3-6 orders of magnitude (e.g., the difference between sensor time series data and process parameter updates). Traditional static modeling methods easily lead to lags in model parameter updates.

[0003] Existing research shows that the timeliness of model parameters directly determines the prediction accuracy of digital twins. When the data sampling interval exceeds the critical value of the physical entity state transition (such as the equipment failure development cycle), the prediction deviation will accumulate in a nonlinear manner. Experimental data show that a parameter update delayed by 10 minutes may increase the error of equipment health assessment by 37%. For this reason, a dynamic sliding window mechanism was introduced to achieve a balance between data freshness and statistical significance by adjusting the window size (configurable from 1 minute to 24 hours): a 60-second window is used for high-frequency vibration data to ensure the capture of state mutations, and an 8-hour window retention period is used for low-frequency environmental parameters. However, a single window strategy is difficult to adapt to mixed data scenarios, and a multi-dimensional optimization function needs to be constructed in combination with the update frequency weight coefficient (adjustable from 0.1-0.9);

[0004] The current technical bottleneck lies in how to achieve adaptive evolution of model update strategies. This requires breaking through the static configuration limitations of parameters such as sliding windows, frequency weights, and anomaly thresholds, and building a dynamic adjustment mechanism with online learning capabilities. Therefore, we provide a smart factory data management system based on digital twins. Summary of the Invention

[0005] The purpose of the present invention is to provide a smart factory data management system based on digital twins to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned object, the present invention provides a smart factory data management system based on digital twins, comprising an abnormal hysteresis unit, a window change unit, and a window update unit;

[0007] As a further improvement of the present technical solution, the change frequency module obtains historical data to calculate the Euclidean norm of the data, and uses the Euclidean norm of the data and the timestamp of the digital twin model update and the timestamp collected in the update exception module to calculate the data change rate value, and then calculates the difference between the digital twin model update timestamp and the timestamp of the historical data and the digital twin model update, and uses the model update timestamp difference to calculate the update frequency of the digital twin model. Abnormal update frequency may indicate potential risks. If the update frequency suddenly drops sharply, it may be due to a failure of the physical entity, a problem with data collection, or a vulnerability in the digital twin model itself. By continuously monitoring and analyzing the update frequency, these potential risks can be discovered in advance, and preventive measures can be taken to avoid serious consequences caused by delayed model parameter updates.

[0008] As a further improvement of the present technical solution, the window definition module obtains the change speed sensitivity coefficient through historical data, and then calculates the update frequency weight coefficient by combining the data change speed value and the digital twin model update frequency with the change speed sensitivity coefficient in the update exception module. The data change speed value, the digital twin model update frequency, the change speed sensitivity coefficient and the update frequency weight coefficient are used to define the sliding window size of the digital twin model. The digital twin model update frequency and the update frequency weight coefficient directly affect the update timing of the digital twin model parameters. By combining these factors with the sliding window size, the update frequency of the digital twin model can be matched with the actual changes in the data. For example, for data with high importance and a large update frequency weight coefficient, a smaller sliding window and a higher update frequency are set to ensure that the digital twin model can update parameters in time according to these key data, thereby avoiding the lag problem caused by update frequency mismatch.

[0009] As a further improvement of the present technical solution, the initial window module integrates the production data of the smart factory collected in the update exception module into an initial data set, sets the condition that the number of collected production data of the smart factory is greater than or equal to the sliding window size of the digital twin model, and selects W consecutive production data collected from the end of the initial data set to form an initial sliding window size of the digital twin model. Using the latest production data to construct the sliding window can make the initial parameters of the digital twin model closer to the actual production situation. In this way, in the initial stage of the digital twin model operation, the error between the output result of the model and the actual production data will be relatively small. The smaller initial error means that the digital twin model does not need to make large-scale parameter adjustments in the subsequent update process, thereby reducing the update lag problem caused by untimely parameter adjustment and improving the stability and accuracy of the model.

[0010] Compared with the prior art, the present invention has the following beneficial effects:

[0011] In this smart factory data management system based on digital twins, a window module is defined to derive a change speed sensitivity coefficient through historical data, and then an update frequency weight coefficient is calculated by combining the data change speed value and the digital twin model update frequency with the change speed sensitivity coefficient in the update exception module. The data change speed value, the digital twin model update frequency, the change speed sensitivity coefficient and the update frequency weight coefficient are used to define the sliding window size of the digital twin model. By adjusting the sliding window size, the latest data can be quickly incorporated, the model parameters can be updated in time, and update delays caused by interference from old data can be avoided, thereby improving data utilization efficiency. At the same time, for data that changes rapidly, a smaller sliding window is used to ensure that the digital twin model can keep up with data changes in a timely manner. For data that changes slowly, the sliding window can be appropriately increased. While ensuring a certain amount of data, subtle changes can also be captured in time, thereby updating the model parameters more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a block diagram of the overall system structure of the present invention;

[0013] Figure 2 It is a module block diagram of the present invention.

[0014] The meaning of each number in the figure is:

[0015] 10. Abnormal hysteresis unit; 101. Update abnormal module; 102. Update hysteresis module;

[0016] 20. Window change unit; 201. Frequency change module; 202. Window definition module;

[0017] 30. Window update unit; 301. Initial window module; 302. Update management module. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] Example 1

[0020] This invention provides a smart factory data management system based on digital twins, please refer to Figure 1-Figure 2 , comprising an abnormal hysteresis unit 10, a window change unit 20 and a window update unit 30;

[0021] The abnormality hysteresis unit 10 includes an update abnormality module 101 and an update hysteresis module 102;

[0022] Update the abnormal module 101 to collect production data of smart factories in real time Input the established digital twin model and record the collected timestamp (i refers to the timestamp of the collection corresponding to the i-th collected smart factory production data) and the number of collected smart factory production data N, the digital twin model needs to collect smart factory production data Update and record the timestamp of the digital twin model update (It can also refer to the timestamp of the most recent parameter update of the digital twin model. Here, i refers to the timestamp of the digital twin model update corresponding to the i-th collected smart factory production data). The absolute value of the timestamp difference is calculated using the collected timestamp and the updated timestamp of the digital twin model. The unit of the absolute value of the timestamp difference, the acquisition timestamp, and the digital twin model update timestamp is seconds.

[0023] When the absolute value of the timestamp difference is known, a delay threshold is set, and the indicator function is used to determine whether the digital twin model has a model update delay anomaly based on the delay threshold τ and the absolute value of the timestamp difference. When the digital twin model is judged to have model update delay anomaly, then when When it is determined that there is no model update delay anomaly in the digital twin model, then Where I(·) is an indicator function, which can convert the complex logical judgment of whether the time interval between each acquisition and model update exceeds the delay threshold into a simple value (0 or 1).

[0024] For example, when analyzing 1,000 collected data, we can quickly identify which data has delay issues; It is clear that the model update time corresponding to the i-th collected smart factory production data exceeds the threshold, which facilitates the staff to conduct targeted investigations;

[0025] The update hysteresis module 102 receives the number of smart factory production data collected in the update abnormality module 101, and satisfies the requirements of the smart factory production data collected by the i-th smart factory production data from the first collected smart factory production data to the N collected smart factory production data. The weighted sum of the collected smart factory production data is calculated to obtain the weighted sum quantity. The weighted sum quantity is divided by the number of collected smart factory production data to obtain the proportion of the difference between collection and update time exceeding the set delay threshold. The specific algorithm formula is: This formula is used to determine the ratio. The digital twin model uses data collection to reflect the real-time status of the physical entity. If this ratio is too high, it indicates that the large amount of production data collected from the smart factory is not updated in a timely manner, and the production status of the smart factory cannot be accurately and real-timely presented. By monitoring this ratio, the digital twin model can be synchronized with the actual smart factory production data, and the smart factory production strategy can be adjusted in a timely manner based on accurate information.

[0026] For example, in a smart factory, the digital twin model of the operating status of production line equipment is monitored. If this ratio suddenly increases, it may mean that there is a problem with the update of the digital twin model, which needs to be checked in time to avoid production failures or increased defective products in the smart factory due to information lag.

[0027] By setting the hysteresis ratio threshold α and the ratio of the acquisition and update time difference exceeding the set delay threshold, it is judged whether the digital twin model has a model parameter update hysteresis problem. When the ratio of the acquisition and update time difference exceeding the set delay threshold is greater than the set hysteresis ratio threshold, it is judged that the digital twin model has a model parameter update hysteresis problem. The specific algorithm formula is: This formula is used to promptly identify and resolve lag issues, ensuring that the digital twin model is always synchronized with actual smart factory production data, maintaining model accuracy.

[0028] For example, we set α = 0.05 (ie 5%), if the calculated If the value of is greater than 0.05, it means that among the N collected smart factory production data, the time interval between data collection and model update of more than 5% exceeds the delay threshold. At this time, it is determined that there is a lag problem in the update of model parameters.

[0029] The window changing unit 20 includes a frequency changing module 201 and a window defining module 202;

[0030] The frequency change module 201 receives the command from the update hysteresis module 102 to determine whether the model parameter update has a hysteresis problem, obtains historical data, and extracts the timestamp from the historical data. Collect production data of smart factories at all times By updating the abnormal module 101 to collect the production data of the smart factory and collect production data from smart factories Calculate the Euclidean norm of data (Euclidean norm is a measure of the "length" or "size" of a vector or data), and then through the Euclidean norm of the data and the timestamp of the digital twin model update Update the timestamp collected in the abnormal module 101 Calculate the data change rate value The formula is used to calculate the data change rate. Recording the data change rate value at each moment in the digital twin model can provide detailed data support for product quality traceability in smart factories.

[0031] For example, when a product produced by a smart factory has quality problems, the production data corresponding to the data change rate value in the digital twin model can be queried to analyze the changes in each link of the production process, identify the cause of the problem and the responsible link, and then take targeted measures to improve it.

[0032] Extract timestamps of digital twin model updates from historical data Timestamp of the digital twin model update in the reconnection update exception module 101 Timestamp of digital twin model updates Calculate the update timestamp difference ΔT of the digital twin model. The specific algorithm formula is:

[0033] The unit of the model update timestamp difference is second / s;

[0034] Calculate the update frequency of the digital twin model using the model update timestamp difference ΔT The digital twin model update frequency (unit: Hz, times / second) indicates the number of model updates per unit time. The numerator value 1 represents the unit time (1 second) and is used to convert the digital twin model update timestamp difference into the standard unit of the digital twin model update frequency f (Hertz, Hz). This formula is used to calculate the digital twin model update frequency. The higher the model update frequency, the more timely the digital twin model can reflect the latest status of physical entities (such as production equipment and production lines). In smart factories, the production process changes rapidly, and the operating parameters of equipment and the flow of materials may change at any time. By increasing the model update frequency, the digital twin model can keep up with the changes in physical entities in real time and provide accurate production information.

[0035] For example, in a smart factory, the temperature and pressure parameters of equipment will change continuously as production progresses. High-frequency model updates can feed these parameter changes into the digital twin model in real time, allowing timely understanding of the equipment's operating status and preventing equipment failures.

[0036] Historical data includes the timestamp of collection Collecting production data from smart factories Timestamp of digital twin model updates

[0037] The definition window module 202 extracts the historical data change speed value v from the historical data ls, the fundamental frequency f of the production data of the smart factory when it is stationary 基线 The fundamental frequency f of Hehui Factory's production data when it is stationary 基线 And the change speed sensitivity coefficient β is obtained, and its specific algorithm formula is The change rate sensitivity coefficient is greater than 0. This formula is used to calculate the change rate sensitivity coefficient. During the production process of a smart factory, production data will have certain normal fluctuations. By reasonably setting the change rate sensitivity coefficient, the digital twin model can effectively distinguish between normal fluctuations and true abnormal changes.

[0038] For example, in the production process of a smart factory, the wear of certain equipment and tools will cause the processing accuracy data to change slowly, which is a normal fluctuation. However, if the processing accuracy data drops sharply in a short period of time and the rate of change exceeds the range set by the sensitivity coefficient, it can be determined as an abnormality, and timely measures can be taken to adjust it to ensure product quality.

[0039] The update frequency weight coefficient γ=f-(v·β) is calculated by combining the data change speed value v and the digital twin model update frequency f in the change frequency module 201 with the change speed sensitivity coefficient β. The update frequency weight coefficient is greater than 0. This formula is used to calculate the update frequency weight coefficient. Smart factories generate massive amounts of production data during the production process. Different types of production data have different importance for production decisions and equipment operation. By setting the update frequency weight coefficient, a higher weight can be given to key data, so that it is updated more frequently.

[0040] For example, the real-time operating parameters of production equipment in a smart factory (such as temperature, pressure, and speed) are crucial to the stable operation of the equipment and product quality. Setting a high weight for this data can ensure that it is updated more frequently and the equipment status is monitored in a timely manner. For relatively less important data, such as non-critical indicators of the factory environment, the update frequency can be reduced to reduce unnecessary resource consumption.

[0041] The sliding window size of the digital twin model is defined by the data change speed value v, the digital twin model update frequency f, the change speed sensitivity coefficient β and the update frequency weight coefficient γ. in, Indicates rounding up, W∈[W min =50,W max =200], this formula is used to define the sliding window size of the digital twin model. The data change rate value and the change rate sensitivity coefficient can help the model keenly perceive the rapid changes in data. When the data change rate exceeds the threshold set by the change rate sensitivity coefficient, the digital twin model can quickly collect the latest collected data by adjusting the sliding window size;

[0042] For example, if the temperature of production equipment in a smart factory suddenly rises during operation and the data changes faster, the sliding window size of the digital twin model can be reduced to focus on the latest temperature data, thereby updating the model parameters in a timely manner and avoiding update lags caused by interference from old data.

[0043] Historical data also includes the historical data change rate value v ls , historical digital twin model update frequency f ls , the fundamental frequency f of the production data of the smart factory when it is stationary 基线 .

[0044] The window updating unit 30 includes an initial window module 301 and an update management module 302;

[0045] The initial window module 301 collects the production data of the smart factory by updating the abnormal module 101 Integrate into initial dataset Among them, x1 refers to the first production data collected from the smart factory, x i It refers to the production data of the i-th smart factory collected. The number N of collected production data of the smart factory is set to be greater than or equal to the sliding window size W of the digital twin model in the definition window module 202, which ensures that the subsequently defined window can select enough data from this data set. The production data of the smart factory collected from the end of the initial data set X is selected continuously. The sliding window size W0 of the initial digital twin model is W0={x i-W+1 ,...,x i};

[0046] For example, if W = 10, N = 100, then W0 contains x 91 to x 100 These 10 collect production data from smart factories;

[0047] The average value is calculated by weighted summing the production data of the smart factory collected in the sliding window of the initial digital twin model and the sliding window size of the digital twin model. Then, the production data of the smart factory is collected in the sliding window of the initial digital twin model. i Square it and get the production data of the smart factory Use Square to collect production data of smart factories The weighted summation of the sliding window size of the digital twin model is used to calculate the average squared value.

[0048] The update management module 302 subtracts the collected timestamps corresponding to each adjacent production data of the smart factory collected in the initial data set in the initial window module 301 to obtain the adjacent timestamp difference, and then compares the adjacent timestamp differences in sequence to find the minimum adjacent timestamp difference. The collected timestamp and the timestamp updated by the digital twin model and the minimum adjacent timestamp difference are used to determine whether to trigger the update condition of the sliding window size of the initial digital twin model. When the value subtracted from the timestamp updated by the digital twin model is greater than or equal to the minimum adjacent timestamp difference, the update condition of the sliding window size of the initial digital twin model is determined to be triggered, and the sliding window size of the initial digital twin model is updated and managed. The oldest production data of the smart factory collected in the sliding window size of the initial digital twin model is removed, and the production data of the smart factory collected most recently is added to the sliding window size of the initial digital twin model, and finally the sliding window size of the updated digital twin model is obtained;

[0049] Suppose we have a digital twin model with a sliding window size W of 3. The initial digital twin model's sliding window size W0 = {x1, x2, x3}, and the corresponding timestamp is {t1, t2, t3}. When the most recently collected production data x4 of the smart factory arrives and the update condition is triggered, the oldest production data x1 of the smart factory is first removed, and then the most recently collected production data x4 of the smart factory is added. The sliding window size of the updated digital twin model is W1 = {x2, x3, x4}.

[0050] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart factory data management system based on digital twins, characterized by: It includes an abnormal hysteresis unit (10), a window change unit (20) and a window update unit (30); The abnormal hysteresis unit (10) collects production data of the smart factory in real time, inputs it into the established digital twin model and updates it, records the timestamp of the digital twin model update to determine whether the digital twin model has a model update delay abnormality, and when a model update delay abnormality exists, calculates the ratio of the difference between the collection and update time exceeding the set delay threshold, and determines whether the digital twin model has a model parameter update hysteresis problem; The window change unit (20) obtains historical data to calculate the data change speed value, the digital twin model update timestamp difference and the change speed sensitivity coefficient, and then calculates the digital twin model update frequency and the update frequency weight coefficient, and uses the update frequency weight coefficient to define the sliding window size of the digital twin model; The window update unit (30) collects the production data of the smart factory from the abnormal hysteresis unit (10) and integrates it into an initial data set, and then uses the initial data set to form a sliding window size of an initial digital twin model to determine whether to trigger the update condition of the sliding window size of the initial digital twin model. When triggered, the sliding window size of the initial digital twin model is updated and managed.

2. The digital twin-based smart factory data management system according to claim 1, characterized in that: The abnormality hysteresis unit (10) includes an abnormality update module (101) and a hysteresis update module (102); The update exception module (101) collects production data of the smart factory in real time, inputs it into the established digital twin model and updates it, records the collected timestamp and the updated timestamp of the digital twin model to calculate the absolute value of the timestamp difference; The indicator function is used to determine whether the digital twin model has model update delay anomalies based on the absolute value of the delay threshold and the timestamp difference.

3. The digital twin-based smart factory data management system according to claim 2 is characterized by: The update hysteresis module (102) receives the number of smart factory production data collected in the update abnormality module (101), performs weighted summation on the number of collected smart factory production data to obtain the weighted sum number, and divides the weighted sum number by the number of collected smart factory production data to obtain the ratio of the difference between the collection and update time exceeding the set delay threshold; By setting the lag ratio threshold and the ratio of the difference between the acquisition and update time exceeding the set delay threshold, it is judged whether the digital twin model has a model parameter update lag problem.

4. The digital twin-based smart factory data management system according to claim 3 is characterized by: The window changing unit (20) includes a frequency changing module (201) and a window defining module (202); The change frequency module (201) obtains historical data to calculate the data Euclidean norm, uses the data Euclidean norm and the timestamp of the digital twin model update and the timestamp collected in the update anomaly module (101) to calculate the data change speed value, and then calculates the digital twin model update timestamp difference ΔT through the historical data and the timestamp of the digital twin model update.

5. The digital twin-based smart factory data management system according to claim 4 is characterized by: The change frequency module (201) calculates the update frequency of the digital twin model using the model update timestamp difference ΔT Among them, the update frequency of the digital twin model represents the number of times the model is updated per unit time. The value 1 in the numerator represents 1 second per unit time. The unit of the update frequency of the digital twin model is Hz, times / second.

6. The digital twin-based smart factory data management system according to claim 5 is characterized by: The definition window module (202) obtains a change speed sensitivity coefficient through historical data, wherein the change speed sensitivity coefficient is greater than 0, and then calculates an update frequency weight coefficient by combining the data change speed value and the digital twin model update frequency with the change speed sensitivity coefficient in the update anomaly module (101), wherein the update frequency weight coefficient is greater than 0.

7. The digital twin-based smart factory data management system according to claim 6, characterized in that: The window definition module (202) defines the sliding window size of the digital twin model using the data change speed value v, the digital twin model update frequency f, the change speed sensitivity coefficient β and the update frequency weight coefficient γ in, Indicates rounding up.

8. The digital twin-based smart factory data management system according to claim 2, characterized in that: The window updating unit (30) includes an initial window module (301) and an updating management module (302); The initial window module (301) integrates the production data of the smart factory collected in the update exception module (101) into an initial data set, sets a condition that the amount of collected production data of the smart factory is greater than or equal to the sliding window size of the digital twin model, and selects W consecutive production data collected from the end of the initial data set to form a sliding window size of an initial digital twin model.

9. The digital twin-based smart factory data management system according to claim 8, characterized in that: The average value is calculated by weighted summing up the production data of the smart factory collected in the sliding window of the initial digital twin model, and then the production data of the smart factory collected in the sliding window of the initial digital twin model is squared to obtain the squared production data of the smart factory. The squared production data of the smart factory collected is weighted summed up to calculate the average value of the square.

10. The digital twin-based smart factory data management system according to claim 9, characterized in that: The update management module (302) subtracts the collected timestamps corresponding to the adjacent production data of the collected smart factory in the initial data set in the initial window module (301) to obtain the adjacent timestamp differences, and then compares the adjacent timestamp differences in sequence to find the minimum adjacent timestamp difference. The collected timestamps and the timestamps updated by the digital twin model and the minimum adjacent timestamp differences are used to determine whether the update condition of the sliding window size of the initial digital twin model is triggered. When triggered, the sliding window size of the initial digital twin model is updated and managed.

Images