An industrial internet of things device state real-time synchronization communication method and system based on digital twinning

Abstract

The application belongs to the technical field of information technology, and specifically discloses an industrial Internet of Things equipment state real-time synchronization communication method and system based on digital twinning, which comprises the following steps: obtaining a global time reference according to a cross-domain time base hard alignment mechanism, and sampling a physical signal to obtain a multi-dimensional state digital sequence according to the global time reference; extracting state features according to the multi-dimensional state digital sequence, and performing iterative calculation on the state features according to a sparse Jacobian matrix to obtain a posterior state estimation vector and a confidence factor; the application aims to solve the problem of how to overcome the virtual-real mapping deviation caused by software and hardware time delay jitter, computing power resource limitation and network channel fluctuation in the high-frequency interaction scene of multi-source data of the industrial Internet of Things, and realize the high real-time, consistent and strict physical boundary constraint deterministic state synchronization between the physical entity and the virtual model.

Description

Technical Field

[0001] This invention relates to the technical field of information technology, specifically to a method and system for real-time synchronization and communication of industrial Internet of Things (IoT) device status based on digital twins. Background Technology

[0002] With the deep integration of the Industrial Internet of Things (IIoT) into modern intelligent manufacturing and high-precision automatic control systems, digital twin technology, as a core means to achieve full lifecycle monitoring, simulation, and prediction of equipment, relies heavily on the real-time, accurate, and causally consistent synchronous mapping of multi-source heterogeneous data between physical entities and virtual models. However, existing IIoT digital twin systems face underlying physical constraints and architectural bottlenecks in practical deployments: First, traditional systems often rely on upper-layer software protocol stacks (such as NTP) for time synchronization, resulting in operating system scheduling delays between sensor sampling actions and network time stamps. This phase drift between physical sampling and computation clocks can cause timing errors and distortions when processing high-frequency multimodal data at the microsecond level, such as high-frequency vibrations and transient currents. Second, facing massive concurrent heterogeneous sensor data, edge computing nodes often employ full... State fusion using local matrix operations (such as full-dimensional extended Kalman filtering) can easily lead to computational congestion or even buffer overflow in industrial microcontrollers with limited computing power, resulting in delayed synchronization responses. In addition, existing forward error correction or retransmission mechanisms in the communication layer usually adopt fixed parameter transmission strategies, which disconnect the channel characteristics of the physical layer from the device dynamics of the application layer. When faced with network jitter or micro-circuit breaks caused by electromagnetic interference in complex workshops, it is impossible to guarantee the success rate of single transmission of critical states that change abruptly. Blind retransmission can lead to control loop failure. Furthermore, when the digital twin application layer encounters packet loss, it often uses pure data-driven or simple interpolation prediction methods, which lack the constraints of physical and mechanical boundaries (such as maximum acceleration and structural strain limits). This can easily lead to extrapolations and visual jumps that violate common sense, weakening the practical value and decision-making credibility of digital twin systems in flexible manufacturing and extreme working conditions. Summary of the Invention

[0003] This invention provides a method and system for real-time synchronization of the status of industrial IoT devices based on digital twins. The purpose is to solve the problem of how existing technologies can overcome the virtual-real mapping deviation caused by hardware and software latency jitter, computing resource limitations, and network channel fluctuations in the high-frequency interaction scenario of multi-source data in industrial IoT, and achieve high real-time, consistent, and deterministic state synchronization between physical entities and virtual models with strict physical boundary constraints.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for real-time synchronization communication of industrial IoT device status based on digital twins includes: obtaining a global time reference according to a cross-domain time base hard alignment mechanism, and sampling physical signals according to the global time reference to obtain a multi-dimensional state digital sequence; extracting state features from the multi-dimensional state digital sequence, and performing iterative calculation on the state features according to a sparse Jacobian matrix to obtain a posterior state estimation vector and a confidence factor; obtaining a controlled trend gradient variable according to physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods; assembling and generating state metadata according to the time reference, spatial parameters, the controlled trend gradient variable, and the confidence factor, and encoding and obtaining a synchronization message according to the state metadata; determining an error correction redundancy ratio according to a measured network latency index and the confidence factor, and performing channel coding transmission on the synchronization message according to the error correction redundancy ratio; parsing the transmitted synchronization message to obtain a true state vector, and obtaining a predicted state vector according to a trajectory extrapolation operator when there is a transmission delay; fusing the predicted state vector and the delayed true state vector according to a smoothing mixing function to obtain a smoothed state vector and drive virtual model rendering.

[0005] In one aspect of the invention, obtaining the global time reference based on the cross-domain time base hard alignment mechanism includes: Based on the clock signal of the local clock module and the preset phase-locked loop processing logic, obtain the aligned trigger input signal; The sampling operation is driven by the trigger input signal, and a hardware timestamp is pushed into each sampled data node to obtain the global time reference.

[0006] In one aspect of the invention, the step of performing iterative calculations on the state features based on the sparse Jacobian matrix to obtain a posterior state estimation vector and a confidence factor includes: The sparse Jacobian matrix is ​​obtained by removing zero elements, and the Kalman gain matrix is ​​obtained by transposing and inverting the sparse Jacobian matrix. The fusion operation is performed based on the Kalman gain matrix and the observation vector to obtain the posterior state estimation vector; The confidence factor is obtained based on the innovation covariance inverse matrix and the observed innovation residual vector.

[0007] In one aspect of the invention, obtaining the controlled trend gradient variable based on physical boundary truncation logic and the posterior state estimation vector over multiple consecutive periods includes: Based on the time mean and state mean of the posterior state estimation vector for the continuous multiple periods, obtain the unconstrained derivative variable; The physical limit benchmark value is established based on the sign parameter of the unconstrained derivative variable, the preset maximum rate of change constant, and the discount factor; The controlled trend gradient variable is obtained by comparing the absolute value of the unconstrained derivative variable with the physical limit benchmark value and performing a truncation operation.

[0008] In one aspect of the invention, determining the error correction redundancy ratio based on the measured network latency index and the confidence factor includes: Based on the comparison result between the network latency index and the preset latency threshold, the step state value is obtained; The error correction redundancy ratio is obtained by performing a combination operation based on the basic redundancy ratio constant, the sensitivity coefficient, the difference parameter of the confidence factor, and the step state value. The error correction redundancy ratio is increased when the confidence factor decreases and network latency exists.

[0009] In one aspect of the invention, after performing channel coding transmission on the synchronization message according to the error correction redundancy ratio, the method further includes: Obtain the receiving timestamp and decoding time constant based on the received control commands; The elastic waiting time variable is obtained based on the network gating parameters and the scheduling cycle constant; The physical action trigger time is obtained by combining the received timestamp, the decoding time constant, and the elastic waiting time variable.

[0010] In one aspect of the invention, fusing the predicted state vector and the delayed-arriving true state vector according to a smoothing mixing function to obtain a smoothed state vector includes: An exponential expression is constructed based on the regression rate constant, the system time variable, and the data arrival time constant to obtain the dynamically changing smooth mixing coefficients; The predicted state vector and the true state vector are weighted and summed according to the smoothing mixing coefficient to obtain the smoothed state vector.

[0011] In another aspect, the present invention also relates to a real-time device status synchronization system based on digital twins, comprising: The data acquisition module is used to acquire a global time base according to the cross-domain time base hard alignment mechanism, and to sample the physical signal according to the global time base to acquire a multi-dimensional state digital sequence. The state extraction module is used to extract state features based on the multidimensional state number sequence, and perform iterative calculations on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor. The gradient calculation module is used to obtain the controlled trend gradient variable based on the physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods. The message generation module is used to assemble and generate state metadata based on the time base, spatial parameters, the controlled trend gradient variable and the confidence factor, and to obtain the synchronization message based on the state metadata encoding. The transmission control module is used to determine the error correction redundancy ratio based on the measured network latency index and the confidence factor, and to perform channel coding transmission on the synchronization message according to the error correction redundancy ratio. The twin rendering module is used to parse the transmitted synchronization message to obtain the real state vector, and when there is a transmission delay, to obtain the predicted state vector according to the trajectory extrapolation operator; and to fuse the predicted state vector and the delayed real state vector according to the smoothing blending function to obtain the smooth state vector and drive the rendering of the virtual model.

[0012] In one aspect of the invention, the state extraction module is specifically used to obtain the sparse Jacobian matrix according to the zero-element removal operation, calculate the Kalman gain matrix according to the sparse Jacobian matrix, then combine the observation vector fusion to obtain the posterior state estimation vector, and obtain the confidence factor according to the innovation covariance inverse matrix.

[0013] In one aspect of the invention, the gradient calculation module is specifically used to obtain unconstrained derivative variables based on the posterior state estimation vectors of the consecutive multiple periods, establish physical limit benchmark values ​​based on the maximum rate of change constant, and then perform a truncation operation on the absolute value of the unconstrained derivative variables to obtain the controlled trend gradient variables.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention introduces a cross-domain information physical time base hard alignment mechanism at the physical sensing layer. By binding the network protocol clock to the trigger pin of the underlying analog-to-digital converter through an on-chip phase-locked loop, it eliminates random latency caused by operating system software scheduling and achieves global time synchronization. At the edge processing layer, a state-space estimation method based on sparse Jacobian matrices is employed, reducing the computational complexity of matrix inversion from cubic to linear, thus mitigating the computational limitations of industrial microcontrollers when processing high-frequency, multi-dimensional data and enabling efficient feature extraction. Finally, controlled physical boundary truncation logic is introduced at the message encapsulation stage. This uses the actual electromechanical limits of the device to constrain the evolution trend of purely mathematical fitting, avoiding divergence caused by sudden noise. The derivative of the property ensures the physical reliability of the transmitted data; in the heterogeneous network transmission layer, the state confidence and forward error correction redundancy rate are coupled across layers, and the error correction rate is adaptively adjusted when the network jitter occurs to ensure the transmission of critical states. With the help of a microsecond-level computing network joint scheduling mechanism, the edge control execution is aligned with the physical time slots of the time-sensitive network to form a real-time closed loop; in the digital twin mapping layer, for the network micro-circuit failure condition, the kinematic trajectory extrapolation constrained by physical boundaries and the time decay smooth reconstruction mechanism maintain the continuity of the virtual space during the communication delay, avoid the model position jump, and build a deterministic virtual-real mapping base with real-time performance, robustness and fidelity in complex industrial environments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram illustrating the steps of a real-time synchronization communication method for the status of industrial IoT devices based on digital twins according to the present invention.

[0017] Figure 2 This is a schematic diagram of sub-step S1 in the real-time synchronization communication method for industrial IoT device status based on digital twins according to the present invention.

[0018] Figure 3 This is a schematic diagram of sub-step S2 in the real-time synchronization communication method for industrial IoT device status based on digital twins of the present invention.

[0019] Figure 4 This is a schematic diagram of sub-step S3 in the real-time synchronization communication method for industrial IoT device status based on digital twins of the present invention.

[0020] Figure 5This is a schematic diagram of sub-step S4 in the real-time synchronization communication method for industrial IoT device status based on digital twins of the present invention.

[0021] Figure 6 This is a schematic diagram of sub-step S5 in the real-time synchronization communication method for industrial IoT device status based on digital twins of the present invention.

[0022] Figure 7 This is a schematic diagram of sub-step S6 in the real-time synchronization communication method for industrial IoT device status based on digital twins of the present invention.

[0023] Figure 8 This is a comparison chart of the end-to-end latency of the present invention after joint scheduling and physical boundary truncation, and the traditional MQTT / JSON+TCP retransmission scheme.

[0024] Figure 9 This is a comparison chart of the dynamic tracking error between the performance of the present invention after joint scheduling and physical boundary truncation and the traditional MQTT / JSON+TCP retransmission scheme. Detailed Implementation The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0025] Please see Figure 1 As shown in the figure, this embodiment discloses a method for real-time synchronization and communication of the status of industrial IoT devices based on digital twins, specifically including: This invention provides a method for real-time synchronization and communication of the status of industrial IoT devices based on digital twins, such as... Figure 1 As shown, in one embodiment, the method includes: S1. Obtain a global time reference based on the cross-domain time base hard alignment mechanism, and sample the physical signal based on the global time reference to obtain a multi-dimensional state digital sequence; S2. Extract state features from the multidimensional state digital sequence, and perform iterative calculations on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor. S3. Based on the physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods, obtain the controlled trend gradient variable; S4. Assemble and generate state metadata based on the time base, spatial parameters, the controlled trend gradient variable and the confidence factor, and obtain the synchronization message based on the state metadata encoding. S5. Determine the error correction redundancy ratio based on the measured network latency index and the confidence factor, and perform channel coding transmission on the synchronization message according to the error correction redundancy ratio; S6. Obtain the true state vector by parsing the transmitted synchronization message. When there is a transmission delay, obtain the predicted state vector by the trajectory extrapolation operator. Fuse the predicted state vector and the delayed true state vector by the smoothing mixing function to obtain a smooth state vector and drive the rendering of the virtual model.

[0026] In this embodiment, it should be noted that, in S1, in order to solve the operating system scheduling delay problem between the physical sampling action of sensors and the network time stamp in the industrial Internet of Things environment, a cross-domain time base hard alignment mechanism is constructed through S1.

[0027] Specifically, traditional synchronization schemes often rely on software protocol stacks for time synchronization, resulting in random phase drift between the physical action and the network time stamp. This scheme, through logical binding of the underlying hardware circuitry, directly connects the output signal of the local clock module maintaining the precise time protocol in the network interface controller to the trigger terminal driving the analog-to-digital converter for sampling, after processing. This logic eliminates the interference of software interrupts, ensuring consistency in the temporal causality of multi-source heterogeneous sensors at the moment of acquisition. It unifies discrete physical sampling points into a single time frame, providing a time-domain fidelity data sequence for subsequent extraction of kinematic features such as velocity and acceleration.

[0028] In S2, to address the limitations of computing power and matrix computation congestion issues that exist when edge computing nodes process massive concurrent multimodal data, S2 introduces state-space estimation and confidence evaluation based on sparse Jacobian matrices.

[0029] Specifically, when dealing with high-frequency, multi-dimensional data, conventional full matrix inversion operations can lead to computational lags of microseconds to milliseconds at edge nodes. This solution generates a sparse Jacobian matrix by pre-identifying weakly coupled terms in the target device's physical topology and performing zero-element removal. This sparse matrix is ​​then used for iterative state calculations, which not only reduces the computational time complexity while maintaining physical constraints but also simultaneously quantifies the fit between the current sensor observation data and the physical model (i.e., the confidence factor). This transforms the original heterogeneous sensor data into a posterior state estimation vector with physical semantics and a confidence metric.

[0030] In S3 and S4, to overcome the problem of extrapolation exceeding mechanical limits that easily occurs when pure mathematical data fitting encounters sudden high-frequency noise, rigid physical boundary truncation logic and semantic message encapsulation are introduced. Specifically, the system extracts continuous periodic state data to calculate unconstrained derivatives and forcibly introduces the maximum rate of change constant determined by the electromechanical parameters of the equipment as a physical limit benchmark. Through numerical comparison and truncation, the calculated controlled trend gradient variable is always kept within the mechanical envelope of the actual equipment. Subsequently, the physically constrained gradient variable, confidence factor, time benchmark, etc., are assembled and binary encoded to generate a synchronization message, ensuring the physical logical rationality of the transmitted data.

[0031] In S5, to address the bandwidth waste or error correction failure caused by the fixed redundancy strategy in existing industrial network transmission under complex channels, S5 couples the data confidence of the application layer with the channel coding strategy of the physical layer across layers.

[0032] Specifically, the system changes the best-effort fixed transmission mode. When the network control plane detects an increase in the delay standard deviation and a decrease in the confidence factor calculated by the application layer, the system dynamically increases the forward error correction redundancy ratio. This calculation logic directionally increases the transmission redundancy of critical states when the communication environment deteriorates, ensuring the single-pass delivery rate for sudden track change states and avoiding delay blind spots caused by blind retransmissions.

[0033] In S6, to address the issues of virtual model stagnation and spatial location jumps caused by data flow delays due to network micro-disruptions, S6 implements a trajectory extrapolation and smoothing hybrid mechanism constrained by physical boundaries.

[0034] like Figure 8 as well as Figure 9 As shown, specifically, when a data packet does not arrive on time, the application layer performs kinematic predictions using the pre-order controlled trend gradient as the initial condition. After the delayed real physical message arrives, the engine does not directly force a location overwrite; instead, it uses a smooth blending function that includes a time variable to smoothly transition the final rendering state from prediction-dominated to reality-dominated. This logic allows the virtual space to operate according to the inertial laws of objects during data gaps and to recover the trajectory within visual blind spots upon data recovery, maintaining the spatiotemporal mapping continuity of the digital twin system.

[0035] like Figure 2 As shown, in one specific embodiment, S1 includes: S11. Obtain the aligned trigger input signal based on the clock signal of the local clock module and the preset phase-locked loop processing logic; S12. Drive the sampling operation according to the trigger input signal, push the hardware timestamp into each sampled data node, and obtain the global time reference.

[0036] In this embodiment, it should be noted that in S11, the system needs to calculate the network transmission asymmetry time offset between the master and slave clock nodes. The calculation formula is as follows:

[0037] The characters in the formula are defined as follows: This represents the asymmetric time offset in network transmission between the master clock node and the slave clock node; Represents the precise physical layer timestamp of the synchronization message sent by the master clock node hardware; Represents the precise physical layer timestamp of the synchronization message received from the clock node hardware; Represents the precise physical layer timestamp from which the delay request message was sent from the clock node hardware; This represents the precise physical layer timestamp indicating when the master clock node hardware receives a delay request message.

[0038] Based on this offset The internal hardware counter automatically adjusts the phase, and after processing by the on-chip phase-locked loop, the aligned trigger input signal is output through the hardware circuit.

[0039] In S12, the analog-to-digital converter synchronously performs high-frequency sampling under the drive of the trigger input signal. Since the trigger signal comes directly from the compensated hardware clock pin, the moment the sampling action occurs is automatically timestamped with the corresponding hardware timestamp, eliminating interference from the operating system and forming a multidimensional state digital sequence with strict time order.

[0040] like Figure 3 As shown, in one specific embodiment, S2 includes: S21. Obtain the sparse Jacobian matrix by performing the zero-element removal operation, and obtain the Kalman gain matrix by performing the transpose and inversion operations of the sparse Jacobian matrix. S22. Perform a fusion operation based on the Kalman gain matrix and the observation vector to obtain the posterior state estimation vector; S23. Obtain the confidence factor based on the innovation covariance inverse matrix and the observed innovation residual vector.

[0041] In this embodiment, it should be noted that before performing the state estimation update, the system first calculates the Mahalanobis distance of the current data using a sliding window to determine whether an update calculation needs to be triggered. The formula for calculating the Mahalanobis distance is:

[0042] The characters in the formula are defined as follows: Indicates time The scalar value of Mahalanobis distance corresponding to the state feature vector; Indicates time A multidimensional state feature column vector obtained from the sensor and converted into a digital sequence; Indicates length is The column vector of mean values ​​of state features within the sliding time window; Indicates length is The inverse matrix of the state feature covariance matrix within the sliding time window; This indicates that a vector or matrix is ​​transposed.

[0043] In S21 and S22, when a calculation is triggered, the system uses the sparse Jacobian matrix obtained by pre-removing zero elements to perform a fusion derivation and calculate the Kalman gain matrix. And output the posterior state estimation vector. The specific formula is as follows:

[0044]

[0045] The characters in the formula are defined as follows: Indicates time The calculated Kalman gain matrix; Indicates time The prior error covariance matrix; This represents the transpose of the sparse observation Jacobian matrix obtained after performing the offline zero-element removal operation; This represents the sparse observation Jacobian matrix obtained after performing the offline zero-element removal operation; Indicates time The observation noise covariance matrix; This indicates that the matrix is ​​inverted. Indicates time The posterior state estimation vector of the fused output; Indicates time The prior state estimation vector; Indicates time The acquired sensor observation vectors; This represents a nonlinear observation function that maps the state space to the observation space.

[0046] In S23, to quantify the reliability of the posterior state, the system calculates the confidence factor. The calculation formula is:

[0047] The characters in the formula are defined as follows: This represents the confidence factor of the calculated output, and its value ranges from 0 to 1. Represented by natural constant Exponential functions with base 0; Indicates time The observational innovation residual column vector originates from the difference between the actual observed value and the prior predicted observed value; This represents the transpose of the column vector of observed innovation residuals; This represents the inverse of the innovation covariance matrix generated during the state observer update process.

[0048] like Figure 4 As shown, in one specific embodiment, S3 includes: S31. Obtain the unconstrained derivative variable based on the time mean and state mean of the posterior state estimation vector for the continuous multiple periods; S32. Based on the sign parameter of the unconstrained derivative variable, the preset maximum rate of change constant, and the discount factor, establish the physical limit benchmark value; S33. Based on the numerical comparison and truncation operation between the absolute value of the unconstrained derivative variable and the physical limit benchmark value, obtain the controlled trend gradient variable.

[0049] In this embodiment, it should be noted that in steps S31 to S33, to limit the extrapolation distortion caused by mathematical fitting when processing high-frequency sensing noise, the system uses the physical mechanical limit as a boundary constraint condition for the mathematical calculation. The system acquires the posterior state estimation vector for five consecutive sampling periods and calculates the original unconstrained derivative variable using the least squares fitting principle. The calculation formula is as follows:

[0050] The characters in the formula are defined as follows: This represents the unconstrained derivative variable calculated based on continuous sampling periods; The index number representing the historical sampling period, taking the value of an integer of 1, 2, 3, 4, or 5; Indicates the first Hardware timestamp for each sampling period; This represents the time average of the hardware timestamps over 5 sampling periods; Indicates the first The posterior state estimation vector components corresponding to each sampling period; This represents the mean state value of the posterior state estimation vector components over 5 sampling periods; Indicates subscript Perform a cumulative summation operation on the sequence from 1 to 5.

[0051] After obtaining the unconstrained derivative variable, the system extracts the physical characteristic parameters of the target equipment, sets the maximum rate of change constant, and introduces a safety margin. Through numerical comparison and sign recovery, the controlled trend gradient variable is obtained, and its calculation formula is as follows:

[0052] The characters in the formula are defined as follows: This represents the controlled trend gradient variable obtained after truncation logic processing; This represents a sign function, whose internal independent variable takes the value 1 when it is greater than 0, -1 when it is less than 0, and 0 when it is equal to 0. This function represents the cutoff function that takes the smaller value of the two input variables within the parentheses. This represents the absolute value of the unconstrained derivative variable; This represents the maximum rate of change constant set by the electromechanical characteristics of the equipment at the factory; This represents the safety margin discount factor constant.

[0053] Specifically, the preset maximum rate of change constant and discount factor This data was obtained through statistical analysis of a limited number of historical extreme operating condition tests conducted on the target equipment. Before system deployment, multiple sets of historical continuous motion state data of the equipment under extreme operating conditions such as full-load start-up and emergency stop were collected. The true rate of change for each adjacent sampling period was calculated, and these rate of change data were plotted as probability density distribution curves. The statistical upper limit extreme value covering 99.9% of physically reasonable jumps was selected as the maximum rate of change constant. To prevent high-frequency oscillations at the control boundary, a one-sided safety margin is added as a discount factor to this extreme value. For example, by collecting 100,000 angular acceleration data points from extreme acceleration and deceleration tests of a certain type of industrial robot, it was found that 99.9% of the peak angular acceleration values ​​did not exceed 0.105 rad / s². Considering mechanical rigidity and buffering, [the following was selected]. The value is 0.1 rad / s², and a discount factor is set. The value is 0.95, thus ensuring physical authenticity while leaving a 5% safety control redundancy.

[0054] This operational logic ensures that the extracted gradient variables are always confined within the physical envelope of the device, eliminating non-physical transient jumps caused by electromagnetic interference.

[0055] like Figure 5 As shown, in one specific embodiment, S4 includes: S41. Obtain the step state value based on the comparison result between the network latency index and the preset latency threshold; S42. Perform a combination operation based on the basic redundancy ratio constant, sensitivity coefficient, the difference parameter of the confidence factor, and the step state value to obtain the error correction redundancy ratio, and increase the error correction redundancy ratio when the confidence factor decreases and network latency exists.

[0056] In this embodiment, it should be noted that in S41 to S42, in order to limit the shortcomings of traditional industrial communication in adapting to dynamic electromagnetic interference in the workshop due to the use of fixed forward error correction configuration, the system strongly couples the "state reliability" of the application layer with the "channel jitter" of the physical link across layers, and uses it as the boundary constraint condition for dynamically calculating communication redundancy.

[0057] In S41, the "network latency metric" characterizes the stability of the current underlying communication link, and the "step state value" is used to quantify whether the network has experienced substantial jitter exceeding limits. The system detects the latency status in real time through the network control plane and obtains the step state value, whose logical expression, including the step function, is as follows:

[0058] The characters in the formula are defined as follows: This represents a step function, where the function value is 1 when the internal independent variable is greater than 0, and the function value is 0 when it is less than or equal to 0. The network latency metric, which is measured in real time, is specifically the standard deviation of the latency of the communication link in this embodiment. This represents the preset delay threshold, which characterizes the maximum safe jitter constant that the network can tolerate.

[0059] Specifically, preset delay threshold This calibration is based on a limited number of historical network communication test data combined with the fault tolerance requirements of the digital twin system. During the system initialization phase, fixed-size probe messages are continuously sent to the industrial Ethernet to record network latency under various workshop interference conditions. The latency standard deviation sequence within different time windows is calculated, and the maximum latency standard deviation corresponding to no significant lag in the system twin screen is extracted as the benchmark. For example, 5000 sets of latency jitter data were recorded during on-site communication tests in the workshop. When the network latency standard deviation reached 12ms, the twin model rendering began to show slight jumps. Combined with the system's allowed 20% tolerance, the preset latency threshold was set. The offline calibration time is 10ms, which serves as a rigid decision boundary for triggering adaptive error correction control.

[0060] In S42, the "error correction redundancy ratio" refers to the proportion of check bytes used to recover damaged data packets to the effective data in forward error correction coding. The system obtains the indicators from the preceding steps and performs cross-layer combination operations to dynamically obtain the current error correction redundancy ratio. Its calculation formula is as follows:

[0061] The characters in the formula are defined as follows: This represents the error correction redundancy ratio dynamically calculated within the current sampling period; This represents the basic redundancy ratio constant set by the network under ideal congestion-free conditions; A sensitivity coefficient constant representing the magnitude of the control redundancy expansion; This represents the confidence factor output by the state extraction module; The difference parameter representing the confidence factor is used to characterize the severity or unpredictability of the current physical entity's state change. This represents the step state value obtained in step S41.

[0062] Specifically, the basic redundancy ratio constant and sensitivity coefficient This result is derived from fitting data from a limited number of historical packet loss recovery tests conducted under weak network conditions. By artificially injecting random packet loss with gradients ranging from 0% to 20% into the test network, the single-pass decoding success rate of data packets under different redundancy configurations was recorded. The nonlinear curve of redundancy versus decoding success rate was fitted using the least squares method, and the lowest redundancy percentage required to maintain a 99% decoding rate under ideal channel conditions was taken as the basic redundancy ratio constant. The slope of the curve during the worsening packet loss rate phase is taken as the sensitivity coefficient. For example, processing data from 10,000 historical channel fading tests revealed that only 5% redundancy is needed to achieve lossless transmission in the absence of interference. Therefore, the following setting was adopted: The redundancy is 0.05. However, when network jitter and drastic changes occur, the redundancy needs to increase rapidly at a ratio of 1:2 as the confidence level decreases in order to compensate for channel loss. Therefore, the sensitivity coefficient is obtained through linear regression. It is 0.2.

[0063] This operational logic ensures that the extracted error correction redundancy ratio dynamically adapts to the physical envelope of the device and the network channel state: when the confidence factor decreases (the device state changes abruptly) and there is network latency (the step state value is 1), the system will automatically increase the error correction redundancy ratio.

[0064] In S4, the main solution addresses the technical problem that traditional communication "best-effort" or fixed-parameter retransmission strategies are prone to causing "latency blind spots" or even "freezing" of digital twin images due to network packet loss when dealing with sudden interference in industrial sites. It achieves the technical effect of targeted enhancement of the single-pass decoding success rate of critical sudden states under poor channel conditions in bandwidth-constrained industrial networks, eliminating the control delay caused by blind retransmissions.

[0065] Specifically, the application scenario is the remote digital twin takeover of unmanned mining trucks: The confidence factor is calculated when the mining truck is traveling smoothly in a straight line and the 5G network is good. Measured network latency metrics ms, preset latency threshold ms. Due to The obtained step state value Substitute the values ​​into the formula to obtain the error correction redundancy ratio. (For example, set to 0.05). The system allocates only 5% of the parity redundancy bandwidth, saving communication resources.

[0066] When the mining truck suddenly performs emergency obstacle avoidance (a drastic change in physical state causes a decrease in model matching accuracy, calculated as follows) And it just so happens to have driven into an area deep in the mine where multipath reflections are severe (network degradation, calculated) ms). At this time, the step state value Assuming sensitivity coefficient Substitute the values ​​into the formula to obtain the error correction redundancy ratio. The edge gateway increases the redundancy of channel coding to 25%, ensuring that even if some data packets of this synchronization message containing critical emergency obstacle avoidance poses are lost due to strong interference during transmission, the cloud receiving layer can use the high redundancy bits to restore the original state in one go, eliminating the collision risk caused by blind retransmission.

[0067] like Figure 6 As shown, in one specific embodiment, S5 includes: S51. Obtain the receiving timestamp and decoding time constant according to the received control command; S52. Obtain the elastic waiting time variable based on the network gating parameters and scheduling cycle constant; S53. Obtain the physical action trigger time based on the combined calculation of the received timestamp, the decoding time constant, and the elastic waiting time variable.

[0068] Specifically, the preset Mahalanobis distance sensitivity threshold, scheduling cycle constant, and decoding time constant are obtained through statistical analysis based on a limited number of historical device operating status and network processing test data. During the debugging phase before system commissioning, the Mahalanobis distance of the massive multidimensional state feature vectors collected during normal and stable operation of the equipment is calculated. The upper limit extreme value containing 99% of normal fluctuation samples is calibrated as the Mahalanobis distance sensitivity threshold. At the same time, by cyclically sending standard test messages to the edge gateway processor, the processing clock cycles required for the message to be received, completely unpacked, and verified are counted, and the 95th percentile of this time is taken as the decoding time. The constant is configured based on the basic time slice length of the underlying TSN switch gating list. For example, by analyzing 5000 sets of normal operating status data, their Mahalanobis distances are all distributed below 0.5. Combined with a certain fault tolerance, the Mahalanobis distance sensitivity threshold is set to 0.6. At the same time, the test records show that the processor unpacking time is concentrated between 180 microseconds and 210 microseconds. The decoding time constant is calibrated to 200 microseconds. Combined with the 1000 microsecond cyclic slice of the switch, the scheduling cycle constant is set to 1000 microseconds, thereby ensuring that the time parameters of the scheduling algorithm are completely consistent with the actual hardware capabilities.

[0069] In this embodiment, it should be noted that in S51 to S53, in order to eliminate the random dwell time delay caused by control commands in the operating system buffer queue, the system executes the network-computer joint scheduling logic to perform cross-layer hard alignment between network transmission time slots and terminal computing time. The system obtains the reception timestamp and decoding time constant based on the received control commands, and calculates the elastic waiting time variable by combining the network gating parameters and scheduling cycle constant. The calculation formula is as follows:

[0070] The characters in the formula are defined as follows: This represents the calculated elastic waiting time variable; The relative time offset constant (i.e., the network gating parameter) represents the next time a high-priority transmission window will be opened in the gating list that supports time-sensitive network protocols. The received timestamp indicates the downlink control command obtained by the physical layer network interface card; This represents the modulo operator; This represents the preset scheduling period constant; The decoding time constant represents the time required for the edge processor to perform message decoding.

[0071] After obtaining the elastic waiting time variable, the system performs a combination operation to determine the absolute physical time point at which the action execution pulse signal is finally sent to the underlying driver, and obtains the physical action trigger time. The calculation formula is as follows:

[0072] The characters in the formula are defined as follows: This indicates the calculated trigger time of the physical action; Indicates the received timestamp; This represents the decoding time constant; This represents the elastic waiting time variable.

[0073] This operational logic ensures that edge control commands are rigidly activated at predetermined physical action trigger times, eliminating the impact of software-layer scheduling randomness on hardware execution timing, eliminating action jitter, and achieving hard real-time closed-loop control.

[0074] like Figure 7 As shown, in one specific embodiment, S6 includes: S61. Construct an exponential expression based on the regression rate constant, system time variable, and data arrival time constant to obtain the dynamically changing smooth mixing coefficient; S62. The predicted state vector and the true state vector are weighted and summed according to the smoothing mixing coefficient to obtain the smoothed state vector.

[0075] In this embodiment, it should be noted that in S61 to S62, to address the issue of screen flickering or coordinate shifts caused by the virtual model directly updating the real state after network micro-disconnection recovery, the system introduces a time-decay-based weighted algorithm. The system first constructs an exponential expression based on the regression rate constant, system time variable, and data arrival time constant to obtain a dynamically changing smoothing mixing coefficient. The calculation formula is as follows:

[0076] The characters in the formula are defined as follows: The smoothing mixing coefficient is obtained by dynamically changing over time; Represents the base of the natural logarithm; This represents the preset regression rate constant; The system time variable representing the current operation of the 3D engine; The data arrival time constant represents the actual data packets arriving at the system.

[0077] Specifically, the regression rate constant and data arrival time constant This is based on a limited number of historical tests of human visual persistence characteristics and system rendering refresh rate processing; during the system offline debugging phase, the average absolute time difference between the data packet sent from the edge gateway and the received data by the twin rendering module is recorded as the data arrival time constant. Simultaneously simulating multiple data delay pullback scenarios, for different... The smoothness of the image under a given value is scored, and the optimal slope that allows the S-shaped smooth curve to complete the transition within the human visual blind spot is selected as the regression rate constant. For example, by analyzing historical network transmission logs, the average arrival delay of data packets under normal fluctuations was found to be 10ms, therefore, the following setting was made: With a 10ms response time and a 60Hz rendering refresh rate, to ensure a seamless transition from extrapolated prediction to the actual state within the visual blind zone, a regression rate constant controlling the smooth transition curve was determined through mathematical fitting. It is 0.5.

[0078] After obtaining the smoothing mixing coefficient, the system uses this coefficient to seamlessly fuse the predicted state vector generated by trajectory extrapolation with the real state vector parsed after delayed arrival to obtain a smooth state vector. The calculation formula is as follows:

[0079] The characters in the formula are defined as follows: This represents the calculated smooth state vector; This represents the smoothing mixing coefficient; This represents the predicted state vector independently generated by the trajectory extrapolation operator; This represents the true state vector parsed after the delay.

[0080] This operational logic assigns progressively increasing weights to the real state, enabling the virtual model to smoothly revert to its real trajectory within a very short time window in accordance with the inertial laws of objects. This eliminates visual abrupt changes and ensures the continuity of the virtual-real mapping in terms of time and spatial location.

[0081] The present invention also provides a real-time device status synchronization system based on digital twins, the system comprising: The data acquisition module is used to acquire a global time base according to the cross-domain time base hard alignment mechanism, and to sample the physical signal according to the global time base to acquire a multi-dimensional state digital sequence. The state extraction module is used to extract state features based on the multidimensional state number sequence, and perform iterative calculations on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor. The gradient calculation module is used to obtain the controlled trend gradient variable based on the physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods. The message generation module is used to assemble and generate state metadata based on the time base, spatial parameters, the controlled trend gradient variable and the confidence factor, and to obtain the synchronization message based on the state metadata encoding. The transmission control module is used to determine the error correction redundancy ratio based on the measured network latency index and the confidence factor, and to perform channel coding transmission on the synchronization message according to the error correction redundancy ratio. The twin rendering module is used to parse the transmitted synchronization message to obtain the real state vector, and when there is a transmission delay, to obtain the predicted state vector according to the trajectory extrapolation operator; and to fuse the predicted state vector and the delayed real state vector according to the smoothing blending function to obtain the smooth state vector and drive the rendering of the virtual model.

[0082] In one specific implementation, the state extraction module is specifically used to obtain the sparse Jacobian matrix by removing zero elements, obtain the Kalman gain matrix by transposing and inverting the sparse Jacobian matrix, obtain the posterior state estimation vector by fusing the Kalman gain matrix and the observation vector, and obtain the confidence factor by fusing the innovation covariance inverse matrix and the observation innovation residual vector. The gradient calculation module is specifically used to obtain the unconstrained derivative variable based on the time mean and state mean of the posterior state estimation vector over multiple consecutive periods, establish the physical limit benchmark value based on a preset maximum rate of change constant and discount coefficient, and obtain the controlled trend gradient variable by comparing the absolute value of the unconstrained derivative variable with the physical limit benchmark value.

[0083] In some optional embodiments, in order to enable those skilled in the art to fully understand and implement the technical solutions described in this specification, the following provides a full-process data analysis of the implementation principle of the real-time synchronization method and system for device status based on digital twins, combined with a specific application scenario.

[0084] Imagine a real-time synchronization scenario in a smart manufacturing workshop, with the target equipment being a running industrial robot. The data acquisition module performs clock synchronization, acquiring the master clock's transmitted timestamp. Receive timestamp Receive timestamps from the clock Send timestamp Substituting the above values ​​into the offset calculation formula for the cross-domain time base hard alignment mechanism:

[0085] The asymmetric time offset is calculated as follows:

[0086] The hardware performs phase-locked loop compensation based on this 5ns offset to ensure that the multidimensional state digital sequence acquired at a sampling frequency of 2000Hz carries an accurate global time base.

[0087] After acquiring the data sequence, the state extraction module performs feature determination. Current time-time feature vector. Sliding window mean vector Inverse covariance matrix Substituting into the Mahalanobis distance formula The calculation shows that: Once the value is determined to be out of bounds, iterative calculation is triggered. The sparse Jacobian matrix is ​​then read. Prior error covariance matrix Observation noise covariance matrix .

[0088] Then substitute into the formula Obtain the Kalman gain matrix Given the observation vector and prior state estimation vector Substitute into the formula Obtain the posterior state estimation vector Extracting innovative residuals from observations Calculations yielded Substitute into the confidence factor formula. Obtain confidence factor .

[0089] The gradient calculation module extracts the posterior state values ​​from five consecutive sampling periods to calculate the trend and time mean. State mean Substitute into the formula:

[0090] Calculate the unconstrained derivative variables Let the maximum rate of change be constant. Discount coefficient Substitute into the cutoff formula By comparing 0.084 with the physical limit baseline value of 0.095, the controlled trend gradient variable is obtained. The message generation module then generates a synchronization message based on this.

[0091] The transmission control module measures the current network latency index. Preset delay threshold Step state value Let the basic redundancy ratio constant be... Sensitivity coefficient Substituting into the error correction redundancy formula Obtain the error correction redundancy ratio And execute the transmission. If a downlink command exists, obtain the receive timestamp. Decoding time constant Scheduling cycle constant Network gating parameters Substitute into the formula Obtain the elastic waiting time variable Substitute into the trigger formula Get the physical action trigger time .

[0092] When transmission is delayed, the twin rendering module obtains the predicted state vector based on trend extrapolation. The true state vector is parsed after the delay. Let the system time variable be... Data arrival time constant Regression rate constant Substitute into the mixing coefficient formula Obtain the smooth mixing coefficient Substitute into the smoothing blending function Obtain the smooth state vector The twin rendering module ultimately uses this value to drive rendering, maintaining spatiotemporal consistency between the virtual and real states.

[0093] 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. A method for real-time synchronization and communication of industrial IoT device status based on digital twins, characterized in that, include: A global time reference is obtained based on a cross-domain time base hard alignment mechanism, and a multi-dimensional state digital sequence is obtained by sampling the physical signal based on the global time reference. State features are extracted from the multidimensional state digital sequence, and iterative calculations are performed on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor. Based on the physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods, the controlled trend gradient variable is obtained; State metadata is generated by assembling the time base, spatial parameters, the controlled trend gradient variable, and the confidence factor, and a synchronization message is obtained by encoding the state metadata. The error correction redundancy ratio is determined based on the measured network latency index and the confidence factor, and channel coding transmission is performed on the synchronization message according to the error correction redundancy ratio. The true state vector is obtained by parsing the transmitted synchronization message. When there is a transmission delay, the predicted state vector is obtained by the trajectory extrapolation operator. The predicted state vector and the delayed true state vector are fused by the smoothing mixing function to obtain a smooth state vector and drive the rendering of the virtual model.

2. The method for real-time synchronization and communication of industrial IoT device status based on digital twins according to claim 1, characterized in that, The step of obtaining the global time base based on the cross-domain time base hard alignment mechanism includes: Based on the clock signal of the local clock module and the preset phase-locked loop processing logic, obtain the aligned trigger input signal; The sampling operation is driven by the trigger input signal, and a hardware timestamp is pushed into each sampled data node to obtain the global time reference.

3. The method for real-time synchronization and communication of industrial IoT device status based on digital twins according to claim 1, characterized in that, The step of performing iterative calculations on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor includes: The sparse Jacobian matrix is ​​obtained by removing zero elements, and the Kalman gain matrix is ​​obtained by transposing and inverting the sparse Jacobian matrix. The fusion operation is performed based on the Kalman gain matrix and the observation vector to obtain the posterior state estimation vector; The confidence factor is obtained based on the innovation covariance inverse matrix and the observed innovation residual vector.

4. The method for real-time synchronization and communication of industrial IoT device status based on digital twin as described in claim 1, characterized in that, The process of obtaining the controlled trend gradient variable based on the physical boundary truncation logic and the posterior state estimation vector over multiple consecutive periods includes: Based on the time mean and state mean of the posterior state estimation vector for the continuous multiple periods, obtain the unconstrained derivative variable; The physical limit benchmark value is established based on the sign parameter of the unconstrained derivative variable, the preset maximum rate of change constant, and the discount factor; The controlled trend gradient variable is obtained by comparing the absolute value of the unconstrained derivative variable with the physical limit benchmark value and performing a truncation operation.

5. The method for real-time synchronization and communication of industrial IoT device status based on digital twins according to claim 1, characterized in that, The step of determining the error correction redundancy ratio based on the measured network latency index and the confidence factor includes: Based on the comparison result between the network latency index and the preset latency threshold, the step state value is obtained; The error correction redundancy ratio is obtained by performing a combination operation based on the basic redundancy ratio constant, the sensitivity coefficient, the difference parameter of the confidence factor, and the step state value. The error correction redundancy ratio is increased when the confidence factor decreases and network latency exists.

6. The method for real-time synchronization and communication of industrial IoT device status based on digital twin as described in claim 1, characterized in that, After performing channel coding transmission on the synchronization message according to the error correction redundancy ratio, the method further includes: Obtain the receiving timestamp and decoding time constant based on the received control commands; The elastic waiting time variable is obtained based on the network gating parameters and the scheduling cycle constant; The physical action trigger time is obtained by combining the received timestamp, the decoding time constant, and the elastic waiting time variable.

7. The method for real-time synchronization and communication of industrial IoT device status based on digital twin as described in claim 1, characterized in that, The step of fusing the predicted state vector and the delayed-arriving true state vector according to a smoothing mixing function to obtain a smoothed state vector includes: An exponential expression is constructed based on the regression rate constant, the system time variable, and the data arrival time constant to obtain the dynamically changing smooth mixing coefficients; The predicted state vector and the true state vector are weighted and summed according to the smoothing mixing coefficient to obtain the smoothed state vector.

8. A real-time synchronization communication system for the status of industrial IoT devices based on digital twins, characterized in that, include: The data acquisition module is used to acquire a global time base according to the cross-domain time base hard alignment mechanism, and to sample the physical signal according to the global time base to acquire a multi-dimensional state digital sequence. The state extraction module is used to extract state features based on the multidimensional state number sequence, and perform iterative calculations on the state features based on the sparse Jacobian matrix to obtain the posterior state estimation vector and confidence factor. The gradient calculation module is used to obtain the controlled trend gradient variable based on the physical boundary truncation logic and the posterior state estimation vector for multiple consecutive periods. The message generation module is used to assemble and generate state metadata based on the time base, spatial parameters, the controlled trend gradient variable and the confidence factor, and to obtain the synchronization message based on the state metadata encoding. The transmission control module is used to determine the error correction redundancy ratio based on the measured network latency index and the confidence factor, and to perform channel coding transmission on the synchronization message according to the error correction redundancy ratio. The twin rendering module is used to parse the transmitted synchronization message to obtain the real state vector, and when there is a transmission delay, to obtain the predicted state vector according to the trajectory extrapolation operator; and to fuse the predicted state vector and the delayed real state vector according to the smoothing blending function to obtain the smooth state vector and drive the rendering of the virtual model.

9. The system according to claim 8, characterized in that, The state extraction module is specifically used to obtain the sparse Jacobian matrix by removing zero elements, calculate the Kalman gain matrix based on the sparse Jacobian matrix, then combine the observation vector fusion to obtain the posterior state estimation vector, and obtain the confidence factor based on the innovation covariance inverse matrix.

10. The system according to claim 8, characterized in that, The gradient calculation module is specifically used to obtain unconstrained derivative variables based on the posterior state estimation vectors of the continuous multiple periods, establish physical limit benchmark values ​​based on the maximum rate of change constant, and then perform a truncation operation on the absolute value of the unconstrained derivative variables to obtain the controlled trend gradient variables.

Images