Magnetically driven rotor twinning control system

Abstract

The invention discloses a magnetic drive rotor twinborn control system. The system comprises a real-time control layer, a monitoring interaction layer and a cloud optimization layer which are connected in sequence, wherein the real-time control layer directly interacts with a magnetic drive rotor physical entity and is used for data acquisition, twin model state iteration and control signal output; the monitoring interaction layer is used for system monitoring, man-machine interaction, data scheduling and instruction management; and the cloud optimization layer is used for executing deep analysis and model optimization. According to the invention, through deep cooperation of the three-layer architecture, the defects of the existing general digital twin framework in coping with the requirements of high real-time control and deep model optimization of the magnetic drive rotor are effectively solved, and full life cycle management from microsecond control to long-term performance optimization is realized.

Description

Technical Field

[0001] This invention relates to a digital twin system, and more particularly to a magnetically driven rotor twin control system. Background Technology

[0002] Magnetic drive rotor systems (such as magnetic drive rotors and magnetic levitation flywheel energy storage devices) are advanced electromechanical energy conversion and drive devices. With their significant advantages such as contactless support, high-speed operation, and low losses, they show broad application prospects in key fields such as high-end manufacturing, aerospace, and energy storage. However, these systems use magnetic force as the driving and levitation medium for the rotor, making them typical strongly coupled nonlinear dynamic systems. Their operation involves deep interaction between electromagnetic fields, rotor dynamics, and control systems, and is often accompanied by time-varying parameters and external disturbances. This poses severe challenges to high-precision modeling, real-time stability control, and full life-cycle state monitoring of the system. Digital twin technology, by constructing a virtual mapping of physical entities, provides a new technical path to solve the perception and control problems of these complex systems.

[0003] Currently, while general-purpose industrial digital twin frameworks exist in the market, most focus on non-real-time macroscopic data visualization and offline fault diagnosis. For magnetically driven rotor systems, which have extremely high requirements for dynamic response speed, control determinism, and robustness, existing general-purpose frameworks have significant limitations: firstly, existing architectures often employ loosely coupled hierarchical structures, resulting in significant data transmission and processing delays, making it difficult to meet the microsecond-level real-time control and state synchronization requirements of magnetically driven rotors; secondly, the lack of online model correction mechanisms for rotor dynamics characteristics causes the virtual model to easily deviate from the actual state after long-term operation, failing to form an effective "perception-modeling-control" closed loop. Therefore, there is an urgent need to develop a digital twin control system specifically for magnetically driven rotors with deep hardware and software collaboration capabilities to overcome the shortcomings of existing technologies in real-time performance, control integration, and model self-adaptation. Summary of the Invention

[0004] Purpose of the invention: The technical problem to be solved by the present invention is to provide a magnetically driven rotor twin control system to address the shortcomings of the prior art.

[0005] To address the aforementioned technical problems, this invention discloses a magnetically driven rotor twin control system, comprising:

[0006] The system consists of a real-time control layer, a monitoring and interaction layer, and a cloud optimization layer, connected sequentially.

[0007] The real-time control layer interacts directly with the physical entity of the magnetically driven rotor and is used for data acquisition, twin model state iteration, and control signal output.

[0008] The monitoring and interaction layer is used for system monitoring, human-computer interaction, data scheduling, and command management;

[0009] The cloud optimization layer is used to perform in-depth analysis and model optimization.

[0010] Furthermore, the real-time control layer includes:

[0011] The high-frequency data acquisition and data preprocessing module is connected to the sensing system of the magnetically driven rotor to perform synchronous data acquisition and hardware filtering to obtain preprocessed data.

[0012] The twin model state iteration module iteratively updates the twin model based on preprocessed data;

[0013] The fast control decision and output module executes a control algorithm based on the predicted state of the twin model and outputs control signals.

[0014] Furthermore, the real-time control layer is implemented using a hardware architecture, including:

[0015] Field-programmable gate arrays (FPGAs), microprocessors, and AD / DA conversion modules; among which,

[0016] Data is synchronously acquired within a Field Programmable Gate Array (FPGA), and multiple parallel FIR digital filters are instantiated for real-time signal denoising.

[0017] Furthermore, the twin model state iteration module models the magnetically driven rotor system in a state-space form and predicts the system state at the next moment. The state-space model is represented as follows:

[0018]

[0019] in, Let be the system state vector. For the system input vector, This is the system output vector; , , , These are the state matrix, input matrix, output matrix, and feedforward matrix, respectively.

[0020] The state-space model is iteratively computed on a field-programmable gate array (FPGA) hardware platform.

[0021] Furthermore, the rapid control decision and output module controls the system based on the system state predicted by the twin model state iteration module at the next moment; wherein,

[0022] The control method adopts a distributed proportional-integral-derivative control strategy, which decouples the multi-degree-of-freedom magnetic bearing control system into independent control channels. An independent PID controller is designed for each channel to control key parameters in parallel and independently. The control quantity is output to the power amplifier through the digital-to-analog converter module to drive the magnetic drive rotor actuator, forming a closed-loop control circuit.

[0023] Furthermore, the monitoring interaction layer includes:

[0024] The system monitoring and visualization module is used to dynamically display the operating status of the magnetically driven rotor through a graphical human-machine interface;

[0025] The instruction issuance and configuration management module is used to receive and securely verify operation instructions or model update parameters;

[0026] The data aggregation and short-term storage module is used to cache high-frequency data and perform preliminary processing and feature extraction.

[0027] Furthermore, the cloud optimization layer includes:

[0028] The timing data storage module is used to store the entire lifecycle operation data of the magnetic drive rotor;

[0029] The twin model parameter update module is used to invert and optimize the twin model parameters based on machine learning or data assimilation techniques.

[0030] The integrated intelligent application service module is used to provide predictive maintenance, energy efficiency optimization, and anomaly diagnosis services.

[0031] Furthermore, the online parameter update module for the twin model performs hierarchical progressive parameter inversion and calibration for the time-varying characteristics of the rotor-bearing system in the magnetically driven rotor, including:

[0032] The first layer, the fast response layer, uses online estimation algorithms based on high-frequency operating data from a sliding time window to fine-tune rapidly changing parameters in the model in real time and track short-term fluctuations in the equipment's state.

[0033] The second layer, the deep optimization layer, is activated periodically. Based on labeled historical data over a longer period, it uses a deep learning model that incorporates physical constraints or an evolutionary algorithm that adapts to covariance to perform global deep optimization of the slowly varying parameters in the model.

[0034] Furthermore, the twin model parameter update module's workflow includes:

[0035] Step 1: Collect historical and real-time operational data;

[0036] Step 2, Data preprocessing and feature extraction;

[0037] Step 3: Construct the objective function for parameter optimization;

[0038] Step 4: Solve iteratively using an optimization algorithm;

[0039] Step 5: Output the optimized model parameters;

[0040] Step 6: Generate a parameter update package and send it to the monitoring interaction layer.

[0041] Beneficial effects:

[0042] 1. This invention proposes a layered heterogeneous hardware and software collaborative architecture. By constructing a three-tiered system of a real-time control layer, a monitoring and interaction layer, and a cloud optimization layer, and by rationally allocating hardware resources according to the computing needs of each layer, it achieves full-process coverage from microsecond-level closed-loop control to long-term model self-evolution.

[0043] 2. This invention utilizes a real-time control layer accelerated by FPGA hardware to ensure microsecond-level high-frequency closed-loop control and precise synchronization between the virtual and real worlds; it relies on a cloud-based optimization layer to achieve continuous optimization of data-driven model parameters, ensuring the high fidelity of the digital twin throughout its entire lifecycle; and its layered heterogeneous architecture optimizes the allocation of computing resources, taking into account the system's real-time performance, deployment cost, and scalability. Attached Figure Description

[0044] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0045] Figure 1 This is a schematic diagram of the overall architecture of the magnetically driven rotor digital twin system of the present invention.

[0046] Figure 2 This is a schematic diagram of the core functional logic architecture of the real-time control layer of the present invention.

[0047] Figure 3 This is a flowchart of the digital twin parameter optimization and calibration method of the present invention. Detailed Implementation

[0048] This invention aims to provide a magnetically driven rotor twin control system to address the shortcomings of existing general-purpose digital twin frameworks, such as insufficient real-time performance and difficulty in meeting the high-precision control requirements of magnetically driven rotors due to their single architectural hierarchy and non-dedicated hardware platforms. Specifically, this invention designs a heterogeneous hierarchical computing architecture with deep hardware and software collaboration and modular deployment. This architecture constructs a dedicated digital twin system capable of achieving highly deterministic, low-latency data interaction and closed-loop control, thereby supporting precise synchronization between the physical entity and the virtual model. This provides a high-performance platform foundation for the state monitoring, optimized control, and predictive maintenance of magnetically driven rotors, improving system reliability and intelligence.

[0049] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0050] The core of this invention lies in constructing a layered heterogeneous hardware and software collaborative architecture that balances the dual requirements of high real-time control and complex intelligent analysis for magnetically driven rotor digital twin systems. Based on the real-time performance, complexity, and functional requirements of the computational tasks, the system architecture is divided into three layers: a real-time control layer, a monitoring and interaction layer, and a cloud-based optimization layer. These layers are interconnected through high-speed communication interfaces, enabling a full-stack digital twin system application from real-time control to long-term optimization.

[0051] The real-time control layer is the front-end unit for direct interaction between the digital twin system and the physical entity of the magnetically driven rotor. It undertakes the critical tasks of microsecond-level high-speed data acquisition, twin model state iteration, and control signal output. This layer uses a Field Programmable Gate Array (FPGA) as its core computing unit, supplemented by a high-performance microprocessor (ARM Cortex series) and a high-precision analog-to-digital / digital-to-analog (AD / DA) conversion module to ensure deterministic system response and low latency. The core functional modules of this layer include: a high-frequency data acquisition and preprocessing module, a twin model state iteration module, and a fast control decision and output module.

[0052] The high-frequency data acquisition and data preprocessing module is directly connected to the sensing system of the magnetic drive rotor system, which mainly includes displacement sensors, current sensors, vibration sensors and other devices. It realizes signal operations such as synchronous data acquisition and hardware filtering in the FPGA to ensure data quality and timeliness from the source.

[0053] The twin model state iteration module, serving as the core of the digital twin real-time control layer, employs a reasonably simplified high-fidelity physical model to balance model fidelity and computational complexity. Based on data preprocessed by the high-frequency data acquisition and preprocessing module, this module performs iterative calculations at a high-frequency rate to predict and update system state information such as the rotor speed, center displacement, and control current of the twin model, achieving a high degree of synchronization with the physical motor.

[0054] Furthermore, the twin model state iteration module models the magnetically driven rotor system in a state-space format, effectively balancing the trade-off between model fidelity and computational complexity. Specifically, the state-space model is represented as follows:

[0055]

[0056] in, This is the system state vector, which includes key state variables such as rotor radial displacement, speed, and control current. This is the system input vector, including control inputs such as the power amplifier drive voltage; This is the system output vector; , , , These are the state matrix, input matrix, output matrix, and feedforward matrix, respectively. By abstracting the dynamic characteristics of the magnetically driven rotor into a state-space representation, not only are the core physical laws of rotor dynamics and electromagnetic forces preserved, but the computational dimensionality is also significantly reduced, enabling the model to achieve high-speed iterative computation at the microsecond level on the FPGA hardware platform.

[0057] Based on the preprocessed data from the high-frequency data acquisition and data preprocessing module, this module solves the state-space equations using numerical integration in each control cycle. It then updates the system state information in the twin model, such as rotor speed, center displacement, and control current, using forward prediction, thereby achieving a high degree of synchronization between the virtual model and the physical motor's operating state.

[0058] The rapid control decision and output module executes a high-real-time control algorithm based on the system state predicted by the twin model state iteration module for the next moment. Specifically, the control algorithm adopts a distributed proportional-integral-derivative control strategy, decoupling the multi-degree-of-freedom magnetic bearing control system into independent control channels. An independent PID controller is designed for each channel, enabling parallel and independent control of key parameters such as rotor displacement and current. This distributed control structure effectively avoids control lag and stability problems caused by inter-channel coupling effects, significantly improving the system's dynamic response speed and control accuracy. Control parameters are preset according to the actual operating characteristics of the magnetically driven rotor and can be dynamically adjusted through the upper-level system. Finally, the control quantity is output to the power amplifier through a high-precision digital-to-analog converter module, driving the magnetically driven rotor actuator to form a high-determinism, low-latency, high-frequency closed-loop control loop, thereby significantly improving the system's control bandwidth and operational stability.

[0059] The monitoring and interaction layer, serving as the human-computer interaction and data scheduling center, runs on an industrial-grade computer or an operating system equipped with real-time patches, and undertakes functions such as system monitoring, command scheduling, and data preprocessing. The core functional modules of this layer include: a system monitoring and visualization module, a command issuance and configuration management module, and a data aggregation and short-term storage module.

[0060] The system monitoring and visualization module dynamically displays the motor operating status, digital twin parameters, alarm information and performance indicators through a graphical human machine interface (HMI).

[0061] The instruction issuance and configuration management module receives operation instructions or model update parameters issued by the cloud optimization layer, and forwards them to the real-time control layer in an orderly manner after security verification.

[0062] The data aggregation and short-term storage module caches high-frequency data reported by the real-time control layer, performs lightweight processing and statistical feature extraction, and uploads data for long-term analysis to the cloud optimization layer.

[0063] The cloud-based optimization layer constructs an online model optimization system for adaptive compensation of magnetically driven rotor performance degradation. Deployed on a high-performance cloud server cluster, this layer, through efficient collaboration with its subordinate layers, specifically performs long-term, large-scale, in-depth analysis and online model optimization tasks targeting the dynamic characteristics of the magnetically driven rotor, ensuring that the digital twin model maintains extremely high synchronization fidelity with the physical entity throughout its entire lifecycle. The core functional modules of this layer include: a time-series data storage module, an online parameter update module for the dedicated magnetically driven rotor twin model, and a comprehensive intelligent application service module.

[0064] The time-series data storage module is used to store the full life cycle operation data of multiple magnetic drive rotors in the subordinate subsystem, and classifies and processes the data according to relevant specifications to facilitate the use of other functional modules.

[0065] The time-series data storage module adopts a tag-based storage architecture oriented towards the multi-modal fault characteristics of magnetically driven rotors. This module automatically tags the aggregated data from the monitoring interaction layer with operating condition labels (e.g., normal operation, acceleration / deceleration, suspected vibration anomaly, current overload, etc.), and constructs a time-series data matrix indexed by "equivalent operating hours" based on the magnetically driven rotor bearing life model. This provides an efficient and accurate data foundation for subsequent parameter optimization and intelligent analysis.

[0066] The online parameter update module for the dedicated twin model of the magnetically driven rotor is the key to realizing the large closed-loop control and long-term performance optimization of the present invention. Its innovation lies in the design of a "hierarchical progressive" parameter inversion and calibration mechanism for the time-varying characteristics of the rotor-bearing system in the magnetically driven rotor (such as mass imbalance drift, electromagnetic bearing stiffness decay, etc.).

[0067] First layer: Fast response layer. Based on high-frequency operating data with a sliding time window, online estimation algorithms such as recursive least squares are used to quickly and lightweightly fine-tune rapidly changing parameters in the model (such as transient damping coefficients) in real time, so as to quickly track short-term state fluctuations of the equipment.

[0068] The second layer: Deep optimization layer. This layer is initiated periodically, using labeled historical data over longer periods. It employs deep learning models incorporating physical constraints (such as Physics-Informed Neural Networks, PINN) or evolutionary algorithms based on covariance adaptive adjustment to perform global deep optimization of slowly varying parameters in the model (such as bearing equivalent stiffness and mass eccentricity). The objective function of this process not only minimizes the error between the model output and the actual data but also introduces constraints on physical indicators such as the stability margin and harmonic components of the magnetically driven rotor. This ensures that the optimized parameters not only fit the data but also conform to physical laws, avoiding ill-conditioned solutions.

[0069] The module's final output is an incremental parameter update package, which includes the logic for parameter changes, version numbers, and rollback strategies, ensuring the security and traceability of model updates.

[0070] The integrated intelligent application service module provides intelligent services such as predictive maintenance, energy efficiency optimization, anomaly diagnosis and root cause analysis based on the optimized model and data, and pushes the optimization parameters and strategies to the monitoring and interaction layer.

[0071] The system's collaborative workflow forms a closed-loop data flow, including downlink data flow and uplink data flow;

[0072] The downlink data stream supports the cloud optimization layer to securely send the optimization parameter package to the real-time control layer for model update after verification by the monitoring and interaction layer; operation instructions are transmitted along the same path.

[0073] The uplink data stream supports the real-time control layer to upload system operation status data to the monitoring and interaction layer, and after data aggregation, it is uploaded to the cloud optimization layer for in-depth analysis.

[0074] Example:

[0075] This embodiment provides a magnetically driven rotor twin control system, the overall architecture of which is shown in the attached figure. Figure 1 As shown, its core lies in constructing a layered, heterogeneous hardware and software collaborative architecture, which includes: a real-time control layer, a monitoring and interaction layer, and a cloud optimization layer. The real-time control layer interacts directly with the physical entity of the magnetically driven rotor; the physical entity transmits sensor data to the real-time control layer, and the real-time control layer transmits control signals to the physical entity. The monitoring and interaction layer serves as a hub for human-machine interaction and data scheduling; the real-time control layer transmits status data to the monitoring and interaction layer, and the monitoring and interaction layer transmits update instructions to the real-time control layer. The cloud optimization layer provides deep data analysis and model evolution capabilities; the monitoring and interaction layer transmits aggregated data to the cloud optimization layer, and the cloud optimization layer transmits optimization parameters to the monitoring and interaction layer.

[0076] The real-time control layer is the front-end unit for direct interaction between the digital twin system and the physical entity of the magnetically driven rotor. It undertakes the critical tasks of microsecond-level high-speed data acquisition, twin model state iteration, and control signal output. This layer uses a Field Programmable Gate Array (FPGA) as its core computing unit, supplemented by a high-performance microprocessor (ARM Cortex series) and a high-precision analog-to-digital / digital-to-analog (AD / DA) conversion module to ensure deterministic system response and low latency. A schematic diagram of the core functional logic architecture of the real-time control layer is attached. Figure 2 As shown;

[0077] The high-frequency data acquisition and data preprocessing module is directly connected to the sensing system of the magnetic drive rotor system, which mainly includes displacement sensors, current sensors, vibration sensors and other devices. It realizes synchronous data acquisition, hardware filtering and other signal operations within the FPGA, ensuring data quality and timeliness from the source.

[0078] The FPGA uses the Xilinx Kintex series to implement highly parallel, low-latency digital logic circuits. Specifically, to achieve synchronous data acquisition, the analog-to-digital converter (ADC) has a sampling rate of at least 1 MS / s and a resolution of at least 16 bits, used to acquire signals from displacement, current, and voltage sensors; the digital-to-analog converter (DAC) has an update rate of at least 1 MS / s and a resolution of at least 12 bits, used to output control signals. To implement hardware filtering, multiple parallel finite impulse response (FIR) digital filters are instantiated in the FPGA to perform real-time noise reduction on the original signal. These FIR digital filters are designed based on time-domain difference equations.

[0079]

[0080] in, For the current moment The output, For the filter's first One coefficient, For the current moment Before Input data for one clock cycle, denoted as the order of the filter.

[0081] The twin model state iteration module, serving as the core of the digital twin real-time control layer, employs a reasonably simplified high-fidelity physical model to balance model fidelity and computational complexity. Based on data preprocessed by the high-frequency data acquisition and preprocessing module, this module performs iterative calculations at a high-frequency rate to predict and update system state information such as the rotor speed, center displacement, and control current of the twin model, achieving a high degree of synchronization with the physical motor.

[0082] Furthermore, the twin model state iteration module, as the core of the digital twin real-time control layer, adopts a reasonably simplified high-fidelity physical model. This model innovatively uses a state-space form to model the magnetically driven rotor system, effectively balancing the contradictory relationship between model fidelity and computational complexity. Specifically, the state-space model is represented as follows:

[0083]

[0084] in, This is the system state vector, which includes key state variables such as rotor radial displacement, speed, and control current. This is the system input vector, including control inputs such as the power amplifier drive voltage; This is the system output vector; , , , These are the state matrix, input matrix, output matrix, and feedforward matrix, respectively. By abstracting the dynamic characteristics of the magnetically driven rotor into a state-space representation, not only are the core physical laws of rotor dynamics and electromagnetic forces preserved, but the computational dimensionality is also significantly reduced, enabling the model to achieve high-speed iterative computation at the microsecond level on the FPGA hardware platform.

[0085] Based on the preprocessed data from the high-frequency data acquisition and data preprocessing module, this module solves the state-space equations using numerical integration in each control cycle. It then updates the system state information in the twin model, such as rotor speed, center displacement, and control current, using forward prediction, thereby achieving a high degree of synchronization between the virtual model and the physical motor's operating state.

[0086] The fast control decision and output module executes a high real-time control algorithm based on the system state predicted by the twin model state iteration module for the next time step. Specifically, the control algorithm employs a distributed proportional-integral-derivative (PID) control algorithm. This algorithm, designed for the independent control of multiple degrees of freedom of the magnetically driven rotor, decouples the system into multiple single-input single-output (SISO) control channels. For any given control channel, its control law is defined by the following discrete PID equation:

[0087]

[0088] in, For the current number The control quantity calculated in the current control cycle is the control quantity of the current control cycle. Control deviation per control cycle; Next-moment state predicted by the twin model With respect to the system's expected settings The difference constitutes, i.e. ; , , These are the proportional, integral, and differential gain coefficients for that channel, respectively. This refers to the system's real-time control cycle.

[0089] Each channel in the distributed control system has an independent and parallel-executing PID controller and an independent parameter set. (where subscript) (Representing channel identifiers) effectively eliminates dynamic coupling interference between multiple channels, simplifies controller design, and allows for precise parameter tuning based on the specific dynamic characteristics of each degree of freedom, thereby significantly improving the system's response speed and control accuracy.

[0090] The initial control parameters are preset based on the dynamic model of the magnetically driven rotor and engineering experience. Finally, the calculated control quantities for each channel are... The high-precision digital-to-analog converter (DA) synchronously outputs current to the corresponding power amplifier, which then drives the magnetically driven rotor, forming a high-frequency closed-loop control loop based on a combination of model predictive feedforward and distributed PID feedback. This loop significantly improves the system's control bandwidth, anti-interference capability, and operational stability. Furthermore, all control parameters can be dynamically configured and updated through the monitoring and interaction layer.

[0091] The monitoring and interaction layer, serving as the human-computer interaction and data scheduling center, runs on an industrial-grade computer or an operating system equipped with real-time patches, and undertakes functions such as system monitoring, command scheduling, and data preprocessing. The core functional modules of this layer include: a system monitoring and visualization module, a command issuance and configuration management module, and a data aggregation and short-term storage module.

[0092] The system monitoring and visualization module dynamically displays the motor operating status, digital twin parameters, alarm information and performance indicators through a graphical human machine interface (HMI).

[0093] The HMI is developed based on the high-level programming languages ​​C++ / Python and the Qt graphics library. The HMI supports real-time display of magnetic drive rotor speed, rotor trajectory, magnetic bearing control current and alarm information, and also supports monitoring and adjustment of key parameters of the digital twin model.

[0094] The instruction issuance and configuration management module receives operation instructions or model update parameters issued by the cloud optimization layer, and forwards them to the real-time control layer in an orderly manner after security verification.

[0095] The data aggregation and short-term storage module runs a real-time database and performs a high-frequency data caching service. It caches the high-frequency data reported by the real-time control layer, performs lightweight processing and statistical feature extraction, and uploads the data for long-term analysis to the cloud optimization layer.

[0096] The cloud optimization layer is deployed on a cloud server cluster based on Kubernetes container orchestration and integrates the distributed big data processing platform Apache Spark and the time-series database TimescaleDB. Its core innovation lies in performing online model optimization tasks for adaptive compensation of magnetically driven rotor performance degradation, rather than simply offline massive data storage. The core functional modules of this layer include: a time-series data storage module, an online parameter update module for a dedicated twin model of the magnetically driven rotor, and a comprehensive intelligent application service module.

[0097] The time-series data storage module adopts a tag-based storage architecture oriented towards the multi-modal fault characteristics of magnetically driven rotors. This module automatically attaches operating condition tags (such as "smooth operation," "acceleration process," and "axial vibration anomaly warning") generated by the intelligent application service module or preset rules to the aggregated data uploaded from the monitoring interaction layer. Based on the magnetically driven rotor bearing life model, it constructs a time-series data matrix indexed by "equivalent operating hours," thereby achieving efficient and accurate data retrieval and analysis.

[0098] The online parameter update module for the dedicated twin model of the magnetically driven rotor is one of the key innovations of this embodiment, and its workflow is shown in the attached figure. Figure 3 As shown, the specific steps include:

[0099] Step S1: Collect historical and real-time operational data;

[0100] The data originates from the real-time control layer, is aggregated by the monitoring and interaction layer, and is stored in a specific data cube of the time-series data storage module.

[0101] Step S2: Data Preprocessing and Feature Extraction

[0102] Perform multi-timescale data preprocessing and feature extraction. For rapid response requirements, extract data within the most recent sliding time window, standardize it, and calculate short-term statistical features (such as mean and variance). For deep optimization requirements, extract data under specific operating condition labels over longer periods and perform frequency domain feature analysis (such as extracting key harmonic components obtained through Fourier transform) to capture performance degradation trends.

[0103] Step S3: Construct the objective function for parameter optimization

[0104] Construct a hierarchical and progressive objective function for parameter optimization; the first objective function is for rapidly changing parameters, aiming to minimize the instantaneous error between recent data and model output; the second objective function is for slowly changing parameters, aiming to minimize the average error between long-term data and model output, and introduces physical constraint terms to ensure that the system stability margin of model calculation is greater than the safety threshold.

[0105] Step S4: Solve iteratively using an optimization algorithm.

[0106] A hierarchical optimization algorithm is used for iterative solution. For rapidly changing parameters, recursive least squares method is used for online real-time estimation. For slowly changing parameters, a deep learning model that incorporates physical constraints (such as Physics-Informed Neural Networks, PINN) or an evolutionary algorithm based on covariance adaptive adjustment is used to periodically perform global optimization. The optimized parameters mainly include bearing stiffness coefficient, damping coefficient, mass imbalance, etc.

[0107] Step S5: Output the optimized model parameters

[0108] Generate a traceable parameter update package, which encapsulates the optimized parameters, version number, checksum, and rollback instructions into an incremental update package;

[0109] Step S6: Generate a parameter update package and send it to the monitoring interaction layer;

[0110] The update package is pushed to the monitoring and interaction layer via an encrypted interface, and then forwarded to the real-time control layer to complete the online model update.

[0111] The operational depth of the integrated intelligent application service module depends on the high-fidelity model generated by the aforementioned online parameter update module. Specifically, it includes:

[0112] Predictive maintenance services: Using optimized models, the operating status of motors under different load spectra in the future is simulated to accurately predict the remaining useful life (RUL) of key components (such as electromagnetic bearings).

[0113] Anomaly Diagnosis and Root Cause Analysis Service: When the system alarms, this service quickly identifies abnormal operating conditions such as model parameter drift and sensor failure by reproducing the operating conditions at the moment of failure in the cloud and comparing the output differences between the optimized model and the mismatched model.

[0114] The system's collaborative workflow forms a closed-loop data flow, including downlink data flow and uplink data flow;

[0115] The downlink data stream supports the cloud optimization layer to securely send the optimization parameter package to the real-time control layer for model update after verification by the monitoring and interaction layer; operation instructions are transmitted along the same path.

[0116] The uplink data stream supports the real-time control layer to upload system operation status data to the monitoring and interaction layer, and after data aggregation, it is uploaded to the cloud optimization layer for in-depth analysis.

[0117] This invention provides a monitoring and interaction layer software interface that meets the operational requirements of magnetically driven rotor systems. It can intuitively characterize the system's operating status, key control parameters, and fault conditions, effectively supporting the realization of the aforementioned functions.

[0118] In summary, through the above specific embodiments, this invention discloses a relatively complete magnetic drive rotor digital twin control system and its construction method, which constructs a high-performance, high-reliability dedicated magnetic drive rotor digital twin system.

[0119] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention content of a magnetically driven rotor twin control system provided by this invention, as well as some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0120] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0121] This invention provides a concept and method for a magnetically driven rotor twin control system. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A magnetically driven rotor twin control system, characterized in that, include: The system consists of a real-time control layer, a monitoring and interaction layer, and a cloud optimization layer, connected sequentially. The real-time control layer interacts directly with the physical entity of the magnetically driven rotor and is used for data acquisition, twin model state iteration, and control signal output. The monitoring and interaction layer is used for system monitoring, human-computer interaction, data scheduling, and command management; The cloud optimization layer is used to perform in-depth analysis and model optimization.

2. The magnetically driven rotor twin control system according to claim 1, characterized in that, The real-time control layer includes: The high-frequency data acquisition and data preprocessing module is connected to the sensing system of the magnetically driven rotor to perform synchronous data acquisition and hardware filtering to obtain preprocessed data. The twin model state iteration module iteratively updates the twin model based on preprocessed data; The fast control decision and output module executes a control algorithm based on the predicted state of the twin model and outputs control signals.

3. The magnetically driven rotor twin control system according to claim 2, characterized in that, The real-time control layer is implemented using a hardware architecture, including: Field-programmable gate arrays (FPGAs), microprocessors, and AD / DA conversion modules.

4. A magnetically driven rotor twin control system according to claim 3, characterized in that, The field-programmable gate array (FPGA) performs synchronous data acquisition and instantiates multiple parallel FIR digital filters for real-time signal denoising.

5. A magnetically driven rotor twin control system according to claim 4, characterized in that, The twin model state iteration module models the magnetically driven rotor system in a state-space form and predicts the system state at the next moment. The state-space model is iteratively calculated on a field-programmable gate array (FPGA) hardware platform.

6. A magnetically driven rotor twin control system according to claim 5, characterized in that, The rapid control decision and output module controls the system based on the system state predicted by the twin model state iteration module at the next moment; wherein, The control method adopts a distributed proportional-integral-derivative control strategy, which decouples the multi-degree-of-freedom magnetic bearing control system into independent control channels. An independent PID controller is designed for each channel to control key parameters in parallel and independently. The control quantity is output to the power amplifier through the digital-to-analog converter module to drive the magnetic drive rotor actuator, forming a closed-loop control circuit.

7. A magnetically driven rotor twin control system according to claim 6, characterized in that, The monitoring interaction layer includes: The system monitoring and visualization module is used to dynamically display the operating status of the magnetically driven rotor through a graphical human-machine interface; The instruction issuance and configuration management module is used to receive and securely verify operation instructions or model update parameters; The data aggregation and short-term storage module is used to cache high-frequency data and perform preliminary processing and feature extraction.

8. A magnetically driven rotor twin control system according to claim 7, characterized in that, The cloud optimization layer includes: The timing data storage module is used to store the entire lifecycle operation data of the magnetic drive rotor; The twin model parameter update module is used to invert and optimize the twin model parameters based on machine learning or data assimilation techniques. The integrated intelligent application service module is used to provide predictive maintenance, energy efficiency optimization, and anomaly diagnosis services.

9. A magnetically driven rotor twin control system according to claim 8, characterized in that, The online parameter update module for the twin model performs hierarchical, progressive parameter inversion and calibration for the time-varying characteristics of the rotor-bearing system in the magnetically driven rotor, including: The first layer, the fast response layer, uses online estimation algorithms based on high-frequency operating data from a sliding time window to fine-tune rapidly changing parameters in the model in real time and track short-term fluctuations in the equipment's state. The second layer, the deep optimization layer, is activated periodically. Based on labeled historical data over a longer period, it uses a deep learning model that incorporates physical constraints or an evolutionary algorithm that adapts to covariance to perform global deep optimization of the slowly varying parameters in the model.

10. A magnetically driven rotor twin control system according to claim 9, characterized in that, The twin model parameter update module has the following workflow: Step 1: Collect historical and real-time operational data; Step 2, Data preprocessing and feature extraction; Step 3: Construct the objective function for parameter optimization; Step 4: Solve iteratively using an optimization algorithm; Step 5: Output the optimized model parameters; Step 6: Generate a parameter update package and send it to the monitoring interaction layer.

Images