A magneto-rheological damper online identification control system and method

Abstract

This invention relates to the field of magnetorheological damper (MRD) control technology, and discloses an online identification control system and method for MRDs, comprising: a multimodal data acquisition and preprocessing module, an online identification module, an FDI module, an adaptive control reconstruction module, a reinforcement learning-optimized control strategy module, and an MRD precision execution and actuator state monitoring module. This invention achieves real-time, high-precision online identification of key MRD operating parameters and can dynamically track their changes, laying the foundation for precise control; it enables intelligent and autonomous operation of the MRD system, improving its overall reliability, stability, and service life in structural vibration control, and providing stronger technical support for health monitoring and active vibration reduction of complex engineering structures.

Description

Technical Field

[0001] This invention relates to the field of magnetorheological damper control technology, and more specifically, to an online identification control system and method for magnetorheological dampers. Background Technology

[0002] Magnetorheological dampers (MRDs) are core actuators for vibration control of complex engineering structures due to their advantages of fast response, adjustable damping, and low energy consumption. Their control effect is highly dependent on accurate modeling and real-time control of nonlinear damping characteristics.

[0003] Existing MRD control systems suffer from numerous technical defects in practical engineering applications, hindering their reliability and control performance: First, they cannot update parameters in real time based on system operating status, relying solely on fixed parameter models identified offline. Parameter drift caused by factors such as temperature and aging continuously increases the deviation between the model and the actual system, reducing control accuracy. Second, they fail to accurately detect and diagnose faults based on sensor data and model parameters, lacking real-time isolation capabilities for complex faults in the MRD itself and sensors, easily leading to control system malfunctions after a fault occurs. Third, they cannot adaptively adjust control strategies based on fault information, lacking an effective self-healing mechanism; even small-scale faults can cause a continuous decline in control performance. Fourth, AI control strategies do not integrate real-time system status and fault information, relying solely on idealized model training, resulting in insufficient robustness to parameter changes and faults. Fifth, the functional modules are independent, failing to form a logical closed loop of "data acquisition - parameter identification - fault detection - control reconfiguration - execution monitoring," resulting in low system integration and an inability to achieve intelligent and autonomous operation.

[0004] To address the aforementioned issues, there is an urgent need to develop an MRD control system based on logical data flow and module collaboration to achieve fully closed-loop online identification, self-diagnosis, and self-healing, thereby improving the system's performance under complex operating conditions. Summary of the Invention

[0005] To achieve the above objectives, this application provides an online identification and control system for magnetorheological dampers, comprising: a multimodal data acquisition and preprocessing module, an online identification module, an FDI module, an adaptive control reconstruction module, a reinforcement learning optimized control strategy module, and an MRD precise execution and actuator state monitoring module; The multimodal data acquisition and preprocessing module is used to: acquire multimodal data, preprocess the multimodal data, obtain identification data, and transmit the identification data to the online identification module, the FDI module, and the reinforcement learning optimization control strategy module. The online identification module is used to: construct an identification model, update the identification model parameters according to the identification data, detect the changing trend of the identification model parameters and predict the future parameter evolution, obtain the identification model parameters and parameter prediction information, and transmit the parameter prediction information to the FDI module and the identification model parameters to the reinforcement learning optimization control strategy module. The FDI module is used to: generate fault information based on identification data and parameter prediction information, and transmit the fault information to the adaptive control reconstruction module and the reinforcement learning optimized control strategy module; The adaptive control reconfiguration module is used to: analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, generate control strategy information, and transmit the control strategy information to the reinforcement learning optimization control strategy module. The reinforcement learning optimization control strategy module is used to: generate the optimal control command based on the identification data, identification model parameters, fault information and control strategy information, and transmit the optimal control command to the MRD precise execution and actuator status monitoring module; The MRD precision execution and actuator status monitoring module is used to: calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD operating status in real time, acquire operating status data, and transmit the operating status data to the multimodal data acquisition and preprocessing module.

[0006] Furthermore, the multimodal data acquisition and preprocessing module includes: an MRD state sensor, a drive circuit state sensor, and a controlled structure response sensor; The MRD status sensor is used to: measure the displacement and velocity of the MRD piston, measure the temperature of the MRD coil, and measure the magnetic field strength of the coil; The drive circuit status sensor is used to measure the output voltage and current of the drive circuit. The controlled structure response sensor is used to: collect vibration response data of the controlled structure.

[0007] Furthermore, the construction of the identification model includes the following steps: S101. Select a physical model to describe the nonlinear hysteresis effect of MRD; S102. Update the key parameters of the physical model using an online estimation algorithm; S103. Train the neural network to learn the mapping from the input to the output of the physical model and build the recognition model framework; S104. Introduce a dynamic model with parameters into the identification model framework to obtain the identification model.

[0008] Furthermore, the physical model supports the Bouc-Wen model, specifically expressed as follows: in, It is the damping force output by the MRD. It is the speed of the piston. It is the displacement of the piston. It is an internal variable, and its evolution satisfies a nonlinear differential equation. It is control Evolutionary parameters.

[0009] Furthermore, the FDI module includes: a fault detection unit, a fault diagnosis unit, and a fault isolation unit; The fault detection unit is used to: compare the identified data and parameter prediction information, calculate the residual, analyze the residual sequence through the fault detection neural network, identify whether a fault has occurred, and set a fault threshold. The fault diagnosis unit is used to: train a neural network to learn the mapping from identification data to fault categories, build a fault model, and obtain the fault category through the fault model; The fault isolation unit is used to: analyze and identify the response pattern of data when a fault occurs, and isolate the source of the fault through the fault isolation unit neural network.

[0010] Furthermore, the adaptive control reconfiguration module includes: a fault impact analysis unit and an adaptive control strategy adjustment unit; The fault impact analysis unit is used to: analyze the impact of faults on the performance of the control system based on fault information and obtain fault impact information; The adaptive control strategy adjustment unit is used to: adjust the control strategy according to fault information and fault impact information, and generate control strategy information.

[0011] Furthermore, adjusting the control strategy based on fault information and fault impact information includes the following steps: S201. Determine the severity of the fault based on the fault impact information. If the fault severity is minor, proceed to step S202. If the fault severity is moderate, proceed to step S203. If the fault severity is severe, proceed to step S204. S202. Adjust the control algorithm parameters according to the fault information. The control algorithm parameters include: sensor software calibration and controller gain adjustment. S203. Replace the identification model with a simplified model and reconfigure the control algorithm; S204, Switch to redundant sensors and backup control channels.

[0012] Furthermore, the step of generating the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information includes the following steps: S301. Construct a fusion state space using a data fusion algorithm based on identification data, identification model parameters, fault information, and control strategy information. S302. Obtain the expected output of MRD and construct the motion space based on the expected output of MRD. S303. Introduce a reward function and a DRL agent. The DRL agent generates the optimal control instructions based on the state space, action space and reward function. The reward elements of the reward function include: control performance, system robustness, MRD energy consumption, and fault compensation; The DRL agent employs an Actor-Critic class algorithm.

[0013] Furthermore, the MRD precision execution and actuator status monitoring module includes: an MRD control signal generation unit, an MRD coil drive unit, and an actuator status feedback unit; The MRD control signal generation unit is used to: calculate the voltage and current signals applied to the MRD according to the optimal control command through an inverse model; The MRD coil drive unit is used to: apply voltage and current to the MRD coil according to the voltage and current signals applied to the MRD, generate an electromagnetic field, and control the MRD damping force; The actuator status feedback unit is used to: monitor the actual voltage, current and temperature of the MRD coil in real time, monitor the working status of the MRD coil drive unit, acquire monitoring information, and feed the monitoring information back to the FDI module and the online identification module.

[0014] An online identification and control method for a magnetorheological damper, implemented through the aforementioned online identification and control system for a magnetorheological damper, includes the following steps; S401. Collect multimodal data and preprocess the multimodal data to obtain identification data; S402. Load the identification model, update the identification model parameters according to the identification data, detect the trend of the identification model parameter change and predict the future parameter evolution, and obtain the identification model parameters and parameter prediction information. S403. Generate fault information based on identification data and parameter prediction information; S404. Analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, and generate control strategy information. S405. Generate the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information; S406. Calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD running status in real time, and obtain running status data. The running status data is used to obtain identification data in step S401. S407. Repeat steps S401-S406 to achieve full closed-loop real-time online identification and control of the MRD until the online identification and control of the MRD ends.

[0015] The beneficial effects of this invention are as follows: 1. Real-time, high-precision online identification of key MRD operating parameters, and dynamic tracking of their changes, laying the foundation for precise control; 2. The FDI module integrates AI-driven technology, enabling rapid, accurate, and reliable detection, diagnosis, and isolation of MRD and related sensor faults; 3. A self-healing mechanism is achieved through the adaptive control reconfiguration module, enabling the system to automatically adjust control strategies or parameters when a small-scale fault or performance degradation is detected, in order to compensate for the impact of the fault, maintain system performance, and extend uptime. 4. The reinforcement learning optimization control strategy module enhances the robustness and adaptability of the control algorithm, and can work in conjunction with the online identification and FDI modules to ensure that the control performance of MRD remains excellent and reliable under various complex working conditions and fault conditions. 5. To achieve intelligent and autonomous operation of the MRD system, improve its overall reliability, stability and service life in structural vibration control, and provide stronger technical support for health monitoring and active vibration reduction of complex engineering structures. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structure of the online identification and control system for magnetorheological dampers provided in an embodiment of the present invention; Figure 2 This is a flowchart of the online identification and control method for magnetorheological dampers provided in an embodiment of the present invention. Detailed Implementation

[0017] The specific implementation of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] like Figure 1 As shown, the present invention provides an online identification and control system for magnetorheological dampers, comprising: a multimodal data acquisition and preprocessing module, an online identification module, an FDI module, an adaptive control reconstruction module, a reinforcement learning optimized control strategy module, and an MRD precise execution and actuator state monitoring module; FDI stands for Fault Detection and Isolation.

[0019] The multimodal data acquisition and preprocessing module is used to: acquire multimodal data, preprocess the multimodal data, obtain identification data, and transmit the identification data to the online identification module, the FDI module, and the reinforcement learning optimization control strategy module. The multimodal data acquisition and preprocessing module includes: an MRD state sensor, a drive circuit state sensor, and a controlled structure response sensor; MRD status sensors include: linear encoder and LDV-laser Doppler velocimeter; The MRD status sensor is used to: measure the displacement and velocity of the MRD piston, measure the temperature of the MRD coil, and measure the magnetic field strength of the coil; The drive circuit status sensor is used to measure the output voltage and current of the drive circuit. The controlled structure response sensor is used to: collect vibration response data of the controlled structure.

[0020] In this embodiment, the multimodal data acquisition and preprocessing module further includes: an environmental parameter sensor; Environmental parameter sensors, including anemometers and wave height meters, are used for more comprehensive system perception in order to make more optimized control decisions.

[0021] Each sensor needs to collect data at a high sampling rate and perform strict time synchronization to ensure accurate correspondence between data from different sensors. In this embodiment, the sampling rate is set to above 1kHz, and the specific frequency depends on the dynamic characteristics requirements of the application scenario.

[0022] Preprocessing of multimodal data includes filtering, calibration, and outlier detection; Specifically, filtering includes low-pass filtering to remove high-frequency noise and median filtering to handle occasional spike signals; calibration compensates for sensor zero-point drift and gain error to ensure accurate readings; and outlier detection identifies and addresses obviously unreasonable sensor readings.

[0023] The online identification module is used to: construct an identification model, update the identification model parameters according to the identification data, detect the changing trend of the identification model parameters and predict the future parameter evolution, obtain the identification model parameters and parameter prediction information, and transmit the parameter prediction information to the FDI module and the identification model parameters to the reinforcement learning optimization control strategy module. The construction of the identification model includes the following steps: S101. Select a physical model to describe the nonlinear hysteresis effect of MRD; The physical model supports the Bouc-Wen model, specifically expressed as follows: in, It is the damping force output by the MRD. It is the speed of the piston. It is the displacement of the piston. It is an internal variable, and its evolution satisfies a nonlinear differential equation. It is control Evolutionary parameters.

[0024] In this embodiment, the Bouc-Wen model or a variant thereof is used to describe the nonlinear hysteresis characteristics of the MRD; these parameters describe the linear damping, viscous damping, hysteresis loop shape, and saturation characteristics of the MRD.

[0025] S102. Update the key parameters of the physical model using an online estimation algorithm; Key parameters of the Bouc-Wen model are updated in real time using online estimation algorithms, such as recursive least squares or extended Kalman filters. These algorithms can use newly acquired data to gradually adjust the model parameters to minimize prediction errors.

[0026] S103. Train the neural network to learn the mapping from the input to the output of the physical model and build the recognition model framework; Neural networks such as Long Short-Term Memory (LSTM) or Gated Recurrent Units (GRUs) can be used to directly learn the mapping from input to output, such as the mapping from input parameters like magnetic field strength, velocity, and displacement to damping force, or more advancedly, to learn the real-time values ​​of model parameters; this approach can better capture complex nonlinear dynamics.

[0027] S104. Introduce a dynamic model with parameters into the identification model framework to obtain the identification model.

[0028] Dynamic models that introduce parameters, for example, can assume that certain parameters have a linear or nonlinear relationship with temperature, or decay exponentially with usage time; the AI ​​model will dynamically adjust the parameter estimates based on real-time temperature, working time and other information to make them closer to reality.

[0029] The AI ​​online identification module in this application is not just a simple parameter update, but can actively detect the changing trends of parameters, such as the rate of change and acceleration, and predict the future evolution of parameters based on the changing trends.

[0030] For example, if an increase in temperature is detected causing a certain hysteresis-related parameter to drift faster, the AI ​​will predict its future value based on historical data and current trends, and adjust the control strategy in advance, such as increasing the expected magnetic field strength to compensate for possible performance degradation, rather than waiting until the parameter has completely drifted before compensation is performed.

[0031] The FDI module is used to: generate fault information based on identification data and parameter prediction information, and transmit the fault information to the adaptive control reconstruction module and the reinforcement learning optimized control strategy module; The FDI module includes: a fault detection unit, a fault diagnosis unit, and a fault isolation unit; The fault detection unit is used to: compare the identified data and parameter prediction information, calculate the residual, analyze the residual sequence through the fault detection neural network, identify whether a fault has occurred, and set a fault threshold. Specifically, the identification data and parameter prediction information are compared, the residuals are calculated, and the statistical characteristics and patterns of the residual sequence are analyzed using a fault detection neural network to identify whether a fault has occurred and to set a fault threshold. Among them, the fault detection neural network is a support vector machine (SVM), an isolation forest (Isolation Forest), or a specially trained fault detection neural network.

[0032] The fault diagnosis unit is used to: train a neural network to learn the mapping from identification data to fault categories, build a fault model, and obtain the fault category through the fault model; The neural network, such as a multilayer perceptron (MLP), decision tree, or attention-based neural network, is trained to learn the mapping from identification data to fault categories. In this embodiment, in addition to identification data, residuals, identified model parameters, and drive signals are also included. Fault categories include "sensor A zero-point drift", "coil aging", and "drive circuit overload".

[0033] The fault isolation unit is used to: analyze and identify the response pattern of data when a fault occurs, and isolate the source of the fault through the fault isolation unit neural network.

[0034] Specifically, the fault isolation unit neural network supports mining based on association rules or graph neural networks (GNNs), the latter being particularly suitable for analyzing dependencies between components; thereby determining which component or parameter has a problem.

[0035] The FDI module does not merely passively detect faults, but utilizes AI to analyze the changing trends of model parameters, providing early warnings of potential faults and distinguishing between sensor problems and MRD-related issues. For example, if the reading of a displacement sensor continuously deviates from the estimated value derived from other information, it is determined that the sensor is faulty; if the actual damping force output by the MRD is consistently lower than the theoretical value predicted by the drive signal and the identified model parameters, it is determined that the MRD performance has degraded.

[0036] The adaptive control reconfiguration module is used to: analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, generate control strategy information, and transmit the control strategy information to the reinforcement learning optimization control strategy module. The adaptive control reconfiguration module includes: a fault impact analysis unit and an adaptive control strategy adjustment unit; The fault impact analysis unit is used to: analyze the impact of faults on the performance of the control system based on fault information and obtain fault impact information; The adaptive control strategy adjustment unit is used to: adjust the control strategy according to fault information and fault impact information, and generate control strategy information.

[0037] The adjustment of the control strategy based on fault information and fault impact information includes the following steps: S201. Determine the severity of the fault based on the fault impact information. If the fault severity is minor, proceed to step S202. If the fault severity is moderate, proceed to step S203. If the fault severity is severe, proceed to step S204. S202. Adjust the control algorithm parameters according to the fault information. The control algorithm parameters include: sensor software calibration and controller gain adjustment. For example, if the output capability of the MRD damping force decreases, the gain of the controller is increased in order to achieve the same damping effect; if the sensor zero point drifts, software calibration is automatically performed to correct the sensor reading.

[0038] S203. Replace the identification model with a simplified model and reconfigure the control algorithm; For example, if the displacement sensor fails, the state is estimated based solely on accelerometer feedback and model predictions, and the controller is adjusted accordingly.

[0039] S204, Switch to redundant sensors and backup control channels.

[0040] If redundant sensors or backup control channels exist, automatic switching is implemented in this step to maintain the basic functions of the system.

[0041] The core of the adaptive control reconfiguration module is the "flexible" adjustment of the control strategy. That is, it does not pursue the complete reproduction of the performance when there is no fault after a fault occurs. Instead, it aims to maintain the basic stability and acceptable control effect of the system as quickly as possible and with the least performance loss, so as to avoid system collapse.

[0042] The reinforcement learning optimization control strategy module is used to: generate the optimal control command based on the identification data, identification model parameters, fault information and control strategy information, and transmit the optimal control command to the MRD precise execution and actuator status monitoring module; The process of generating the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information includes the following steps: S301. Construct a fusion state space using a data fusion algorithm based on identification data, identification model parameters, fault information, and control strategy information. In this embodiment, the predicted value of the structural response is also included as the state space.

[0043] S302. Obtain the expected output of MRD and construct the motion space based on the expected output of MRD. S303. Introduce a reward function and a DRL agent. The DRL agent generates the optimal control instructions based on the state space, action space and reward function. The reward elements of the reward function include: control performance, system robustness, MRD energy consumption, and fault compensation; Specifically, control performance refers to minimizing structural response, such as tower top acceleration and displacement; system robustness specifically refers to the ability to tolerate certain model uncertainties and faults; MRD energy consumption refers to minimizing the energy consumption of the drive coil; fault compensation, also known as self-healing, specifically refers to providing a certain "survival" reward when the system is in fault compensation mode to encourage the system to maintain operation under harsh conditions.

[0044] The reward function explicitly includes evaluations of system robustness, fault tolerance, and "self-healing" performance.

[0045] This means that during the learning process, the DRL agent will be guided to find control strategies that still perform well under complex operating conditions and minor faults.

[0046] The DRL agent employs an Actor-Critic class algorithm.

[0047] Actor-Critic type algorithms, such as DDPG, SAC, and PPO, are particularly suitable for continuous action spaces.

[0048] During system operation, the intelligent agent, or Actor, generates control actions using real-time collected state information; the Critic network evaluates the value of these actions; and by updating the Actor and Critic network online, the control strategy can continuously adapt to changes in system state and fault conditions.

[0049] The DRL agent not only learns how to perform optimal control based on normal system states, but also learns how to adjust its action output when a fault or parameter drift is detected, in order to adapt to a new system model or control mode; this achieves deep collaboration between AI control strategies and FDI and control reconfiguration modules.

[0050] The MRD precision execution and actuator status monitoring module is used to: calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD operating status in real time, acquire operating status data, and transmit the operating status data to the multimodal data acquisition and preprocessing module.

[0051] The MRD precision execution and actuator status monitoring module includes: an MRD control signal generation unit, an MRD coil drive unit, and an actuator status feedback unit; The MRD control signal generation unit is used to: calculate the voltage and current signals applied to the MRD according to the optimal control command through an inverse model; Specifically, the inverse model refers to the inverse model of MRD or the AI-driven inverse model. In this embodiment, it is the inverse model of MRD, taking into account the real-time values ​​of the Bouc-Wen model parameters.

[0052] The MRD coil drive unit is used to: apply voltage and current to the MRD coil according to the voltage and current signals applied to the MRD, generate an electromagnetic field, and control the MRD damping force; The actuator status feedback unit is used to: monitor the actual voltage, current and temperature of the MRD coil in real time, monitor the working status of the MRD coil drive unit, acquire monitoring information, and feed the monitoring information back to the FDI module and the online identification module.

[0053] By feeding back the monitoring information to the FDI module and the online identification module, a closed-loop monitoring system for the MRD body and the drive system is formed.

[0054] The MRD precision execution and actuator status monitoring module realizes a closed loop from AI control output to MRD actual output, and the operating status of the MRD body and drive circuit is also included in the monitoring scope of FDI and online identification, forming a complete MRD subsystem-level self-diagnosis and adaptive closed loop.

[0055] An online identification and control method for a magnetorheological damper, implemented through the aforementioned online identification and control system for a magnetorheological damper, includes the following steps; S401. Collect multimodal data and preprocess the multimodal data to obtain identification data; S402. Load the identification model, update the identification model parameters according to the identification data, detect the trend of the identification model parameter change and predict the future parameter evolution, and obtain the identification model parameters and parameter prediction information. S403. Generate fault information based on identification data and parameter prediction information; S404. Analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, and generate control strategy information. S405. Generate the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information; S406. Calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD running status in real time, and obtain running status data. The running status data is used to obtain identification data in step S401. S407. Repeat steps S401-S406 to achieve full closed-loop real-time online identification and control of the MRD until the online identification and control of the MRD ends.

[0056] This invention achieves real-time, high-precision online identification of key operating parameters of the MRD (Mechanical Reduction Device) and can dynamically track its changes, laying the foundation for precise control. The FDI (Fixed-Input Difference) module integrates AI-driven capabilities, enabling rapid, accurate, and reliable detection, diagnosis, and isolation of faults in the MRD and its related sensors. An adaptive control reconfiguration module implements a self-healing mechanism, allowing the system to automatically adjust control strategies or parameters to compensate for fault effects, maintain system performance, and extend uptime when small-scale faults or performance degradation are detected. A reinforcement learning-optimized control strategy module enhances the robustness and adaptability of the control algorithm, working collaboratively with the online identification and FDI modules to ensure excellent and reliable MRD control performance under various complex operating conditions and fault scenarios. This enables intelligent and autonomous operation of the MRD system, improving its overall reliability, stability, and service life in structural vibration control, and providing stronger technical support for health monitoring and active vibration reduction of complex engineering structures.

[0057] The above-disclosed embodiments are merely a few specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. An online identification and control system for magnetorheological dampers, characterized in that, include: Multimodal data acquisition and preprocessing module, online identification module, FDI module, adaptive control reconstruction module, reinforcement learning optimized control strategy module, MRD precise execution and actuator state monitoring module; The multimodal data acquisition and preprocessing module is used to: acquire multimodal data, preprocess the multimodal data, obtain identification data, and transmit the identification data to the online identification module, the FDI module, and the reinforcement learning optimization control strategy module. The online identification module is used to: construct an identification model, update the identification model parameters according to the identification data, detect the changing trend of the identification model parameters and predict the future parameter evolution, obtain the identification model parameters and parameter prediction information, and transmit the parameter prediction information to the FDI module and the identification model parameters to the reinforcement learning optimization control strategy module. The FDI module is used to: generate fault information based on identification data and parameter prediction information, and transmit the fault information to the adaptive control reconstruction module and the reinforcement learning optimized control strategy module; The adaptive control reconfiguration module is used to: analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, generate control strategy information, and transmit the control strategy information to the reinforcement learning optimization control strategy module. The reinforcement learning optimization control strategy module is used to: generate the optimal control command based on the identification data, identification model parameters, fault information and control strategy information, and transmit the optimal control command to the MRD precise execution and actuator status monitoring module; The MRD precision execution and actuator status monitoring module is used to: calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD operating status in real time, acquire operating status data, and transmit the operating status data to the multimodal data acquisition and preprocessing module.

2. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The multimodal data acquisition and preprocessing module includes: an MRD state sensor, a drive circuit state sensor, and a controlled structure response sensor; The MRD status sensor is used to: measure the displacement and velocity of the MRD piston, measure the temperature of the MRD coil, and measure the magnetic field strength of the coil; The drive circuit status sensor is used to measure the output voltage and current of the drive circuit. The controlled structure response sensor is used to: collect vibration response data of the controlled structure.

3. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The construction of the identification model includes the following steps: S101. Select a physical model to describe the nonlinear hysteresis effect of MRD; S102. Update the key parameters of the physical model using an online estimation algorithm; S103. Train the neural network to learn the mapping from the physical model input to the output, and build the recognition model framework; S104. Introduce a dynamic model with parameters into the identification model framework to obtain the identification model.

4. The online identification and control system for magnetorheological dampers according to claim 3, characterized in that, The physical model supports the Bouc-Wen model, specifically expressed as follows: in, It is the damping force output by the MRD. It is the speed of the piston. It is the displacement of the piston. It is an internal variable, and its evolution satisfies a nonlinear differential equation. It is control Evolutionary parameters.

5. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The FDI module includes: a fault detection unit, a fault diagnosis unit, and a fault isolation unit; The fault detection unit is used to: compare the identified data and parameter prediction information, calculate the residual, analyze the residual sequence through the fault detection neural network, identify whether a fault has occurred, and set a fault threshold. The fault diagnosis unit is used to: train a neural network to learn the mapping from identification data to fault categories, build a fault model, and obtain the fault category through the fault model; The fault isolation unit is used to: analyze and identify the response pattern of data when a fault occurs, and isolate the source of the fault through the fault isolation unit neural network.

6. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The adaptive control reconfiguration module includes: a fault impact analysis unit and an adaptive control strategy adjustment unit; The fault impact analysis unit is used to: analyze the impact of faults on the performance of the control system based on fault information and obtain fault impact information; The adaptive control strategy adjustment unit is used to: adjust the control strategy according to fault information and fault impact information, and generate control strategy information.

7. The online identification and control system for magnetorheological dampers according to claim 6, characterized in that, The adjustment of the control strategy based on fault information and fault impact information includes the following steps: S201. Determine the severity of the fault based on the fault impact information. If the fault severity is minor, proceed to step S202. If the fault severity is moderate, proceed to step S203. If the fault severity is severe, proceed to step S204. S202. Adjust the control algorithm parameters according to the fault information. The control algorithm parameters include: sensor software calibration and controller gain adjustment. S203. Replace the identification model with a simplified model and reconfigure the control algorithm; S204, Switch to redundant sensors and backup control channels.

8. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The process of generating the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information includes the following steps: S301. Construct a fusion state space using a data fusion algorithm based on identification data, identification model parameters, fault information, and control strategy information. S302. Obtain the expected output of MRD and construct the motion space based on the expected output of MRD. S303. Introduce a reward function and a DRL agent. The DRL agent generates the optimal control instructions based on the state space, action space and reward function. The reward elements of the reward function include: control performance, system robustness, MRD energy consumption, and fault compensation; The DRL agent employs an Actor-Critic class algorithm.

9. The online identification and control system for magnetorheological dampers according to claim 1, characterized in that, The MRD precision execution and actuator status monitoring module includes: an MRD control signal generation unit, an MRD coil drive unit, and an actuator status feedback unit; The MRD control signal generation unit is used to: calculate the voltage and current signals applied to the MRD according to the optimal control command through an inverse model; The MRD coil drive unit is used to: apply voltage and current to the MRD coil according to the voltage and current signals applied to the MRD, generate an electromagnetic field, and control the MRD damping force; The actuator status feedback unit is used to: monitor the actual voltage, current and temperature of the MRD coil in real time, monitor the working status of the MRD coil drive unit, acquire monitoring information, and feed the monitoring information back to the FDI module and the online identification module.

10. A method for online identification and control of a magnetorheological damper, characterized in that, This is achieved through the online identification and control system for magnetorheological dampers as described in any one of claims 1-9. Includes the following steps; S401. Collect multimodal data and preprocess the multimodal data to obtain identification data; S402. Load the identification model, update the identification model parameters according to the identification data, detect the trend of the identification model parameter change and predict the future parameter evolution, and obtain the identification model parameters and parameter prediction information. S403. Generate fault information based on identification data and parameter prediction information; S404. Analyze the impact of faults on the performance of the control system based on fault information, adjust the control strategy, and generate control strategy information. S405. Generate the optimal control command based on the identification data, identification model parameters, fault information, and control strategy information; S406. Calculate the drive signal according to the optimal control command, output the drive signal to drive the MRD to work, monitor the MRD running status in real time, and obtain running status data. The running status data is used to obtain identification data in step S401. S407. Repeat steps S401-S406 to achieve full closed-loop real-time online identification and control of the MRD until the online identification and control of the MRD ends.

Images