Circuit damper and control method

Abstract

The invention discloses a circuit damper and a control method, and aims to overcome the defect that semi-active control cannot be independently completed by an electromagnetic branch circuit damper in the prior art. The circuit damper comprises a magnetic coupling system, a magnetic circuit centering system, a reset mechanism, a microcontroller and an external adjustable resistor. A permanent magnet group in the magnetic coupling system is connected with a loading rod and generates induced voltage on a coil along with vibration of a controlled structure. The control method is based on the damper, induced voltage is processed into analog quantity through the execution measurement module, the microcontroller receives the analog quantity and then calls the gradient lifting regression model to invert the relative speed, and then the resistance value of the external adjustable resistor is adjusted according to the relative speed to suppress vibration. According to the whole scheme, an external sensor is not needed, independent semi-active control over the damper is achieved through self-sensing and a damping adjustment closed loop, and cost and system complexity are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vibration control, and in particular to a circuit damper and a control method. BACKGROUND

[0002] In the field of building structure vibration resistance and mechanical system vibration reduction, semi-active control has become a key technology direction for vibration suppression due to its control effect and energy economy. Among them, the traditional electromagnetic shunt circuit damper (EMSD) can flexibly adapt to different vibration conditions by adjusting the external variable resistor to change the loop damping characteristics, and is an important executive element in the semi-active control system.

[0003] The existing EMSD scheme has a fundamental technical limitation. It does not build a control parameter sensing mechanism based on the structure and electromagnetic properties of the damper itself, and completely relies on external displacement or acceleration sensors to obtain key vibration parameters. This dependency directly leads to two major technical problems. First, the system independence and economy are insufficient. Since it cannot complete the semi-active control closed loop independently without external sensors, it needs to additionally lay out high-precision sensors, as well as design a power supply circuit, data communication interface and regular calibration device for the sensors. This not only significantly increases the hardware cost of the entire vibration reduction system, but also expands the system size due to the need for multiple component integration, making it difficult to adapt to small and lightweight application scenarios. Second, the control stability and precision are limited. The displacement or acceleration signals measured by the external sensor need to be filtered by a microcontroller unit (MCU) before the relative speed is indirectly obtained through numerical derivative estimation. In this process, signal noise is easily amplified, and derivative estimation has inherent lag and discrete error, resulting in jitter in the output speed signal, making it difficult to stably drive the discrete damper controller, and ultimately affecting the consistency and reliability of the vibration suppression effect.

[0004] In summary, the existing EMSD scheme cannot balance cost control and self-adaptive ability due to the design defect of separate sensing and control. To break through this limitation, a new type of electromagnetic vibration reduction technology scheme is urgently needed that does not rely on external sensors, can directly obtain the relative speed based on the properties of the damper itself in real time, and can complete the damping closed loop regulation in the resource-limited MCU. SUMMARY

[0005] The purpose of the present application is to provide a circuit damper and a control method to overcome the deficiency that the prior art cannot rely on the electromagnetic shunt circuit damper to independently complete semi-active control.

[0006] To achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, the present application provides a circuit damper, comprising: The magnetic coupling system comprises a permanent magnet group and a coil, the permanent magnet group is arranged inside the coil, the permanent magnet group is connected with a loading rod for connecting a controlled structure to load vibration displacement, and the permanent magnet group is driven to move in the coil to generate induced voltage; The magnetic circuit centering system comprises a top partition plate and a bottom partition plate arranged at the top and bottom of the coil respectively to limit the stroke range of the permanent magnet group; The reset mechanism is connected with the permanent magnet group to reset the permanent magnet group; The coil is connected with a microcontroller and an external adjustable resistor, and the microcontroller adjusts the external adjustable resistor according to the induced voltage of the magnetic coupling system.

[0007] The support structure comprises a top cover arranged above the top partition plate and a base arranged below the reset mechanism; the top partition plate and the bottom partition plate are connected together through a support rod, one end of the support rod passes through the top partition plate and is connected with the top cover, and the other end passes through the bottom partition plate and is connected with the base.

[0008] The reset mechanism is a compression spring, the lower end of the compression spring is supported on the base, and the upper end of the compression spring is in contact with the bottom of the permanent magnet group through a spring pad to form support.

[0009] The permanent magnet group adopts a ten-stage alternating magnetic pole axial arrangement configuration, and a space of 2-4 mm is left between the permanent magnet group and the coil.

[0010] In the second aspect, the application provides a circuit damper control method based on the above-mentioned circuit damper, and the method comprises the following steps: The induced voltage generated by the magnetic coupling system is processed into voltage data analog quantity; The relative speed of the permanent magnet group (10) and the coil (5) is inversely calculated according to the voltage data analog quantity by calling a gradient boosting regression model; The resistance value of the external adjustable resistor is adjusted according to the relative speed to suppress the vibration of the controlled structure, and the circuit damper control is completed.

[0011] The gradient boosting regression model is obtained by the following way: The induced voltage and the corresponding relative speed data of the magnetic coupling system under different vibration states are collected to form a training data set; The induced voltage data is preprocessed to reduce quantization noise; The first derivative and the change direction symbol feature of the preprocessed voltage data are extracted, and a lag dimension input sequence is constructed; Based on the input sequence and the corresponding relative speed data, a gradient boosting regression algorithm is used for model training to obtain an initial model, with the square error as the loss function; The initial model is compressed to obtain a gradient boosting regression model suitable for microcontroller deployment.

[0012] The specific way of constructing the hysteresis dimension input sequence is: The feature vector is constructed in the form of , wherein is the voltage value at the current time, is the first derivative of the voltage, is the change direction sign function, is the hysteresis order.

[0013] The gradient boosting regression algorithm is combined by multiple regression trees to form a strong prediction model, and the output expression is:

[0014] wherein, wherein is the number of trees, is the learning rate, is the first regression tree.

[0015] The regression tree adopts a weak regression tree, and the depth of each weak regression tree does not exceed 3.

[0016] The root mean square error and the determination coefficient are used as the model performance evaluation indexes in the model training process.

[0017] Compared with the prior art, the present application has the following beneficial technical effects: In the first aspect, the present application provides a circuit damper, which realizes the basic functions of self-sensing and damping adjustment through a magnetic coupling system: the permanent magnet group moves inside the coil to generate an induced voltage when the controlled structure vibrates, and this voltage is directly used as the control input parameter of the microcontroller, without relying on an external sensor to obtain vibration information, thereby avoiding the problems of cost increase and independence loss caused by additional sensor arrangement in the prior art. The magnetic circuit centering system limits the stroke range of the permanent magnet group, and the reset mechanism ensures its reset after vibration, both of which together ensure the stable output of the induced voltage of the magnetic coupling system and provide a reliable signal source for the microcontroller. The microcontroller adjusts the external adjustable resistor according to the induced voltage, forming a complete closed loop of vibration-induced voltage, voltage-driven resistor adjustment, and resistor-changed loop damping.

[0018] In a second aspect, the present application provides a circuit damper control method, which relies on the above-mentioned damper structure, and processes the induced voltage generated by the magnetic coupling system into voltage data analog quantity through the execution of a measurement module, without relying on external sensors to collect vibration parameters, thereby avoiding the problem of loss of independence caused by sensor dependence in the prior art. After receiving the voltage data analog quantity, the microcontroller directly inverts the relative speed of the permanent magnet group and the coil by calling a gradient boosting regression model, without the need for external equipment to assist in calculating the speed parameter. Then, the microcontroller adjusts the external adjustable resistance value according to the relative speed obtained by inversion, realizes the suppression of the vibration of the controlled structure, forms a complete control process from the processing of the induced voltage to the damping adjustment, and enables the damper to independently complete the semi-active control. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 It is a schematic diagram of the overall structure of a circuit damper in an embodiment of the present application.

[0020] Figure 2 It is a schematic diagram of a circuit damper control system framework in an embodiment of the present application.

[0021] Figure 3 It is a schematic diagram of a circuit damper control method in an embodiment of the present application.

[0022] Figure 4 It is a schematic diagram of the gradient boosting regression model acquisition process in an embodiment of the present application.

[0023] In the figure, 1 is a top cover, 2 is a loading rod, 3 is a top partition plate, 4 is a hexagonal nut, 5 is a coil, 6 is a support rod, 7 is a bottom partition plate, 8 is a compression spring, 9 is a base, 10 is a permanent magnet group, 11 is a spring pad, 12 is a middle reinforcing nut, and 13 is a lower fixing nut. DETAILED DESCRIPTION

[0024] In the field of building structure vibration resistance and mechanical system vibration reduction, semi-active control has become a key direction for vibration suppression because it takes into account both control effect and energy consumption economy. Traditional electromagnetic shunt circuit damper (EMSD) can adapt to different vibration conditions by adjusting the external variable resistance to change the loop damping, and is an important executive element for semi-active control.

[0025] The existing EMSD has a fundamental limitation: it does not construct a control parameter sensing mechanism based on its own structure and electromagnetic properties, and completely relies on external displacement or acceleration sensors to obtain vibration parameters. This directly leads to two major problems: first, the system is not independent and economical, and requires additional high-precision sensors and supporting power supply, communication, and calibration devices, increasing cost and size, making it difficult to adapt to small-scale scenarios; second, the control stability and precision are limited, and the sensor signal is indirectly obtained by microcontroller unit (MCU) filtering and derivative estimation, which is easy to amplify noise and has lag error, leading to speed signal jitter and making it difficult to stabilize the drive controller.

[0026] In summary, due to the separation of sensing and control, the existing EMSD cannot balance cost and adaptive ability, and there is an urgent need for a new solution that does not require external sensors, can obtain speed in real time based on its own properties, and can close-loop adjust damping in a resource-limited MCU.

[0027] According to the above background, the present application proposes a circuit damper and control method, which overcomes the defects of the prior art by constructing its own sensing mechanism through a magnetic coupling system. In the magnetic coupling system, the permanent magnet group moves inside the coil to generate an induced voltage when the controlled structure vibrates, which can be directly used as a control parameter without relying on external displacement or acceleration sensors, avoiding the cost and size burden of additional sensors and supporting devices. At the same time, the microcontroller adjusts the external adjustable resistor according to the induced voltage, so that the damper itself forms a complete sensing-to-control closed loop without the need for external equipment to complete damping adjustment, realizing independent operation of semi-active control and solving the technical problem that the existing technology cannot independently complete semi-active control relying on the electromagnetic branch circuit damper.

[0028] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0029] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0030] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "provided with", "sleeved / connected", "connected" and the like should be understood in a broad sense, for example, "connected" can be fixedly connected, or detachably connected, or integrally connected; can be mechanically connected, or electrically connected; can be directly connected, or indirectly connected through an intermediate medium, or can be internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0031] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "several" is two or more, unless otherwise explicitly specified and limited.

[0032] Referring to Figure 1 The specific embodiment of the circuit damper provided by the present application is shown, which comprises a magnetic coupling system, a magnetic circuit centering system, a reset mechanism and a control module. The magnetic coupling system includes a permanent magnet group 10 and a coil 5, the permanent magnet group 10 is connected to the controlled structure through a loading rod 2, and the induced voltage is generated inside the coil 5. The magnetic circuit centering system is composed of a top partition plate 3 and a bottom partition plate 7, which are respectively arranged at both ends of the coil 5. The reset mechanism connects the permanent magnet group 10 to realize position reset. The microcontroller adjusts the external adjustable resistor according to the induced voltage.

[0033] Among them, the magnetic coupling system refers to a device that converts mechanical vibration into electrical signal through electromagnetic induction principle, the permanent magnet group 10 can adopt multi-stage alternating magnetic pole axial arrangement configuration, and the coil 5 is wound on the non-magnetic skeleton to form a closed loop. The movement direction of the permanent magnet of the system is parallel to the axis of the coil 5, which ensures the linear relationship between the induced voltage and the speed.

[0034] The magnetic circuit centering system refers to the guide structure for limiting the movement range of the permanent magnet group 10, the top partition plate 3 and the bottom partition plate 7 can be made of high permeability material, which can assist the permanent magnet group 10 to keep the axial movement track through magnetic attraction force, and prevent electromagnetic parameter distortion caused by lateral deviation.

[0035] The reset mechanism refers to a mechanical component that provides elastic restoring force, which can use a spiral spring or a disc spring, and its stiffness coefficient matches the natural frequency of the controlled structure, which ensures the periodic transmission of vibration energy.

[0036] The control module refers to an embedded system including a signal conditioning circuit and a data processing unit, comprising a microcontroller and an external adjustable resistor, which collects the voltage signal between the two ends of the coil 5, executes a speed estimation algorithm and outputs a resistance adjustment instruction.

[0037] Specifically, when the controlled structure vibrates, the loading rod 2 drives the permanent magnet group 10 to move axially in the coil 5. According to Faraday's law of electromagnetic induction, the coil 5 cuts the magnetic induction lines to generate an induced voltage proportional to the speed of movement. The voltage signal is input into the microcontroller after being filtered and amplified, and the real-time relative speed is calculated through the pre-set voltage-speed mapping model. The microcontroller dynamically adjusts the external adjustable resistor according to the speed value, changes the electromagnetic damping force formed by the loop current, and thereby suppresses the vibration energy transmission. The double-layer partition of the magnetic circuit centering system limits the stroke of the permanent magnet to prevent overshoot damage, and at the same time maintains the straightness of the movement trajectory by using magnetic attraction force. The reset spring pushes the permanent magnet back to the equilibrium position in the vibration gap, ensuring that the system has sustained response capability.

[0038] Compared with the prior art, the existing scheme needs an external acceleration sensor and a displacement meter to measure vibration parameters respectively, and then obtains the speed signal through differentiation calculation, resulting in a long signal link and being easily disturbed by noise. The present scheme directly uses the electromagnetic induction voltage to reflect the vibration speed, eliminating the need for sensor installation and signal conversion. The permanent magnet guidance of the traditional damper relies on mechanical slide rails, which has friction loss and gap error, while the present scheme realizes non-contact guidance through the magnetic circuit centering system, improving the movement precision and service life.

[0039] Through the above technical scheme, the present application realizes the functional integration of vibration parameter sensing and damping adjustment, eliminating the additional cost and signal distortion problems caused by external sensors. The electromagnetic induction voltage is used to directly represent the vibration speed, avoiding the phase lag and noise amplification caused by numerical differentiation operation, and improving the response speed and stability of the control system. The synergistic effect of the magnetic circuit centering system and the reset mechanism ensures the reliability of the permanent magnet group 10 in long-term operation, enabling the damper to adapt to the continuous working requirements in high-frequency vibration environment.

[0040] In another specific embodiment of the present application, the circuit damper further comprises a support structure, which comprises a top cover 1 arranged above the magnetic circuit centering system and a base 9 arranged below the reset mechanism; the top partition 3 and the bottom partition 7 are connected together through the support rod 6, one end of which is connected with the top cover 1 by penetrating the top partition 3, and the other end is connected with the base 9 by penetrating the bottom partition 7. The support rod 6 is specifically four through-type support threaded rods, forming a stable quadrilateral support frame to ensure the rigidity of the overall structure.

[0041] Wherein, the top cover 1 refers to the rigid plate structure covering the magnetic circuit centering system, which is specifically processed by acrylic disc, with a central reserved through hole for the center loading rod 2 to pass through, the loading rod 2 is rigidly connected with the controlled structure, and the loading rod 2 and the permanent magnet group 10 are designed in an integrated manner, the vibration displacement can be directly transmitted to the permanent magnet group 10 through the loading rod 2, which is used to provide top support for the magnetic circuit centering system and as an external mounting interface. The base 9 refers to the bearing structure located below the reset mechanism, which can specifically be a metal block with mounting holes, used to fix the reset mechanism and bear the vibration transmission load. The support rod 6 refers to the axial rigid connecting piece connecting the top partition plate 3 and the bottom partition plate 7, which can specifically be a threaded steel rod or a hollow metal pipe, which is threaded through the partition plate and threaded connected with the top cover 1 and the base 9, forming an axial constraint frame.

[0042] Specifically, the top cover 1 and the base 9 form an upper and lower symmetrical support frame through the support rod 6, the top partition plate 3 and the bottom partition plate 7 are rigidly connected after being penetrated by the support rod 6, limiting the lateral deviation of the permanent magnet group 10 in the coil 5. The axial pre-tightening is realized by three-stage locking assembly: the upper end uses lock bolt and hex nut 4 to complete the pre-pressing of the top cover 1; the middle section fixes the coil 5 skeleton through the middle reinforcing nut 12; the lower end configures the lower fixing nut 13 to form the foundation anchoring point, the support threaded rod generates uniform axial tension under the action of three-stage pre-tightening force, ensuring the magnetic gap balance and the assembly coaxiality under high frequency vibration condition, and eliminating the risk of coil 5 sweeping bore. The support rod 6 is detachably threaded connected with the top cover 1 and the base 9, which is convenient for adjusting the distance between the partition plates during assembly. The top cover 1 is directly fixed with the controlled structure, and the base 9 is contacted with the reset mechanism through the spring pad 11, the vibration load is dispersedly transmitted to the top cover 1 and the base 9 through the support rod 6, avoiding stress concentration. The structure integrates the magnetic circuit centering system and the reset mechanism into a closed frame, without external sensor positioning, directly constraining the movement track of the permanent magnet group 10 through mechanical connection.

[0043] The reset mechanism is a compression spring 8, specifically a cylindrical helical compression spring 8, the lower end of which is supported on the base 9, and the upper end is in contact with the bottom of the permanent magnet group 10 through the spring pad 11 to form a support. Among them, the compression spring 8 refers to a helical spring that stores mechanical energy by elastic deformation, which can be realized by winding a cylindrical cross-section steel wire, and generates a linear restoring force proportional to the displacement amount by axial compression. The base 9 refers to the basic component that carries the reset mechanism, which can be made of metal or high-strength composite material, and provides a stable support reference point for the compression spring 8. The spring pad 11 refers to a rigid contact plate arranged between the compression spring 8 and the permanent magnet group 10, which can be a disc-shaped structure with a smooth surface, used to disperse the concentrated stress of the spring reaction force and maintain the axial movement trajectory of the permanent magnet group 10. The cylindrical helical compression spring 8 can accurately reset the permanent magnet group 10 to the center line of the magnetic circuit after the vibration excitation disappears, maintaining the symmetry of the magnetic field and the initial working state.

[0044] The lower end of the compression spring 8 is fixed to the base 9, and the upper end is in contact with the bottom of the permanent magnet group 10 through the spring pad 11. When the permanent magnet group 10 is moved downward under the action of the vibration load, the spring pad 11 transmits the displacement to the compression spring 8, causing it to compress and store elastic potential energy; when the vibration load disappears, the compression spring 8 releases the potential energy to push the spring pad 11 upward, driving the permanent magnet group 10 to return to the initial position. Because the base 9 and the spring pad 11 form a two-way constraint, the deformation direction of the compression spring 8 is limited to the axial direction, avoiding lateral component forces that cause the permanent magnet group 10 to deviate from the movement trajectory. The planar contact design of the spring pad 11 further disperses the local stress, preventing plastic deformation of the bottom of the permanent magnet group 10 due to point contact, and ensuring the uniformity of force transmission during the reset process.

[0045] The permanent magnet group 10 adopts a ten-level alternating magnetic pole axial arrangement configuration, specifically the "N-S-S-N-N-S-S-N-N-S" magnetic pole arrangement, and there is a space of 2-4mm between the permanent magnet group 10 and the coil 5. The space is a uniform magnetic gap, and the coil 5 is a multi-turn enameled copper coil winding. The permanent magnet group 10 and the coil 5 form a dynamic and static magnetic coupling system. Among them, the ten-level alternating magnetic pole axial arrangement configuration refers to ten permanent magnets with alternating polarity arranged continuously along the movement direction, which can be realized by axial magnetization of neodymium iron boron material, and the N and S poles of adjacent permanent magnets are arranged in turn to form a gradient magnetic field. The magnetic field gradient generated by the alternating polarity enhances the magnetic flux change rate, thereby improving the linearity of the induced voltage. The 2-4mm space refers to the radial gap between the outer surface of the permanent magnet and the inner wall of the coil 5, which can be realized by a positioning ring made of non-magnetic material to control the gap. The gap range is determined by comprehensive evaluation of magnetic field penetration depth and mechanical interference risk, which ensures the closure efficiency of the magnetic circuit while avoiding contact between moving parts.

[0046] When the permanent magnet group 10 moves axially under the action of a vibration load, the alternating pole arrangement causes the magnetic flux passing through the coil 5 to exhibit periodic changes, and the polarity reversal of adjacent poles causes the rate of change of the magnetic flux to increase significantly, thereby generating an induced voltage with a higher amplitude and a linear relationship with the speed across the two ends of the coil 5. This induced voltage is generated in accordance with Faraday's Law, and a closed-loop energy dissipation current is formed by connecting an adjustable resistor, and the reverse electromagnetic coupling force generated by the current acts on the loading rod 2, forming a vibration suppression effect. The gap between the coil 5 and the permanent magnet is controlled within the range of 2-4 mm, so that the magnetic field can fully penetrate the coil 5 winding to maintain the electromagnetic induction efficiency, while avoiding the friction loss between the permanent magnet and the coil 5 due to the too small gap. The gap range can balance the contradiction between the magnetic field strength attenuation and the structural compactness through experimental verification.

[0047] The entire module is integrated into a standardized module through the closed-loop frame from the top cover 1 to the base 9 and the three-stage pre-tightening system, and can be directly embedded into a vehicle suspension system, a bridge seismic isolation support or an industrial equipment base. In combination with a digital adjustable resistance module, real-time optimization and energy recovery of damping parameters are realized.

[0048] In the specific embodiment of the present application, a circuit damper control method is also provided, which can be used to control Figure 2 The controlled structure of the circuit damper control system shown is simplified as a single-degree-of-freedom structure model, and the vibration is transmitted to the circuit damper provided in the foregoing embodiment through the loading rod 2, to excite the coil current The coil current After executing the measurement module, the permanent magnet-coil coupling structure in the circuit damper induces a primary voltage in the vibration, and a secondary voltage is formed through a Hall element and an amplifier, and then the voltage data analog quantity is output after rectification and isolation by an AC voltage transmitter-sensing circuit.

[0049] The execution measurement module is connected with a microcontroller and an external adjustable resistance. In the specific embodiment, the microcontroller is an ATMega328P, and the external adjustable resistance is an 8-stage programmable external resistance network. Eight relays S1 to S8 correspond to discrete resistance values of 1Ω, 2Ω, 4Ω, 8Ω, 16Ω, 32Ω, 64Ω and 128Ω, respectively, and the total impedance of the loop can be changed according to the control command .

[0050] After the microcontroller collects the voltage data, it calls the internal Gradient Boosting Regression (GBR) model that has been burned in to real-time inversion of the relative speed, and outputs a high or low level according to the balance logic to drive the relay: when the output is high, the relay is attracted, and the total resistance of the loop increases; when the output is low, the relay is disconnected, and the total resistance of the loop decreases, thereby realizing bidirectional adjustment.

[0051] Thus, a closed-loop process of structural vibration, electromagnetic induction, self-sensing measurement, and digital damping adjustment is formed, completing the integration of damping control and self-sensing in semi-active control.

[0052] The circuit damper control method specifically includes: The magnetic coupling system generates an induced voltage, and the magnetic coupling system is connected with an execution measurement module to process the induced voltage into voltage data analog quantity; The microcontroller receives the voltage data analog quantity, and calls the Gradient Boosting Regression model to invert the relative speed of the permanent magnet group 10 and the coil 5; The microcontroller adjusts the resistance value of the external adjustable resistor according to the relative speed, suppresses the controlled structure vibration, and completes the circuit damper control.

[0053] The Gradient Boosting Regression model refers to a machine learning model established by combining multiple weak regression trees, which can specifically extract the first derivative and change direction symbol features after preprocessing the voltage data, construct a lag dimension input sequence for training, and be used for predicting the relative speed from the voltage signal. The execution measurement module refers to a hardware unit for converting the induced voltage into a standardized analog signal, which is used to eliminate signal noise and adapt to the input requirements of the microcontroller. The external adjustable resistor resistance value adjustment refers to a control behavior of dynamically changing the resistance value according to the predicted speed, which is used to adjust the loop damping characteristics to match the vibration state.

[0054] Specifically, the magnetic coupling system generates an induced voltage when the controlled structure is vibrating, and the voltage is filtered and quantized by the execution measurement module to form a voltage data analog quantity. After the microcontroller receives the analog quantity, it analyzes the voltage variation characteristics through the Gradient Boosting Regression model, learns the nonlinear relationship between voltage and speed from historical data, and directly outputs the relative speed prediction value. Based on the prediction value, the microcontroller generates a resistance adjustment instruction to make the resistance value of the external adjustable resistor match the current vibration energy, forming a closed-loop control circuit without external sensors.

[0055] The traditional scheme relies on external sensors to collect displacement or acceleration signals, which requires additional deployment of sensor hardware and introduces numerical derivative calculation errors. However, the present scheme directly uses the induced voltage of the magnetic coupling system as the original signal source, extracts speed information from the voltage characteristics through a machine learning model, avoiding the installation requirements of external sensors and the error accumulation in the signal processing link.

[0056] By the technical solution, the vibration parameter sensing and control are completed only through the electromagnetic signal of the damper body, the system hardware cost and volume are reduced, the speed estimation accuracy is improved through the machine learning model, the resistance adjustment response is more in line with the actual vibration state, and the stability of the vibration suppression is enhanced.

[0057] In yet another specific embodiment of the present application, the deployment of the gradient boosting regression model in the microcontroller refers to Figure 4 The training and deployment process of the gradient boosting regression model starts with input sample data, including time stamp, coil induced voltage, displacement and resistance parameters; then the displacement data is converted into relative speed through derivative calculation to form a time series data set. In the feature engineering stage, a feature vector containing current voltage, current, voltage change direction and historical voltage and derivative information is constructed to establish an initial model. The model training uses gradient boosting regression algorithm, which iteratively updates through key parameters (learning rate , weak regression tree depth , training subsample ratio ), where the weak regression tree is specifically a CART tree. In the performance evaluation stage, RMSE and R 2 are used to quantify the model accuracy and goodness of fit, and the closer R 2 is to 1, the better the fitting effect is. After training, the model is pruned and optimized to improve the generalization ability and calculation efficiency, and finally burned into the microcontroller to realize embedded deployment, which can complete real-time speed inversion within 20ms cycle, meet the demand of 50Hz semi-active vibration control, and complete the whole process closed loop from data preprocessing to real-time prediction.

[0058] Specifically, the gradient boosting regression model is obtained by: collecting induced voltage and corresponding relative speed data of the magnetic coupling system under different vibration states to form a training data set; preprocessing the induced voltage data to reduce quantization noise, which specifically includes sampling the voltage signal at a frequency of 50Hz, and reducing quantization noise interference by taking average through four fast ADC conversions; extracting the first derivative and change direction symbol feature of the preprocessed voltage data and constructing a lag dimension input sequence, where the first derivative is approximately expressed as using the backward difference formula, and the change direction symbol feature can capture the magnetic hysteresis phenomenon commonly existing in electromagnetic damping; based on the input sequence and the corresponding relative speed data, the gradient boosting regression algorithm is used to train the model to obtain an initial model, taking the squared error as the loss function; the initial model is compressed to obtain a gradient boosting regression model suitable for microcontroller deployment, and the compression processing is specifically realized by using pruning algorithm combined with 16bit quantization compression, and the speed inversion accuracy of the final model is RMSE≈0.035ms -1 .

[0059] The data preprocessing refers to the operation of noise suppression on the original voltage signal, which is used to eliminate the quantization error generated in the analog-to-digital conversion process, and is specifically realized by 50Hz sampling frequency cooperating with four times of ADC conversion to average. The first derivative extraction refers to calculating the slope of the voltage signal changing with time, which is specifically realized by the backward difference formula The change direction symbol feature refers to the binary identification of the positive and negative states of the voltage derivative, which is used to enhance the recognition ability of the model to the change of the speed direction, and can capture the magnetic hysteresis phenomenon commonly existing in the electromagnetic damping. The lag dimension input sequence refers to a multi-dimensional feature vector containing the voltage values at historical time and their derivatives, which is used to improve the memory modeling ability of the model to the dynamic process, and fully reflects the path dependence of the system. The model compression processing refers to the operation of reducing the number and depth of regression trees, which can be specifically realized by pruning algorithm combined with 16bit quantization compression, and is used to adapt to the storage capacity and computing power limitations of the microcontroller, and to ensure that the model can complete real-time speed inversion within 20ms period.

[0060] Specifically, the induced voltage signals under different frequency and amplitude excitations are collected by the vibration table test, and the relative motion speed of the permanent magnet group and the coil is recorded synchronously to form a training data set covering multiple working conditions. After eliminating high-frequency noise from the original voltage signal, it is sampled at 50Hz frequency and further denoised by four times of ADC conversion to average. The central difference method is used to calculate the voltage derivative and extract the symbol feature, which can capture the magnetic hysteresis phenomenon, construct a lag input sequence containing the voltage information at the current time and the previous The processed feature vector and the measured speed data are input into the gradient boosting regression framework, and multiple weak regression trees with limited depth are generated by iteration to form a strong prediction model. The initial model trained is removed by pruning algorithm, and is quantized and compressed to 16bit, and finally a simplified model that can run in real time in the microcontroller is generated, realizing the direct mapping from the voltage signal to the relative speed.

[0061] Compared with the prior art, the traditional scheme relies on external sensors to obtain vibration parameters, and needs to additionally configure signal conditioning circuit and communication module, while the present scheme directly uses the induced voltage signal generated by the damper itself to construct the speed inversion model, eliminating the dependence on external sensors. When the prior art uses numerical differentiation to estimate the speed, it is easily disturbed by signal noise, and the present scheme effectively suppresses the noise amplification effect by using gradient boosting regression model to nonlinearly model the voltage dynamic characteristics. In addition, the prior art does not consider the computing limitation of embedded devices, and the present scheme deploys the prediction model to the microcontroller by pruning algorithm combined with 16bit quantization compression, realizing the real-time requirement of closed-loop control.

[0062] By the technical solution, the application constructs an endogenous sensing mechanism based on the characteristics of the electromagnetic damper body, without external sensors to obtain the relative speed parameter in real time, reducing the system hardware cost and volume. The data-driven regression model is used to replace the traditional numerical differential method, improving the speed inversion accuracy and control stability. The model compression processing ensures efficient operation of the algorithm in the resource-limited microcontroller, forming a complete self-sensing closed-loop control system.

[0063] Considering that the actual response of the electromagnetic damper has a memory effect and non-stationarity, the historical voltage value, its derivative and direction information are added to the feature vector to construct an input sequence of lag dimension:

[0064] wherein is the current voltage value, is the first derivative of the voltage, is the change direction sign function, is the lag order.

[0065] wherein, the current voltage value is the induced voltage signal output by the magnetic coupling system in real time, which can be realized by sampling using a high-precision analog-to-digital converter, directly reflecting the relative motion state of the permanent magnet group and the coil. The first derivative of the voltage is the rate of change of the voltage signal with time, which can be realized by using a difference operation or a digital filter, and is used to capture the dynamic characteristics of the voltage signal. The change direction sign function is a binary representation of the positive and negative states of the voltage derivative, which can be realized by using a comparator circuit or a logic operation, and is used to eliminate the interference of noise on trend judgment. The lag order is the time window length of the historical data, which can be realized by using a circular buffer storage, and is used to establish a time series correlation model of the voltage signal.

[0066] Specifically, the construction of the multi-dimensional feature vector realizes the autonomous perception of vibration parameters by fusing real-time voltage values and historical state data. The current voltage value as a basic signal source, directly represents the instantaneous energy conversion state of the magnetic coupling system. The superimposed first derivative of the voltage can enhance the sensitivity to changes in the direction of motion, and the dynamic characteristics of the voltage signal are extracted by differential operation. The sign function converts the continuous derivative signal into a discrete direction identifier, effectively suppressing false judgments caused by noise. The introduction of the lag order integrates the voltage value and its derivative , forming a sliding window in time dimension, and capturing the implicit patterns of vibration velocity by constructing time series correlation model. This multi-dimensional feature construction method extends the single-dimensional voltage signal to a high-dimensional vector containing space-time features, providing sufficient information input for the subsequent gradient boosting regression model.

[0067] The gradient boosting regression algorithm combines multiple regression trees to form a strong prediction model, and its output expression is:

[0068] wherein, wherein is the number of trees, is the learning rate, is the th regression tree.

[0069] Gradient boosting regression algorithm refers to an ensemble learning method that combines multiple weak prediction models iteratively. It can be implemented using additive models and forward step-by-step algorithms, and the overall accuracy is improved by gradually correcting the prediction residuals of the previous model. Regression tree refers to a tree-shaped prediction model based on feature space partitioning, which can be implemented using a binary tree structure. Each non-leaf node contains feature selection and segmentation threshold judgment rules. Learning rate is a hyperparameter that controls the contribution of a single tree to the final prediction result. It can be implemented using a fixed value in the range of 0.1 to 0.3, which balances the model convergence speed and overfitting risk.

[0070] The voltage signal is preprocessed to form a multi-dimensional feature vector containing the current voltage value, first derivative, and historical data. Each regression tree generates prediction rules by recursively partitioning the feature space. The first tree is fitted to the original features, and each subsequent tree is trained on the prediction residuals of the previous model. The learning rate accumulates the prediction results of each tree in a weighted manner to form the final velocity prediction value. The low depth of the weak regression tree ensures that the decision path of a single tree does not exceed three feature judgments. The combined model maintains the ability to fit non-linearities while controlling the parameter size within the storage capacity of a microcontroller. During the model training phase, the gradient descent method is used to determine the optimal partition point for each tree, gradually reducing the mean square error between the predicted value and the true velocity.

[0071] The regression tree adopts a weak regression tree, and the depth of each tree is not more than 3. The weak regression tree refers to a regression tree model with weak prediction ability but low computational complexity, and can be specifically implemented by a shallow decision tree with small prediction error. The gradient boosting algorithm is used to combine multiple weak trees to form a strong prediction model. The depth of the tree is not more than 3, which limits the number of splits and the number of levels of the regression tree. Specifically, the maximum number of split layers of the tree can be set to achieve this. This parameter directly constrains the complexity of the tree structure. The weak regression tree reduces the prediction accuracy of a single tree to improve the computational efficiency, and the shallow tree structure reduces the number of branch nodes to reduce the number of model parameters. In the gradient boosting framework, the prediction results of multiple weak trees are weighted and superimposed to form a stable estimate of the relative speed. Each tree only performs conditional judgment not more than three times, so that the number of calculation steps required for a single inference is significantly reduced. During the model training process, the number of trees and the learning rate parameter are optimized and configured to ensure that the overall prediction error converges to an acceptable range. This design compresses the operation time and memory usage within the microcontroller resource limit while maintaining the inversion accuracy.

[0072] In this embodiment, the gradient boosting regression model uses the square error as the loss function, and the goal is to minimize the overall empirical risk. The square error loss function is as follows:

[0073] During the model training process, the root mean square error (RMSE) and the coefficient of determination R2 are used as evaluation indicators. As a model performance evaluation indicator, wherein The root mean square error is the square root of the arithmetic mean of the square of the deviation between the predicted value and the true value. Specifically, the square root of the mean of the error square sequence of the predicted speed and the measured speed can be used to achieve this. This indicator can amplify the weight of large error items, and can quantify the prediction deviation of the model under extreme vibration conditions, thereby guiding the optimization direction of the model parameters.

[0074] Specifically, in the training phase of the gradient boosting regression model, the combination weight and split node of the regression tree are dynamically adjusted by continuously monitoring the root mean square error of the predicted speed and the true speed. When the model has a prediction bias at a certain vibration frequency or amplitude, the sensitivity of the root mean square error can quickly identify the abnormal data interval, and gradually reduce the dispersion degree of the prediction error by iteratively optimizing the depth and branch condition of the regression tree. The error feedback mechanism established in this way can effectively suppress the speed inversion error caused by voltage signal noise or numerical derivative calculation lag, ensure that the relative speed signal output by the model is synchronized with the true motion state, provide accurate input for subsequent damping adjustment, meet the control requirements, and the smooth speed curve output by the model can be used as the key input of the balance logic damping control, forming a closed loop of circuit damper self-sensing, relative speed prediction and circuit damper control.

[0075] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A circuit damper, characterized by, include: The magnetic coupling system includes a permanent magnet assembly (10) and a coil (5). The permanent magnet assembly (10) is located inside the coil (5). The permanent magnet assembly (10) is connected to a loading rod (2) for connecting the controlled structure to load vibration displacement, thereby driving the permanent magnet assembly (10) to move within the coil (5) to generate an induced voltage. The magnetic circuit alignment system includes a top partition (3) and a bottom partition (7), which are respectively set at the top and bottom of the coil (5) to limit the travel range of the permanent magnet assembly (10); The reset mechanism is connected to the permanent magnet assembly (10) and resets it; The coil (5) is connected to a microcontroller and an external adjustable resistor. The microcontroller adjusts the external adjustable resistor according to the induced voltage of the magnetic coupling system.

2. A circuit damper according to claim 1, characterized in that It also includes a support structure, which includes a top cover (1) set above the top partition (3) and a base (9) set below the reset mechanism; the top partition (3) and the bottom partition (7) are connected together by a support rod (6), one end of the support rod (6) passes through the top partition (3) and is connected to the top cover (1), and the other end passes through the bottom partition (7) and is connected to the base (9).

3. A circuit damper according to claim 2, wherein The reset mechanism is a compression spring (8). The lower end of the compression spring (8) is supported on the base (9), and the upper end is supported by the bottom of the permanent magnet assembly (10) through the spring pad (11).

4. A circuit damper according to claim 1, wherein The permanent magnet assembly (10) adopts a configuration of axial arrangement of ten alternating magnetic poles, and a space of 2-4 mm is left between the permanent magnet assembly (10) and the coil (5).

5. A method of controlling the circuit damper according to any one of claims 1 to 4, characterized by, include: The induced voltage generated by the magnetic coupling system is processed into analog voltage data. The relative velocity between the permanent magnet assembly (10) and the coil (5) is inverted based on the voltage data analog quantity by calling the gradient boosting regression model; The resistance value of the external adjustable resistor is adjusted according to the relative velocity to suppress the vibration of the controlled structure and complete the circuit damper control.

6. The method of claim 5, wherein the step of determining the damping factor comprises the step of: The gradient boosting regression model is obtained through the following method: The induced voltage and corresponding relative velocity data of the magnetic coupling system under different vibration states are collected to form a training dataset; The induced voltage data is preprocessed to reduce quantization noise; Extract the first derivative and sign characteristics of the change direction of the preprocessed voltage data, and construct the input sequence with hysteresis dimension; Based on the input sequence and the corresponding relative velocity data, the model is trained using the gradient boosting regression algorithm with the squared error as the loss function to obtain the initial model. The initial model is compressed to obtain a gradient boosting regression model suitable for microcontroller deployment.

7. The method of claim 6, wherein the step of determining the damping factor comprises the step of: determining the damping factor based on the difference between the first and second voltage values. The specific method for constructing the lagged dimension input sequence is as follows: The feature vector is constructed in the form where is the voltage value at the current moment, is the first derivative of the voltage, is the sign function of the change direction, is the hysteresis order.

8. The method of claim 6, wherein the step of determining the damping factor comprises the step of: determining the damping factor based on the difference between the first and second voltage values. The gradient boosting regression algorithm constructs a strong predictive model by combining multiple regression trees, and its output expression is: wherein, wherein is the number of trees, is the learning rate, is the th regression tree.

9. The method of claim 8, wherein the step of determining the damping factor comprises the step of: determining the damping factor based on the difference between the first and second voltage values. The regression tree is a weak regression tree, and the depth of each weak regression tree does not exceed 3.

10. The method of claim 6, wherein the step of determining the damping factor comprises the step of: determining the damping factor based on the frequency of the input signal. The root mean square error and coefficient of determination are used as performance evaluation metrics during model training.

Images