An intelligent digital twin industrial big data simulation verification system

Abstract

This invention discloses an intelligent digital twin industrial big data simulation and verification system, relating to the fields of industrial digital twins and industrial big data technology. It includes: a data acquisition module for real-time acquisition of multi-source heterogeneous data streams from industrial production sites, including equipment sensor data, process parameter streams, and external environmental variables; and a data preprocessing module for format standardization, missing value imputation, and abnormal data filtering of the multi-source heterogeneous data streams acquired by the data acquisition module. This invention effectively solves the problems of inaccurate synchronization between simulation and physical entities, cumulative error propagation, and lack of data temporal and causal co-verification in existing digital twin systems by constructing a complete technical closed loop through multi-module collaboration. First, the data acquisition module comprehensively acquires multi-source heterogeneous data; then, the preprocessing module optimizes data quality; next, the consistency verification module achieves data temporal alignment; and finally, the causal correlation analysis module constructs logical data relationships.

Description

Technical Field

[0001] This invention relates to the field of industrial digital twin and industrial big data technology, specifically to an intelligent digital twin industrial big data simulation and verification system. Background Technology

[0002] Digital twin technology, by constructing virtual mappings of physical entities and combining real-time data-driven simulation analysis, has become a core support for the intelligent upgrading of industries. This system enables dynamic simulation, condition monitoring, and predictive maintenance of production processes, effectively improving the visibility and controllability of industrial systems. Industrial big data, as a key driving element of digital twins, is massive in scale, complex in structure, and requires high real-time performance, placing higher demands on the accuracy and timeliness of simulation models. Current technologies generally focus on data acquisition and model building, but there is still room for optimization in terms of the depth of virtual-real interaction and dynamic response capabilities.

[0003] However, existing digital twin systems generally suffer from real-time synchronization inaccuracies between simulation models and physical entities when dealing with the high-frequency dynamic injection of multi-source heterogeneous industrial big data. Due to the sudden, multimodal, and time-series interleaved characteristics of industrial field data flows, traditional simulation engines struggle to avoid the spread and propagation of accumulated errors during data-driven iterative calculations. These errors cause the virtual model to gradually deviate from the actual operating state, rendering simulation-based decision verification and optimization suggestions unreliable. Especially when the system needs to simultaneously process equipment sensor data, process parameter streams, and external environmental variables, existing technologies lack a collaborative verification mechanism for data temporal consistency and causal relationships. This leads to simulation results lagging or even deviating from reality in dynamic scenarios, severely limiting the application effectiveness of digital twins in real-time control and precise decision-making. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent digital twin industrial big data simulation and verification system, which solves the problems of inaccurate real-time synchronization between digital twin systems and physical entities, lack of data timing and causal co-verification, and cumulative error propagation in existing technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent digital twin industrial big data simulation and verification system, comprising: The data acquisition module is used to collect multi-source heterogeneous data streams from the industrial production site in real time, including equipment sensor data, process parameter streams, and external environmental variables; The data preprocessing module is used to perform format unification, missing value filling, and abnormal data filtering on the multi-source heterogeneous data streams collected by the data acquisition module to obtain a preprocessed data stream. The data consistency verification module is used to perform time alignment processing on each data sub-stream in the preprocessed data stream; and to determine the time consistency between each data sub-stream based on the arrival timestamps and data association identifiers of each time-aligned data sub-stream. The causal relationship analysis module is used to analyze the business logic relationship and physical coupling relationship between each data sub-stream after the time sequence consistency verification module has completed the verification, and to construct a data causal relationship graph. The collaborative verification module is used to perform collaborative verification on the preprocessed data stream based on the time-series consistency results determined by the data consistency verification module and the causal relationship graph constructed by the causal relationship analysis module, to remove or correct low-confidence data and generate a high-confidence industrial big data set. The model dynamic update module is used to receive the high-confidence industrial big data generated by the collaborative verification module, analyze the key physical quantities and state variables in the dataset, and dynamically adjust the internal parameters and states of the digital twin simulation model based on the high-confidence industrial big data. The simulation execution module is used to perform dynamic simulation calculations of the industrial production process based on the digital twin simulation model adjusted by the model dynamic update module. The simulation result verification module is used to compare and analyze the simulation results generated by the simulation execution module with a high-confidence industrial big data set to determine the accuracy and real-time synchronization of the simulation results. The decision support module receives accurate simulation results determined by the simulation result verification module and generates adjustment suggestions for the industrial production process based on the accurate simulation results.

[0006] Furthermore, the data acquisition module is used to acquire real-time operating parameters of field sensors through an industrial IoT interface; acquire production line process parameters through an industrial control system interface; acquire material supply and order data through an enterprise resource planning system interface; and acquire external environmental data through an environmental monitoring system interface.

[0007] Furthermore, the data preprocessing module is used to perform data type conversion on equipment sensor data, process parameter streams, and external environmental variables; fill missing values ​​using multinomial regression interpolation; and identify and remove abnormal data using statistical methods.

[0008] Furthermore, the data consistency verification module is used to assign a global timestamp to each data sub-stream; perform linear interpolation or resampling processing on each data sub-stream based on the global timestamp to synchronize the time dimension; calculate the time deviation between each data sub-stream after time alignment; and determine that each data sub-stream has time consistency when the time deviation is less than a preset threshold.

[0009] Furthermore, the causal relationship analysis module is used to analyze the lag and leading relationships of each data sub-stream in the time dimension using Granger causality test or mutual information method; identify the direct and indirect causal effects between each data sub-stream; and construct a data causal relationship graph.

[0010] Furthermore, the collaborative verification module is used to mark data points with inconsistent time series based on the time series consistency verification results; identify data associations that contradict known causal relationships based on the causal association graph; prioritize the marked data points and identified data associations; and perform collaborative verification on the preprocessed data stream based on the priority ranking to generate a high-confidence industrial big data set.

[0011] Furthermore, the model dynamic update module is used to receive a high-confidence industrial big data set, analyze the key physical quantities and state variables in the data set; update the motion equation parameters of the digital twin simulation model using Kalman filtering or extended Kalman filtering algorithms; and adjust the behavioral logic parameters of the digital twin simulation model using a deep learning model.

[0012] Furthermore, the simulation execution module is used to simulate the operating status of industrial equipment in a virtual environment based on the digital twin simulation model adjusted by the model dynamic update module; to simulate the material flow and energy conversion process of industrial processes; and to simulate the response characteristics of the production line under different operating conditions.

[0013] Furthermore, the simulation result verification module is used to compare the simulation output data generated by the simulation execution module with the actual operating data in a high-reliability industrial big data set point by point; calculate the mean square error or mean absolute error between the simulation output data and the actual operating data; determine the accuracy of the simulation result when the error value is less than a preset threshold; and determine the real-time synchronization of the simulation result by monitoring the time delay between the simulation output data and the actual operating data.

[0014] Furthermore, the decision support module is used to analyze the wear trend of industrial equipment based on accurate simulation results; predict production bottlenecks and failure risks; determine the target operating parameter settings of the equipment through a multi-objective optimization algorithm under the set production targets; and generate industrial production process adjustment suggestions containing the target operating parameter settings.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention effectively solves the problems of inaccurate synchronization between simulation and physical entities, cumulative error propagation, and lack of data temporal and causal co-verification in existing digital twin systems by constructing a complete technical closed loop through multi-module collaboration. First, a data acquisition module comprehensively acquires heterogeneous data from multiple sources. A preprocessing module optimizes data quality, and a consistency verification module aligns the data temporally. A causal correlation analysis module constructs logical data relationships, and a co-verification module generates a high-reliability dataset, mitigating data errors at the source. A model dynamic update module uses algorithms such as Kalman filtering to adjust model parameters in real time, avoiding the accumulation of deviations caused by fixed parameters. A simulation execution and verification module ensures accurate synchronization of simulation results, and a decision support module generates production adjustment suggestions based on reliable simulation results. Overall, this improves the dynamic response capability and decision reliability of the digital twin system, enhancing the efficiency of precise control in industrial production. Attached Figure Description

[0016] Figure 1 This is a system structure diagram of the present invention; Figure 2 This is a closed-loop diagram of simulation model update and decision support in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1-2 This invention provides an intelligent digital twin industrial big data simulation and verification system, comprising: The data acquisition module is used to collect multi-source heterogeneous data streams from the industrial production site in real time, including equipment sensor data, process parameter streams, and external environmental variables; The data preprocessing module is used to perform format unification, missing value filling, and abnormal data filtering on the multi-source heterogeneous data streams collected by the data acquisition module to obtain a preprocessed data stream. The data consistency verification module is used to perform time alignment processing on each data sub-stream in the preprocessed data stream; and to determine the time consistency between each data sub-stream based on the arrival timestamps and data association identifiers of each time-aligned data sub-stream. The causal relationship analysis module is used to analyze the business logic relationship and physical coupling relationship between each data sub-stream after the time sequence consistency verification module has completed the verification, and to construct a data causal relationship graph. The collaborative verification module is used to perform collaborative verification on the preprocessed data stream based on the time-series consistency results determined by the data consistency verification module and the causal relationship graph constructed by the causal relationship analysis module, to remove or correct low-confidence data and generate a high-confidence industrial big data set. The model dynamic update module is used to receive the high-confidence industrial big data generated by the collaborative verification module, analyze the key physical quantities and state variables in the dataset, and dynamically adjust the internal parameters and states of the digital twin simulation model based on the high-confidence industrial big data. The simulation execution module is used to perform dynamic simulation calculations of the industrial production process based on the digital twin simulation model adjusted by the model dynamic update module. The simulation result verification module is used to compare and analyze the simulation results generated by the simulation execution module with a high-confidence industrial big data set to determine the accuracy and real-time synchronization of the simulation results. The decision support module receives accurate simulation results determined by the simulation result verification module and generates adjustment suggestions for the industrial production process based on the accurate simulation results.

[0019] Specifically, this implementation uses an automotive engine block machining production line as an application scenario. This scenario involves high-frequency interaction of heterogeneous data from multiple sources, including equipment sensor data, process parameters, and external environmental variables. Traditional systems are prone to problems such as synchronization discrepancies between simulation and physical entities, and the spread of accumulated errors. The data acquisition module captures real-time sensor data such as vibration, temperature, and rotational speed from over 200 machining equipment on the production line. It obtains process parameters such as cutting speed, feed rate, and depth of cut for cylinder block machining through the industrial control system. Simultaneously, it collects external environmental variables such as workshop temperature, humidity, and air pressure, as well as related data such as material supply cycles and order priorities. This data undergoes format standardization processing by the data preprocessing module, converting hexadecimal binary data output from different equipment into decimal values. For missing values ​​caused by occasional sensor interruptions, multinomial regression interpolation is used to fill in the missing values. Statistical methods are employed to screen and remove abnormal data caused by equipment failure or signal interference, ensuring the initial reliability of the preprocessed data stream.

[0020] The data consistency verification module assigns a global timestamp based on the NTP protocol to each data sub-stream. Addressing the difference between the device sensor data acquisition frequency of 100Hz and the process parameter stream acquisition frequency of 50Hz, it resamples the process parameter stream using linear interpolation to ensure time synchronization across all data sub-streams. Subsequently, it calculates the time deviation between each data sub-stream and sets a reasonable preset threshold to ensure time consistency, thus resolving the simulation lag problem caused by timing interleaving in traditional systems.

[0021] After completing the temporal consistency verification, the causal correlation analysis module uses Granger causality tests to analyze the temporal correlation of each data sub-stream. For example, it identifies that changes in cutting speed precede changes in the surface roughness of the cylinder block, clarifying the leader-lag relationship between the two. Simultaneously, combining the physical laws of industrial production, it identifies the direct causal impact of tool wear on machining accuracy, as well as the indirect causal relationship where ambient temperature indirectly affects machining accuracy by influencing tool thermal expansion. Ultimately, it constructs a multi-level data causal correlation map that includes equipment parameters, process indicators, and environmental factors.

[0022] The collaborative verification module combines temporal consistency results with causal relationship graphs to identify anomalies where the cutting speed data and machining accuracy data are inconsistent in time. It also identifies contradictory data correlations where increasing ambient temperature does not necessarily lead to a decrease in machining error. Data is prioritized based on its impact on simulation results, with anomalies related to core machining accuracy addressed first. Low-confidence data is corrected through cross-validation, and invalid data that cannot be corrected is removed, generating a high-confidence industrial dataset to provide high-quality data support for subsequent simulations.

[0023] The model dynamic update module analyzes key physical quantities such as tool wear and processing pressure from high-reliability industrial big data sets, as well as state variables such as equipment operating status and production progress. It uses the Kalman filter algorithm to dynamically adjust the motion equation parameters of the digital twin simulation model and uses a deep learning model to optimize the behavioral logic parameters of the model. This enables the simulation model to keep up with the operating status of the physical production line in real time, avoiding the accumulation of deviations caused by the fixed parameters of traditional models.

[0024] The simulation execution module is based on the updated digital twin simulation model. In the virtual environment, it accurately simulates the operating status of equipment such as the rotation of the machine tool spindle and the cutting action of the cutting tool. It reproduces the material flow process of the cylinder from blank to finished product and the energy conversion during the cutting process. At the same time, it simulates the production line response under different working conditions, such as the fluctuation of production efficiency when the material hardness changes and the order batch is adjusted, so as to realize the comprehensive dynamic simulation of the production process.

[0025] The simulation result verification module compares the simulation output data, such as processing accuracy and production efficiency, with actual data from a high-reliability industrial big data dataset point by point, calculating the mean square error and mean absolute error of both. When the error value is less than a preset threshold, the accuracy of the simulation results is confirmed. By monitoring the time delay between simulation data and actual data, the real-time synchronization of simulation results is ensured, solving the problem of simulation being disconnected from reality in traditional systems.

[0026] Based on accurate simulation results, the decision support module analyzes tool wear trends to predict remaining service life, identifies production bottlenecks caused by excessive processing time in certain steps, and identifies failure risks due to abnormal equipment vibration. Under the production objectives of meeting product quality requirements and order delivery deadlines, it uses a multi-objective optimization algorithm to determine optimal equipment operating parameters such as cutting speed and feed rate. This generates industrial production process adjustment suggestions, including parameter adjustment schemes, tool replacement recommendations, and process optimization strategies, helping the production line achieve precise control and efficiency improvement.

[0027] In this embodiment, the data acquisition module is used to acquire real-time operating parameters of field sensors through an industrial IoT interface; acquire production line process parameters through an industrial control system interface; acquire material supply and order data through an enterprise resource planning system interface; and acquire external environmental data through an environmental monitoring system interface.

[0028] Specifically, the data acquisition module achieves comprehensive acquisition of multi-source data through multi-interface fusion, employing corresponding interface solutions for different data types in the industrial production site. The Industrial Internet of Things (IIoT) interface uses the Modbus TCP protocol to establish communication with field sensors, acquiring real-time operating parameters such as spindle vibration, acceleration, bearing temperature, and motor current—parameters that directly reflect the equipment's operating status. The industrial control system interface connects to the production line's PLC system, reading process parameters such as cutting speed, feed rate, and depth of cut during cylinder machining via the OPCUA protocol, ensuring real-time synchronization of process data with the production process.

[0029] The Enterprise Resource Planning (ERP) system interface connects to the enterprise's ERP system via a RESTful API to obtain data such as the delivery cycle, product model, material inventory, and quantity of current production orders, providing a basis for production scheduling and resource allocation. The environmental monitoring system interface uses wireless sensor network technology to collect external environmental data such as temperature, humidity, air pressure, and dust concentration within the workshop. While this data does not directly participate in the processing, it indirectly affects equipment operational stability and product quality.

[0030] For example, in an automotive engine block machining production line, the Industrial Internet of Things (IIoT) interface collects spindle vibration data every 100 milliseconds, the Industrial Control System (ICS) interface acquires process parameters every 500 milliseconds, the Enterprise Resource Planning (ERP) system interface synchronizes material inventory data every hour, and the Environmental Monitoring System (EMS) interface collects temperature and humidity data every minute.

[0031] In this embodiment, the data preprocessing module is used to convert the data types of equipment sensor data, process parameter streams, and external environmental variables; fill missing values ​​using multinomial regression interpolation; and identify and remove abnormal data using statistical methods.

[0032] Specifically, the data preprocessing module addresses the characteristics of multi-source heterogeneous data streams by performing data processing in steps. First, it performs data type conversion, converting the hexadecimal raw data output from the equipment sensors into decimal values, and converting the string data in the process parameter stream into numerical data, ensuring that all data has a unified data format for easier subsequent unified processing and analysis.

[0033] For missing values ​​generated during data acquisition due to sensor malfunctions or signal transmission interruptions, multinomial regression interpolation is used to fill in the missing values. The formula is as follows: ; in, The missing data values ​​obtained through interpolation. For time variables, These are the coefficients of the polynomial regression. The order of the polynomial is typically 3 to 5. The regression coefficients are obtained by fitting a polynomial to five consecutive valid data points before and after the missing data. This allows for the calculation of values ​​at missing locations, ensuring data continuity.

[0034] Outliers in the data are identified and removed using the 3σ criterion. First, the mean of a given data substream is calculated. and standard deviation When data points satisfy If a data point is deemed an outlier, it is removed. For example, in spindle vibration data, if the vibration acceleration value at a certain moment far exceeds the normal fluctuation range, it is determined to be an outlier by the 3σ criterion and is directly removed from the data stream to avoid interference from outlier data with subsequent processing results.

[0035] In this embodiment, the data consistency verification module is used to assign a global timestamp to each data sub-stream; perform linear interpolation or resampling on each data sub-stream based on the global timestamp to synchronize the time dimension; calculate the time deviation between each data sub-stream after time alignment; and determine that each data sub-stream has time consistency when the time deviation is less than a preset threshold.

[0036] Specifically, the data consistency verification module focuses on time alignment to ensure the synchronization of each data sub-stream in the time dimension. First, a global timestamp is assigned to all incoming data streams, and the local time of each acquisition node is synchronized with the system's global time using a network time protocol, so that each piece of data has a unified standard time identifier, eliminating time deviations between different acquisition devices.

[0037] To address the issue of inconsistent acquisition frequencies among different data substreams, linear interpolation or resampling is performed based on the global timestamp. When the acquisition frequency of a certain data substream is lower than that of other related data substreams, linear interpolation is used to supplement data points, with the following formula: ; in, and The timestamps of two adjacent valid data points and For the corresponding data stream values, The timestamp of the point to be interpolated. The values ​​are obtained through interpolation. When the data substream acquisition frequency is too high, a resampling method is used to extract data points at the target frequency to ensure that the time intervals of all data substreams remain consistent.

[0038] After timing alignment, the time deviation between each data sub-stream is calculated. The time deviation is calculated by averaging the absolute values ​​of the timestamp differences of the corresponding data points. A preset threshold is set, which is determined based on the real-time requirements of the production scenario. For example, it is set to 50 milliseconds in a high-precision processing scenario. When the calculated time deviation is less than this threshold, the data sub-streams are considered to have timing consistency.

[0039] For example, in an automotive engine block machining production line, the data acquisition frequency of the equipment sensors is 100Hz, while the acquisition frequency of the process parameter stream is 50Hz. By resampling, the frequency of the equipment sensor data is adjusted to 50Hz, ensuring that the time interval between the two is 20 milliseconds. The time deviation between the two types of data after adjustment is calculated. If the average deviation is 30 milliseconds, which is less than the preset threshold of 50 milliseconds, then the timing of the two data sets is confirmed to be consistent.

[0040] In this embodiment, the causal association analysis module is used to analyze the lag and leading relationships of each data sub-stream in the time dimension using Granger causality test or mutual information method; identify the direct causal influence and indirect causal influence between each data sub-stream; and construct a data causal association graph.

[0041] Specifically, the causal relationship analysis module, based on time-series consistency verification, delves deeper into the intrinsic relationships between various data sub-streams. It employs the Granger causality test to analyze the lag and leading relationships between data sub-streams over time. The core idea is that if historical data of variable A significantly improves the prediction accuracy of variable B, then A is considered a Granger cause of B. The F-statistic formula for the Granger causality test is as follows: ; in, To constrain the model's residual sum of squares, i.e., the residual sum of squares of the predicted variable B without considering variable A, This is the sum of squared residuals of the unconstrained model, i.e., the sum of squared residuals of the predicted variable B when variable A is considered. Let A be the lag order of variable A. Let be the sample size. The F-statistic is calculated and compared to the critical value. If the F-statistic is greater than the critical value, the null hypothesis is rejected, and variable A is considered a Granger cause of variable B.

[0042] For data substreams with nonlinear relationships, the mutual information method is used to analyze their correlation degree; the larger the mutual information value, the stronger the correlation between the two. During the analysis, not only are direct causal effects identified, such as changes in cutting speed directly leading to changes in machining efficiency, but indirect causal effects are also identified through hierarchical analysis. For example, increased ambient temperature leads to thermal expansion of the tool, which in turn affects machining accuracy; the relationship between ambient temperature and machining accuracy is an indirect causal relationship.

[0043] Based on the analysis results, a data causal relationship graph was constructed. In the graph, nodes represent various data sub-flows, such as cutting speed, machining accuracy, tool wear, and ambient temperature. Directed edges represent causal relationships, and edge weights indicate the strength of the causal influence. For example, in the automotive engine block machining scenario, Granger causality tests determined that cutting speed is a Granger cause of machining accuracy. Mutual information analysis showed that tool wear and machining accuracy have high mutual information values. The final graph clearly presents the direct and indirect causal relationships between the various data sub-flows.

[0044] In this embodiment, the collaborative verification module is used to mark data points with inconsistent time series based on the time series consistency verification results; identify data associations that contradict known causal relationships based on the causal association graph; prioritize the marked data points and identified data associations; and perform collaborative verification on the preprocessed data stream based on the priority ranking to generate a high-confidence industrial big data set.

[0045] Specifically, the collaborative verification module combines the temporal consistency verification results with the causal relationship graph to conduct multi-dimensional data verification. First, based on the temporal consistency verification results, all data points with time deviations exceeding a preset threshold are marked. For example, if the time deviation between the cutting speed data and the machining accuracy data at a certain moment is 80 milliseconds, exceeding the preset threshold of 50 milliseconds, this data point is marked as a temporally inconsistent data point.

[0046] Subsequently, based on the causal relationship graph, data associations that contradict known causal relationships are identified. For example, according to the causal relationship graph, machining efficiency should increase accordingly when cutting speed increases. If a data segment shows a situation where machining efficiency decreases when cutting speed increases, and there are no other reasonable interfering factors, then the data association is determined to contradict the known causal relationship.

[0047] The marked data points and identified data associations are prioritized based on their impact on the accuracy of the simulation results. Data anomalies directly related to core production indicators such as processing accuracy and production efficiency are given higher priority than minor data anomalies such as slight fluctuations in environmental humidity; data associations with contradictory direct causal relationships are given higher priority than data associations with contradictory indirect causal relationships.

[0048] Collaborative verification is conducted based on priority ranking results. For high-priority data points with inconsistent time sequences, cross-validation is performed with valid data from adjacent time points. If the data deviation is within a correctable range, it is corrected; if the deviation is too large to be corrected, it is discarded. For data associations that contradict known causal relationships, verification is conducted in conjunction with the actual situation on the production site. If it is a data acquisition error, the data is discarded; if it is a sudden interference factor, the interference data is supplemented to explain the situation.

[0049] For example, in the automotive engine block machining production line, the timing of high-priority machining accuracy data and cutting speed data was inconsistent. After cross-validation, 30% of the data was corrected and 10% of the data that could not be corrected was eliminated. For data correlations where cutting speed and machining efficiency were contradictory, it was verified that they were caused by temporary sensor failures, and the erroneous data in that period was eliminated.

[0050] In this embodiment, the model dynamic update module is used to receive a high-confidence industrial big data set, analyze the key physical quantities and state variables in the data set, update the motion equation parameters of the digital twin simulation model using Kalman filtering or extended Kalman filtering algorithms, and adjust the behavioral logic parameters of the digital twin simulation model using a deep learning model.

[0051] Specifically, the model dynamic update module is driven by a high-reliability industrial big data set to achieve real-time optimization of the digital twin simulation model. First, it analyzes the key physical quantities and state variables in the dataset. Key physical quantities include physical parameters that directly affect the production process, such as tool wear and machining pressure / cutting force. State variables include variables reflecting the system's operating status, such as normal equipment operation status, early warning of faults, material supply status (sufficient or insufficient), and production progress.

[0052] The Kalman filter algorithm is used to update the motion equation parameters of the digital twin simulation model. The Kalman filter algorithm consists of two stages: prediction and update. The state equation and observation equation in the prediction stage are as follows: ; ; in, Let be the system state vector at time k. Here is the state transition matrix. Let k be the system state vector at time k-1. To control the input matrix, Let k be the control input vector at time k. The process noise vector follows a mean of 0 and a variance of . Gaussian distribution, The observation vector at time k represents the actual data in a high-reliability industrial big data dataset. For the observation matrix, The observed noise vector follows a mean of 0 and a variance of . The Gaussian distribution.

[0053] The Kalman gain and state update formulas for the update phase are as follows: ; ; ; in, The Kalman gain at time k, Let be the prior error covariance matrix at time k. For the transpose of the observation matrix, To observe the noise variance, Let k be the prior state estimate at time k. Let k be the posterior state estimate at time k. It is the identity matrix. Let be the posterior error covariance matrix at time k.

[0054] For nonlinear digital twin simulation models, an extended Kalman filter algorithm is employed. By performing Taylor expansions on the state and observation equations and ignoring higher-order small terms, the nonlinear system is approximated as a linear system before filtering and updating. Simultaneously, deep learning models, such as Long Short-Term Memory (LSTM) networks, are used to adjust the behavioral logic parameters of the digital twin simulation model. LTM networks learn data patterns from high-reliability industrial big data sets to optimize the model's behavioral logic regarding equipment response, process adjustment, and environmental adaptation, enabling the model to more accurately simulate the complex behavior of physical entities.

[0055] For example, in the production line for machining automotive engine cylinder blocks, the evolution parameters of tool wear in the model are updated by using the Kalman filter algorithm, so that the model can reflect the actual wear state of the tool in real time; the production efficiency calculation logic under different combinations of machining parameters in the model is adjusted by using a long short-term memory network, so that the model's response to the optimization of machining parameters is more in line with the actual production situation.

[0056] In this embodiment, the simulation execution module is used to simulate the operating status of industrial equipment in a virtual environment based on the digital twin simulation model adjusted by the model dynamic update module; to simulate the material flow and energy conversion process of industrial processes; and to simulate the response characteristics of the production line under different operating conditions.

[0057] Specifically, the simulation execution module, based on a dynamically updated digital twin simulation model, accurately reproduces the industrial production process in a virtual environment. The virtual environment is constructed using 3D modeling technology to restore key elements such as the physical layout of the production line, equipment structure, and material morphology, including precise modeling of machine tools, conveyor lines, and storage equipment, ensuring a high degree of consistency between the virtual and physical scenes.

[0058] The virtual environment simulates the operation of industrial equipment, driving the movements of equipment components through the model's motion equations, such as the rotation of machine tool spindles, the feed and cutting actions of cutting tools, and the transmission of conveyor lines, presenting the equipment's operating posture and working status in real time. Simultaneously, it simulates the material flow process in industrial workflows, tracking the transfer, processing, and inspection of raw materials between various stages from their entry into the production line, recreating the changes in material shape, position, movement, and state transitions. For example, it shows the complete flow process of a cylinder block blank after cutting, drilling, and inspection to become a finished product.

[0059] The simulation of energy conversion processes primarily focuses on energy transfer and loss during processing, such as the conversion of electrical energy into mechanical energy during cutting, while simultaneously generating heat loss. The model calculates energy conversion efficiency and loss distribution, providing data support for energy-saving optimization of the production process. For different production line operating conditions, such as changes in material batches, equipment load adjustments, and order priority changes, the simulation simulates the production line's response characteristics. For example, it examines the changing trend of production line processing efficiency, equipment load fluctuations, and product quality stability when material hardness increases.

[0060] For example, in the simulation of an automotive engine block machining production line, the virtual environment accurately reproduced the layout of 10 machining production lines, simulated the operating status of more than 200 pieces of equipment in real time, tracked the flow trajectory of each batch of cylinder blocks from blank to finished product, and calculated the energy consumption during the cutting process. When simulating a 20% increase in material hardness, the simulation results showed response characteristics such as decreased processing efficiency, extended processing time, and a slight increase in equipment vibration amplitude, which highly matched the actual response of the physical production line.

[0061] In this embodiment, the simulation result verification module is used to compare the simulation output data generated by the simulation execution module with the actual operating data in a high-reliability industrial big data set point by point; calculate the mean square error or mean absolute error between the simulation output data and the actual operating data; determine the accuracy of the simulation result when the error value is less than a preset threshold; and determine the real-time synchronization of the simulation result by monitoring the time delay between the simulation output data and the actual operating data.

[0062] Specifically, the simulation result verification module ensures the accuracy and real-time synchronization of the simulation results through multi-dimensional comparative analysis. First, the simulation output data generated by the simulation execution module is compared point by point with the actual operating data in a high-reliability industrial big data set. The comparison includes key data such as equipment operating parameters (vibration, temperature), rotational speed process indicators (processing accuracy, processing efficiency), and production status (production progress, material inventory).

[0063] Calculate the mean square error and mean absolute error between the simulation output data and the actual running data. The formula for the mean square error is: ; in, Mean square error, For the number of data samples, For the i-th actual running data value, Let be the i-th simulation output data value. The mean absolute error formula is: ; in, To calculate the mean absolute error, the parameters are defined in the same way as the mean square error formula. Before calculating the error, the simulation output data and the actual running data are normalized to ensure that their dimensions are consistent. Data with different dimensions and orders of magnitude are converted to the same scale, such as between 0 and 1, to facilitate comprehensive calculation and comparison. The normalization formula is: ; in, These are the normalized data values. The original data values, The minimum value in the dataset. This represents the maximum value in the dataset.

[0064] The preset error threshold is set according to the accuracy requirements of the production scenario. For example, in a high-precision machining scenario, the mean square error threshold is set to 0.05 and the mean absolute error threshold is set to 0.03. When the calculated mean square error and mean absolute error are both less than the corresponding preset threshold, the simulation result is considered to be accurate.

[0065] Real-time synchronization is verified by monitoring the time delay between simulation output data and actual running data. The time delay is the absolute value of the difference between the timestamp of the simulation output data and the timestamp of the corresponding actual running data. A preset threshold for the time delay is set, for example, 100 milliseconds. When the monitored time delay is less than this threshold, the real-time synchronization of the simulation results is confirmed.

[0066] For example, in an automotive engine block machining production line, a comparative analysis of 1000 consecutive data points showed a normalized mean square error of 0.02 and a mean absolute error of 0.01, both less than preset thresholds, indicating good accuracy of the simulation results. Simultaneously, the monitored average time delay was 40 milliseconds, less than the preset threshold of 100 milliseconds, ensuring the real-time synchronization of the simulation results.

[0067] In this embodiment, the decision support module is used to analyze the wear trend of industrial equipment based on accurate simulation results; predict production bottlenecks and failure risks; determine the target operating parameter settings of the equipment through a multi-objective optimization algorithm under the set production target; and generate industrial production process adjustment suggestions containing the target operating parameter settings.

[0068] Specifically, the decision support module provides precise adjustment suggestions for industrial production processes based on verified and accurate simulation results. First, it analyzes the wear trends of industrial equipment by comparing the simulated equipment operating parameters, such as vibration and temperature, with actual operating data over a long period. Combined with the equipment's design lifespan and maintenance standards, it identifies the wear patterns of key components, such as the change in tool wear over processing time and the evolution of bearing vibration amplitude over operating time.

[0069] By combining simulation results with production process data, production bottlenecks and failure risks can be predicted. By simulating the operating status of the production line under different production loads, processes with low processing efficiency and prone to material accumulation are identified as production bottlenecks. Based on the changing trends of equipment operating parameters and historical failure data, the types and timing of potential equipment failures can be predicted. For example, when simulation data shows that the amplitude of spindle vibration continues to increase, bearing wear failure is predicted.

[0070] Under the premise of setting production goals such as achieving a product quality pass rate of 99.5% and reducing order delivery cycle by 10%, a multi-objective optimization algorithm, such as the non-dominated sorting genetic algorithm NSGA-II, is used to determine the target operating parameter settings for the equipment. The NSGA-II algorithm constructs a fitness function and comprehensively considers multiple objectives such as product quality, production efficiency, energy consumption, and cost to search for the optimal combination of equipment operating parameters within the feasible region, such as the optimal combination of cutting speed, feed rate, and depth of cut.

[0071] The generated industrial production process adjustment recommendations include target operating parameter settings for equipment, maintenance plan optimization, and process adjustment strategies. For example, based on tool wear trends, it is recommended to replace the tools after machining 500 cylinder blocks; for bottleneck processes, it is recommended to adjust the equipment operating parameters of that process to improve machining efficiency, or increase equipment investment to optimize process layout; for predicted failure risks, it is recommended to perform bearing maintenance after 1000 hours of equipment operation.

[0072] For example, in an automotive engine block machining production line, the decision support module, based on simulation results analysis, concludes that the tool wear rate is relatively fast and predicts that machining accuracy may not meet standards after machining 450 cylinder blocks. Under the set production target, the NSGA-II algorithm determines the target operating parameters for adjusting the cutting speed to 1200 r / min and the feed rate to 0.2 mm / r, generating adjustment suggestions that include parameter adjustment, tool replacement time points, and process optimization schemes.

[0073] In summary, this invention effectively solves the problems of inaccurate synchronization between simulation and physical entities, cumulative error propagation, and lack of data temporal and causal co-verification in existing digital twin systems by constructing a complete technical closed loop through multi-module collaboration. First, the data acquisition module comprehensively acquires heterogeneous data from multiple sources. The preprocessing module optimizes data quality, and the consistency verification module achieves data temporal alignment. Combined with the causal correlation analysis module, logical data relationships are constructed. The co-verification module generates a high-reliability dataset, mitigating data errors at the source. The model dynamic update module uses algorithms such as Kalman filtering to adjust model parameters in real time, avoiding the accumulation of deviations caused by fixed parameters. The simulation execution and verification module ensures accurate synchronization of simulation results, and the decision support module generates production adjustment suggestions based on reliable simulation results. Overall, this improves the dynamic response capability and decision reliability of the digital twin system, enhancing the efficiency of precise control in industrial production.

[0074] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent digital twin industrial big data simulation and verification system, characterized in that, include: The data acquisition module is used to collect multi-source heterogeneous data streams from the industrial production site in real time, including equipment sensor data, process parameter streams, and external environmental variables; The data preprocessing module is used to perform format unification, missing value filling, and abnormal data filtering on the multi-source heterogeneous data streams collected by the data acquisition module to obtain a preprocessed data stream. The data consistency verification module is used to perform timing alignment processing on each data sub-stream in the preprocessed data stream; Based on the arrival timestamps of each data substream after time alignment and the data association identifier, the time consistency between each data substream is determined; The causal relationship analysis module is used to analyze the business logic relationship and physical coupling relationship between each data sub-stream after the time sequence consistency verification module has completed the verification, and to construct a data causal relationship graph. The collaborative verification module is used to perform collaborative verification on the preprocessed data stream based on the time-series consistency results determined by the data consistency verification module and the causal relationship graph constructed by the causal relationship analysis module, to remove or correct low-confidence data and generate a high-confidence industrial big data set. The model dynamic update module is used to receive the high-confidence industrial big data generated by the collaborative verification module, analyze the key physical quantities and state variables in the dataset, and dynamically adjust the internal parameters and states of the digital twin simulation model based on the high-confidence industrial big data. The simulation execution module is used to perform dynamic simulation calculations of the industrial production process based on the digital twin simulation model adjusted by the model dynamic update module. The simulation result verification module is used to compare and analyze the simulation results generated by the simulation execution module with a high-confidence industrial big data set to determine the accuracy and real-time synchronization of the simulation results. The decision support module receives accurate simulation results determined by the simulation result verification module and generates adjustment suggestions for the industrial production process based on the accurate simulation results.

2. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The data acquisition module is used to acquire real-time operating parameters of field sensors through an industrial IoT interface; acquire production line process parameters through an industrial control system interface; acquire material supply and order data through an enterprise resource planning system interface; and acquire external environmental data through an environmental monitoring system interface.

3. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The data preprocessing module is used to convert the data types of equipment sensor data, process parameter streams, and external environmental variables; fill missing values ​​using multinomial regression interpolation; and identify and remove abnormal data using statistical methods.

4. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The data consistency verification module is used to assign a global timestamp to each data sub-stream; Based on the global timestamp, linear interpolation or resampling is performed on each data substream to synchronize the time dimension; the time deviation between each data substream after time alignment is calculated; when the time deviation is less than a preset threshold, it is determined that each data substream has time consistency.

5. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The causal relationship analysis module is used to analyze the lag and leading relationships of each data sub-stream in the time dimension using Granger causality test or mutual information method; identify the direct and indirect causal effects between each data sub-stream; and construct a data causal relationship graph.

6. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The collaborative verification module is used to mark data points with inconsistent time series based on the time series consistency verification results; identify data associations that contradict known causal relationships based on the causal association graph; and prioritize the marked data points and identified data associations. Based on priority sorting, the preprocessed data stream is collaboratively verified to generate a highly reliable industrial big data set.

7. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The model dynamic update module is used to receive a high-reliability industrial big data set, analyze the key physical quantities and state variables in the dataset, update the motion equation parameters of the digital twin simulation model using Kalman filtering or extended Kalman filtering algorithms, and adjust the behavioral logic parameters of the digital twin simulation model using a deep learning model.

8. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The simulation execution module is used to simulate the operating status of industrial equipment in a virtual environment based on the digital twin simulation model adjusted by the model dynamic update module; and to simulate the material flow and energy conversion process of industrial processes. Simulate the response characteristics of the production line under different operating conditions.

9. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The simulation result verification module is used to compare the simulation output data generated by the simulation execution module with the actual operating data in the high-reliability industrial big data set point by point; and to calculate the mean square error or mean absolute error between the simulation output data and the actual operating data. When the error value is less than a preset threshold, the accuracy of the simulation results is determined. The real-time synchronization of simulation results is determined by monitoring the time delay between simulation output data and actual running data.

10. The intelligent digital twin industrial big data simulation and verification system according to claim 1, characterized in that, The decision support module is used to analyze the wear trend of industrial equipment based on accurate simulation results; predict production bottlenecks and failure risks; determine the target operating parameter settings of the equipment through a multi-objective optimization algorithm under the set production targets; and generate industrial production process adjustment suggestions containing the target operating parameter settings.

Images