Digital twinning control platform

Abstract

The invention discloses a digital twin control platform. The platform comprises a multi-protocol adaptive edge intelligent gateway; the dynamic twin modeling layer comprises an online transfer learning engine and a federal learning coordinator; the intelligent decision control layer is integrated with an MPC-DRL hybrid algorithm and an elastic resource scheduler; and the security interaction layer supports RLWE homomorphic encryption and block chain traceability. According to the edge intelligent gateway, through the wavelet denoising algorithm and the improved Huffman encoder, the 5G and TSN networks are combined, the multi-modal data acquisition and transmission efficiency is remarkably improved, the problem of processing lag of a traditional platform is solved, and the industrial real-time requirement is met. The dynamic twin modeling layer adopts a physical constraint loss function of an online transfer learning engine and a differential privacy mechanism of a federated learning coordinator to realize real-time updating of the model and distributed cooperative privacy guarantee, and overcomes the defects of static modeling and data leakage.

Description

Technical Field

[0001] The present invention relates to the technical field of the Internet of Things, and particularly to a digital twin control platform for application. Background Art

[0002] With the advancement of Industry 4.0 and intelligent manufacturing, digital twin technology has been widely applied in fields such as industrial control and new energy management. However, there are significant deficiencies in data acquisition and processing in existing digital twin control platforms. Traditional edge devices have a relatively high processing delay (often exceeding 10 ms), a low data compression rate (<70%), and only support a single protocol, making it difficult to be compatible with the requirements of multi-modal and multi-source data acquisition, resulting in poor real-time performance. In addition, static or offline methods are mostly adopted at the modeling level, making it difficult to dynamically adapt to changes in physical entities, and data sharing in distributed collaboration is prone to privacy leakage, lacking an effective privacy protection mechanism. These problems limit the efficient application of digital twin systems in complex industrial scenarios.

[0003] On the other hand, there are also defects in the control decision-making and security of existing platforms. Traditional control algorithms (such as PID or single MPC) have limited accuracy and are difficult to meet the requirements of dynamic systems. Resource scheduling is mostly static allocation and cannot be optimized according to real-time loads, resulting in waste or shortage of resources. At the same time, data transmission and storage lack strong encryption support, are vulnerable to attacks, and the decision-making process is not traceable, making it difficult to meet the requirements of high security and compliance in industrial scenarios. Therefore, there is an urgent need for a new type of digital twin control platform. Summary of the Invention

[0004] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a digital twin control platform for application. The digital twin control platform for application solves the above problems in the prior art by improving data processing efficiency, modeling dynamics, control accuracy, and security, and provides reliable support for industrial intelligence.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: A digital twin control platform for application, comprising: an edge intelligent gateway with multi-protocol adaptation, used for real-time acquisition of multi-modal data of physical entities; a dynamic twin modeling layer, including an online transfer learning engine and a federated learning coordinator; an intelligent decision-making control layer, integrating an MPC-DRL hybrid algorithm and an elastic resource scheduler; a secure interaction layer, supporting RLWE homomorphic encryption and blockchain traceability.

[0006] Preferably, the edge intelligent gateway is built-in with a wavelet denoising algorithm and an improved Huffman encoder. The data compression rate of the edge intelligent gateway is ≥85%, the end-side processing delay is ≤5 ms, and it supports protocol conversion of OPC UA, MQTT, and Modbus.

[0007] Preferably, the loss function of the online transfer learning engine is , where is the prediction error, is the constraint term of the physical constraint equation, and λ = 0.1 - 0.5.

[0008] Preferably, the elastic resource scheduler allocates FPGA acceleration resources according to the priority preemption strategy based on the real-time load of the container cluster. The priority calculation formula is P = 0.6E + 0.4R, where E is the task urgency and R is the resource demand coefficient.

[0009] Preferably, the polynomial degree n of the RLWE homomorphic encryption is 1024, the modulus , and the private key update period T and the data sampling frequency f satisfy .

[0010] Preferably, the federated learning coordinator adopts the differential privacy mechanism with Gaussian noise injection. The noise standard deviation = 0.1 × data sensitivity, and the privacy budget = 0.5, and the failure probability .

[0011] Preferably, the hardware configuration of the edge intelligent gateway includes the NVIDIA Jetson AGX Xavier chip and the Xilinx Zynq UltraScale+ RFSoC, supports 5G communication and the TSN time-sensitive network, and the transmission bandwidth ≥ 1Gbps.

[0012] Preferably, the physical constraint equation includes the thermodynamic equation , the kinematic equation , or the electrical characteristic equation , where Q is the heat, c is the specific heat capacity, m is the mass of the substance, is the temperature change, F is the resultant force, is the mass of the object, a is the acceleration of the object, V is the voltage, I is the current, R is the resistance, and the calculation method of the constraint term is the weighted sum of squared errors.

[0013] Preferably, the elastic resource scheduler integrates the Prometheus monitoring system, collects the CPU / GPU / FPGA utilization rate of the container cluster in real time, sets the preemption threshold to CPU ≥ 80% or GPU ≥ 90%, and uses the Kubernetes custom scheduling plugin to achieve resource allocation.

[0014] Preferably, the blockchain traceability adopts the Hyperledger structure framework, the consensus mechanism is the practical Byzantine fault tolerance algorithm, the block generation time ≤ 2 seconds, the minimum number of nodes is 4, and the data tampering detection rate is 100%.

[0015] The present invention has the following beneficial effects: 1. The edge intelligent gateway of the present invention realizes a data compression rate ≥ 85% and an end-side processing delay ≤ 5 ms through a wavelet denoising algorithm and an improved Huffman encoder. Combining 5G and TSN networks, it significantly improves the multi-modal data acquisition and transmission efficiency, solves the problem of lagging processing of traditional platforms, and meets the industrial real-time requirements.

[0016] 2. The dynamic twin modeling layer of the present invention adopts the physical constraint loss function of the online transfer learning engine and the differential privacy mechanism of the federated learning coordinator to realize real-time model update and distributed collaborative privacy protection, overcoming the defects of static modeling and data leakage.

[0017] 3. The intelligent decision-making control layer of the present invention integrates the MPC-DRL hybrid algorithm to improve the control accuracy of complex systems. The elastic resource scheduler dynamically allocates resources based on the priority formula and Prometheus monitoring, solving the problems of low accuracy and resource waste of traditional algorithms.

[0018] 4. The security interaction layer of the present invention supports encrypted computing through RLWE homomorphic encryption, and the blockchain traceability realizes a block generation time ≤ 2 seconds and a tampering detection rate of 100%, effectively solving the problems of data security and untraceability, and improving the reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a block diagram of the platform system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0021] Refer to Figure 1 , an application in a digital twin control platform, including: an edge intelligent gateway 101 with multi-protocol adaptation, used for real-time acquisition of multi-modal data of physical entities; a dynamic twin modeling layer 200, including an online transfer learning engine 201 and a federated learning coordinator 204; an intelligent decision-making control layer 300, integrating an MPC-DRL hybrid algorithm 301 and an elastic resource scheduler 304; a security interaction layer 400, supporting RLWE homomorphic encryption 401 and blockchain traceability 402.

[0022] In this embodiment, the edge intelligent gateway 101 is used to collect multi-modal data of physical entities in real time, support multi-protocol adaptation, perform edge preprocessing and transmission. The edge intelligent gateway 101 improves data collection efficiency and compatibility, reduces transmission latency, ensures the real-time nature of multi-source data, and enhances the system's adaptability to industrial scenarios; the dynamic twin modeling layer 200 is used to dynamically update the digital twin model through the online transfer learning engine 201, and the federated learning coordinator 204 coordinates distributed learning to construct a real-time virtual mapping of physical entities. The dynamic twin modeling layer 200 realizes high adaptability and distributed collaboration of the model. The online transfer learning engine 201 improves modeling accuracy, and the federated learning coordinator 204 protects data privacy and optimizes the system's dynamic performance; the intelligent decision-making and control layer 300 is used to integrate the MPC-DRL hybrid algorithm 301 to optimize real-time control decisions, and the elastic resource scheduler 304 dynamically allocates computing resources to support system operation. The intelligent decision-making and control layer 300 improves control accuracy and resource utilization rate. The MPC-DRL hybrid algorithm 301 adapts to complex scenarios, and the elastic resource scheduler 304 ensures efficient operation and improves the system's intelligence level; the secure interaction layer 400 is used to protect data privacy through RLWE homomorphic encryption 401, and the blockchain traceability 402 records operation history to ensure secure and traceable data interaction. The secure interaction layer 400 enhances system security. RLWE homomorphic encryption 401 supports encrypted computing, and blockchain traceability 402 realizes tampering detection, improving data credibility and compliance.

[0023] In the present invention, the edge intelligent gateway 101 is built-in with a wavelet denoising algorithm and an improved Huffman encoder. The data compression rate of the edge intelligent gateway 101 is ≥85%, and the end-side processing delay is ≤5 ms, and it supports protocol conversion of OPC UA, MQTT, and Modbus.

[0024] In this embodiment, the edge intelligent gateway 101 is built-in with a wavelet denoising algorithm and an improved Huffman encoder, collects multi-modal data of physical entities in real time and performs denoising and compression, supports protocol conversion of OPC UA, MQTT, and Modbus, and realizes edge data processing and transmission. The edge intelligent gateway 101 improves data quality through the wavelet denoising algorithm, and the improved Huffman encoder realizes a data compression rate of ≥85% and an end-side processing delay of ≤5 ms, ensuring efficient real-time processing. Multi-protocol support enhances compatibility and optimizes data collection and transmission efficiency in industrial scenarios.

[0025] In the present invention, the loss function of the online transfer learning engine 201 is , where is the prediction error, is the constraint term of the thermodynamic equation or the kinematic equation, = 0.1 - 0.5.

[0026] In this embodiment, the online transfer learning engine 201 dynamically updates the digital twin model through a loss function where is the prediction error, is the constraint term of the physical constraint equation, = 0.1 - 0.5 adjusts the physical constraint weight to achieve real-time modeling. The online transfer learning engine 201 combines data-driven and physical laws to improve the model accuracy and adaptability, ensuring compliance with actual physical characteristics, and the flexible adjustment of

[0027] optimizes the modeling balance, enabling the digital twin control platform to quickly respond to physical entity changes.

[0028] In this embodiment, the elastic resource scheduler 304 allocates FPGA acceleration resources based on the real-time load of the container cluster through a priority preemption strategy. The priority calculation formula is P = 0.6E + 0.4R, where E is the task urgency and R is the resource demand coefficient. The elastic resource scheduler 304 dynamically optimizes the computing resource allocation by integrating the task urgency E and the resource demand coefficient R through the priority calculation formula P = 0.6E + 0.4R. The elastic resource scheduler 304 improves the resource utilization efficiency, and the priority formula P = 0.6E + 0.4R ensures that urgent tasks are executed first. The FPGA acceleration supports high-performance computing, enhancing the response ability of the digital twin control platform to dynamic loads and the system stability.

[0029] In the present invention, the polynomial degree n of the RLWE homomorphic encryption 401 is 1024, and the modulus The private key update period T and the data sampling frequency f satisfy .

[0030] In this embodiment, the RLWE homomorphic encryption 401 uses the RLWE algorithm with a polynomial degree n = 1024 and a modulus to perform homomorphic encryption on the data. The private key update period T is set to according to the data sampling frequency f, ensuring encrypted computing and data security. The RLWE homomorphic encryption 401 supports data processing in the encrypted state. n = 1024 and ensure high security strength, dynamically updates the private key to enhance the anti-attack ability, strengthening the data privacy protection and security of the digital twin control platform.

[0031] In the present invention, the federated learning coordinator 204 adopts a differential privacy mechanism with Gaussian noise injection. The noise standard deviation = 0.1 × data sensitivity, and the privacy budget = 0.5, failure probability .

[0032] In this embodiment, the federated learning coordinator 204 uses a differential privacy mechanism with Gaussian noise injection to coordinate distributed learning. The standard deviation of the noise = 0.1 × data sensitivity, and the privacy budget = 0.5, failure probability . To protect data privacy in multi-party collaboration, the federated learning coordinator 204 precisely controls the noise through = 0.1 × sensitivity, = 0.5 and ensures strong privacy protection and a low failure rate, improves the security and reliability of the distributed modeling of the digital twin control platform, and maintains the model accuracy at the same time.

[0033] In the present invention, the hardware configuration of the edge intelligent gateway 101 includes an NVIDIA Jetson AGX Xavier chip and a Xilinx Zynq UltraScale+ RFSoC, supports 5G communication and TSN time-sensitive network, and the transmission bandwidth ≥ 1 Gbps.

[0034] In this embodiment, the hardware configuration of the edge intelligent gateway 101 includes an NVIDIA Jetson AGX Xavier chip and a Xilinx Zynq UltraScale+ RFSoC, supports 5G communication and TSN time-sensitive network, and the transmission bandwidth ≥ 1 Gbps. It is used to efficiently collect and process multi-modal data of physical entities. The edge intelligent gateway 101 improves the computing power through high-performance chips, and the 5G and TSN networks achieve low-latency transmission with a bandwidth ≥ 1 Gbps, enhancing the data real-time performance and reliability of the digital twin control platform and meeting the high-efficiency communication requirements of industrial scenarios.

[0035] In the present invention, the physical constraint equations include thermodynamic equations , kinematic equations , or electrical characteristic equations . Among them, Q is heat, c is specific heat capacity, m is the mass of the substance, is the temperature change, F is the resultant external force, is the mass of the object, a is the acceleration of the object, V is voltage, I is current, R is resistance, and the calculation method of the constraint term is the weighted sum of squared errors.

[0036] In this embodiment, the online transfer learning engine 201 introduces physical constraint equations and incorporates them into the loss function through the constraint term for constraint violation penalty, ensuring that the model conforms to physical laws. The online transfer learning engine 201 uses physical constraint equations to improve the scientific nature of the model. The weighted sum of squared error quantifies the degree of violation, enhancing the accuracy and reliability of the digital twin control platform modeling and adapting to complex physical scenarios.

[0037] In the present invention, the elastic resource scheduler 304 integrates the Prometheus monitoring system to collect the CPU / GPU / FPGA utilization rates of the container cluster in real time. The preemption threshold is set as CPU ≥ 80% or GPU ≥ 90%, and the Kubernetes custom scheduling plugin is used to implement resource allocation.

[0038] In this embodiment, the elastic resource scheduler 304 integrates the Prometheus monitoring system to collect the CPU / GPU / FPGA utilization rates of the container cluster in real time, sets the preemption threshold, and dynamically allocates resources through the Kubernetes custom scheduling plugin. The elastic resource scheduler 304 uses Prometheus to achieve accurate load monitoring, and the preemption threshold ensures efficient resource utilization. The Kubernetes plugin optimizes the allocation flexibility, improving the computing performance and system stability of the digital twin control platform.

[0039] In the present invention, the blockchain traceability 402 adopts the Hyperledger Fabric framework, that is, the super ledger structure framework, and the consensus mechanism is Practical Byzantine Fault Tolerance, that is, the practical Byzantine fault tolerance algorithm. The block generation time ≤ 2 seconds, the minimum number of nodes is 4, and the data tampering detection rate is 100%.

[0040] In this embodiment, the blockchain traceability 402 adopts the Hyperledger Fabric framework and the PBFT consensus mechanism. The block generation time ≤ 2 seconds, the minimum number of nodes is 4, records the operation history and detects data tampering to ensure the credibility of traceability. The blockchain traceability 402 achieves efficient consensus through Hyperledger Fabric and PBFT. The minimum of 4 nodes guarantees distributed reliability, and the 100% tampering detection rate improves the data credibility and security of the digital twin control platform.

[0041] As mentioned above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered within the protection scope of the present invention.

Claims

1. An application for a digital twin control platform, characterized in that Including: A multi - protocol - adapted edge intelligent gateway (101) for real - time collection of multi - modal data of physical entities; A dynamic twin modeling layer (200), including an online transfer learning engine (201) and a federated learning coordinator (204). The digital twin model is dynamically updated through the online transfer learning engine (201), and the federated learning coordinator (204) coordinates distributed learning to construct a real - time virtual mapping of physical entities; An intelligent decision - making and control layer (300), integrating an MPC - DRL hybrid algorithm (301) and an elastic resource scheduler (304), for optimizing real - time control decisions and dynamically allocating computing resources; A security interaction layer (400), supporting RLWE homomorphic encryption (401) and blockchain traceability (402). Data privacy is protected through the RLWE homomorphic encryption (401), and the blockchain traceability (402) records the operation history for data interaction security and traceability.

2. The one applied to the digital twin control platform according to claim 1, characterized in that The edge intelligent gateway (101) is built - in with a wavelet denoising algorithm and an improved Huffman encoder. The data compression rate of the edge intelligent gateway (101) is ≥85%, the end - side processing delay is ≤5ms, and it supports protocol conversion of OPC UA, MQTT, and Modbus.

3. A digital twin control platform according to claim 1, characterized in that The loss function of the online transfer learning engine (201) is , where is the prediction error, is the constraint term of the physical constraint equation, = 0.1 - 0.

5.

4. A digital twin control platform according to claim 1, wherein The elastic resource scheduler (304) allocates FPGA acceleration resources according to the real - time load of the container cluster and a priority pre - emption strategy. The priority calculation formula is P = 0.6E+0.4R, where P is the priority, E is the task urgency, and R is the resource requirement coefficient.

5. A digital twin control platform according to claim 1, characterized in that The polynomial degree n of the RLWE homomorphic encryption (401) is 1024, and the modulus , the private key update period T and the data sampling frequency f satisfy .

6. The one applied to the digital twin control platform according to claim 1, characterized in that, The federated learning coordinator (204) adopts a differential privacy mechanism with Gaussian noise injection, and the standard deviation of the noise = 0.1 × data sensitivity, and the privacy budget = 0.5, and the failure probability .

7. A digital twin control platform according to claim 1, characterized in that, The hardware configuration of the edge intelligent gateway (101) includes an NVIDIA Jetson AGX Xavier chip and an Xilinx Zynq UltraScale+RFSoC, supports 5G communication and TSN time - sensitive network, and the transmission bandwidth is ≥1Gbps.

8. The one applied to the digital twin control platform according to claim 3, characterized in that, The physical constraint equations include thermodynamic equations , kinematic equations , or electrical characteristic equations , where Q is heat, c is specific heat capacity, m is the mass of a substance, is the temperature change, F is the resultant external force, is the mass of an object, a is the acceleration of the object, V is voltage, I is current, R is resistance, and the calculation method of the constraint term is weighted sum of squared errors.

9. The one applied to the digital twin control platform according to claim 4, characterized in that, The elastic resource scheduler (304) integrates a Prometheus monitoring system, real - time collects the CPU / GPU / FPGA utilization rate of the container cluster, the pre - emption threshold is set as CPU≥80% or GPU≥90%, and resource allocation is implemented using a Kubernetes custom scheduling plugin.

10. A digital twin control platform according to claim 1, characterized in that The blockchain traceability (402) adopts a Hyperledger structure framework, the consensus mechanism is a practical Byzantine fault - tolerance algorithm, the block generation time is ≤2 seconds, the minimum number of nodes is 4, and the data tampering detection rate is 100%.

Images