A method, device and equipment for stiffness compensation of a magnetic bearing rotor

Abstract

The application relates to the technical field of magnetic bearing control, and discloses a rigidity compensation method, device and equipment for a magnetic bearing rotor, wherein the method comprises the following steps: acquiring real-time operation data and target rigidity data of the magnetic bearing rotor, wherein the real-time operation data at least comprises a current bias current; performing gated correction reconstruction on the real-time operation data according to a pre-trained rigidity identification model to obtain target identification data, wherein the rigidity identification model is constructed based on a prior rigidity model; constructing a rigidity error vector and a Jacobian matrix according to the target rigidity data and the target identification data; performing weighted cost matching on the rigidity error vector and the Jacobian matrix to obtain a bias current increment; and determining a target bias current according to the bias current increment and the current bias current, wherein the target bias current is used for performing outer ring supervision control on the magnetic bearing rotor to complete rigidity compensation. The technical scheme provided by the application can realize high-precision identification and compensation of dynamic rigidity of the magnetic bearing rotor.

Description

Technical Field

[0001] This application relates to the field of magnetic bearing control technology, and in particular to a method, device and equipment for stiffness compensation of a magnetic bearing rotor. Background Technology

[0002] As a key component in ultra-high vacuum systems, magnetic levitation turbomolecular pumps experience significant degradation in current and displacement stiffness of their internal magnetic bearing rotors under temperature-speed coupling conditions where high temperature and high speed coexist, thus affecting the operational reliability of the pump.

[0003] In related technologies, control parameters are typically adjusted to adapt to changes in the dynamic characteristics of magnetic bearing rotors. However, when faced with dynamic stiffness variations coupled with temperature and speed, these methods often suffer from insufficient model representation, poor consistency across operating conditions, or overly conservative control, making it difficult to achieve stiffness compensation under coupled operating conditions. Therefore, achieving high-precision identification and online compensation of dynamic stiffness under temperature-speed coupled operating conditions has become a key research focus in the field of magnetic bearing control. Summary of the Invention

[0004] This application provides a method, apparatus, and equipment for stiffness compensation of a magnetic bearing rotor, which can achieve high-precision identification and compensation of the dynamic stiffness of the magnetic bearing rotor.

[0005] The first aspect of this application provides a stiffness compensation method for a magnetic bearing rotor. The method includes: acquiring real-time operating data and target stiffness data of the magnetic bearing rotor; wherein the real-time operating data includes at least the current bias current; performing gated correction and reconstruction on the real-time operating data according to a pre-trained stiffness identification model to obtain target identification data; wherein the stiffness identification model is constructed based on a prior stiffness model; constructing a stiffness error vector and a Jacobian matrix according to the target stiffness data and the target identification data, respectively, and performing weighted cost matching on the stiffness error vector and the Jacobian matrix to obtain the bias current increment; determining a target bias current according to the bias current increment and the current bias current, wherein the target bias current is used to perform outer-loop supervisory control on the magnetic bearing rotor to complete stiffness compensation.

[0006] In one implementation, the prior stiffness model is constructed based on physical genes; gating correction and reconstruction of real-time running data to obtain stiffness identification data includes: performing network correction estimation on real-time running data to obtain target correction data; gating mapping on the target correction data to obtain gating mapping factors; and determining target identification data based on the gating mapping factors and the physical genes.

[0007] In one implementation, the target identification data includes identification bias current and stiffness identification data; determining the target identification data based on the gating mapping factor and the physical gene includes: acquiring the model bias current, determining the identification bias current based on the gating mapping factor and the model bias current, and determining the stiffness identification data based on the identification bias current, the gating mapping factor, and the physical gene.

[0008] In one implementation, the prior stiffness model includes a current stiffness model and a displacement stiffness model; wherein, the current stiffness model is used to characterize the change of current stiffness with bias current and dynamic reluctance, and the displacement stiffness model is used to characterize the change of displacement stiffness with bias current and dynamic reluctance.

[0009] In one implementation, the current stiffness model and the displacement stiffness model are determined based on a bias current term and a dynamic magnetoresistive term. The bias current term characterizes the change of bias current with temperature, and the dynamic magnetoresistive term characterizes the change of dynamic magnetoresistive with rotational speed.

[0010] In one implementation, the stiffness identification model has a neural network; the stiffness identification model is trained as follows: multiple sets of measured data under different working conditions are acquired, including measured rotational speed, measured temperature and measured bias current; the parameters of the neural network are trained based on the measured data and the joint loss function to obtain the trained stiffness identification model.

[0011] In one implementation, the joint loss function includes a data error term and a physical constraint term; wherein the physical constraint term includes a correction magnitude constraint term, a smoothness constraint term, and a bias current consistency constraint term; wherein the constraint weights of the smoothness constraint term and the bias current consistency constraint term are determined by annealing scheduling.

[0012] In one implementation, before determining the target bias current based on the bias current increment and the current bias current, the method further includes: updating the bias current increment based on a rate of change constraint and upper and lower limit constraints to determine the target bias current based on the updated bias current increment.

[0013] A second aspect of this application provides a stiffness compensation device for a magnetic bearing rotor. The device includes: a data acquisition unit for acquiring real-time operating data and target stiffness data of the magnetic bearing rotor; wherein the real-time operating data includes at least the current bias current; a data identification unit for performing gated correction and reconstruction on the real-time operating data according to a pre-trained stiffness identification model to obtain target identification data; wherein the stiffness identification model is constructed based on a prior stiffness model; an increment determination unit for constructing a stiffness error vector and a Jacobian matrix respectively based on the target stiffness data and the target identification data, and performing weighted cost matching on the stiffness error vector and the Jacobian matrix to obtain the bias current increment; and a stiffness compensation unit for determining a target bias current based on the bias current increment and the current bias current, wherein the target bias current is used for outer-loop supervisory control of the magnetic bearing rotor to complete stiffness compensation.

[0014] A third aspect of this application provides a computer device, a memory, and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions. The computer device is used to implement the stiffness compensation of the magnetic bearing rotor described in the first aspect.

[0015] The technical solution provided in this application determines the target bias current by comprehensively considering dynamic reluctance changes and temperature rise current drift through a stiffness identification model, which is then applied to the magnetic bearing rotor for real-time and precise stiffness compensation. Specifically, the real-time operating data is gated and reconstructed using the stiffness identification model to obtain target identification data that ensures smooth and controllable correction. The stiffness identification model embeds a priori stiffness model to characterize the stiffness degradation mechanism of the magnetic bearing rotor under temperature-speed coupling conditions. Furthermore, a stiffness error vector and a Jacobian matrix characterizing the sensitivity of current stiffness and displacement stiffness to the bias current are constructed based on the target identification data. Weighted cost matching is then used to quickly determine the bias current increment. The target bias current is determined by combining the current bias current with the bias current increment, and this is input as an outer-loop supervisory control command to the inner-loop controller, completing online and precise stiffness compensation of the magnetic bearing rotor without altering the inner-loop structure.

[0016] As can be seen, the technical solution provided in this application achieves high-precision dynamic stiffness identification and compensation of magnetic bearing rotor under temperature-speed coupling conditions, effectively restores degraded current stiffness and displacement stiffness, and improves the stability and adaptability of magnetic bearing rotor system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A schematic diagram illustrating the steps of a stiffness compensation method for a magnetic bearing rotor provided in an embodiment of this application; Figure 2 A schematic diagram illustrating the steps of a stiffness identification data acquisition method provided in one embodiment of this application; Figure 3 A schematic diagram illustrating the steps of a stiffness identification model training method provided in one embodiment of this application; Figure 4 A schematic diagram of the structure of a stiffness compensation device for a magnetic bearing rotor provided in one embodiment of this application; Figure 5 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Furthermore, the use of terms such as "first," "second," etc., in this application is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of embodiments in this application, unless otherwise stated, "multiple" means two or more. Additionally, the use of "based on" or "according to" implies openness and inclusiveness, because processes, steps, calculations, or other actions "based on" or "according to" one or more of the stated conditions or values ​​may in practice be based on additional conditions or beyond the stated values.

[0021] With the continuous development of the semiconductor manufacturing industry, higher requirements are placed on stability and process consistency in ultra-high vacuum environments. As a core component of ultra-high vacuum systems, the operational reliability of magnetic levitation turbomolecular pumps directly affects the overall process quality. Typically, magnetic levitation turbomolecular pumps must simultaneously withstand the combined effects of high-temperature heating (such as anti-deposition heating) and high-speed rotation during operation, forming a typical temperature-speed coupling condition. These pumps usually incorporate a drive motor, turbine blades, and a magnetic bearing rotor. Under temperature-speed coupling conditions, the current stiffness and displacement stiffness of the magnetic bearing rotor undergo significant degradation, leading to a decrease in stability margin and a weakening of vibration suppression capability, thus affecting the operational reliability of the pump. Here, the current stiffness represents the change in electromagnetic force when the control current changes by a unit amount. The displacement stiffness represents the change in electromagnetic force when the rotor deviates from its equilibrium position by a unit displacement.

[0022] The physical mechanisms of dynamic stiffness degradation in magnetic bearing rotors mainly include two aspects: First, bias current drift caused by temperature rise. Specifically, coil resistance increases with temperature, and under constant reference voltage conditions, bias current decreases, leading to a reduction in current stiffness and displacement stiffness. Second, rotor eddy current effect induced by high-speed rotation. Specifically, laminated rotors generate eddy currents at high speeds, and their reverse magnetic field weakens the effective magnetic flux, further reducing the stiffness of the magnetic bearing. More complexly, the temperature effect and eddy current effect are not independent. Eddy currents cause additional temperature rise in the rotor, and the temperature rise changes the electromagnetic properties of the material, thus forming a strongly coupled stiffness degradation path of temperature and speed. In related technologies, most methods adapt to changes in system dynamic characteristics by adjusting control law parameters. Some thermal compensation methods correct the bias current by detecting coil temperature. However, when facing dynamic stiffness changes coupled with temperature and speed, these methods often suffer from insufficient model representation, poor consistency across operating conditions, or strong control conservatism, making it difficult to achieve stiffness compensation for coupled operating conditions.

[0023] In view of the above, one or more embodiments of this application provide a method, apparatus and device for stiffness compensation of magnetic bearing rotor, which can solve the above problems. By establishing a stiffness identification model that can explicitly characterize the temperature-speed coupled stiffness degradation mechanism, high-precision identification and online compensation of dynamic stiffness can be achieved.

[0024] Please see Figure 1 One embodiment of this application provides a method for stiffness compensation of a magnetic bearing rotor, which may include the following steps: S1: Obtain real-time operating data and target stiffness data of the magnetic bearing rotor; wherein the real-time operating data includes at least the current bias current.

[0025] The aforementioned real-time operating data may include the current rotational speed, current temperature, and current bias current, reflecting the current temperature-speed coupling condition. During actual operation, this real-time operating data is acquired in real time via sensors or a controller. The aforementioned target stiffness data includes target current stiffness and target displacement stiffness, representing the preset current stiffness and displacement stiffness of the magnetic bearing under the current operating conditions, respectively. The bias current, as a controllable variable of the magnetic bearing rotor, can be used to adjust both the current stiffness and displacement stiffness.

[0026] Optionally, the target stiffness data can be read from the controller's internal registers or an external configuration file, representing a preset target stiffness value for the current operating condition. Optionally, the value can be directly given by the engineer according to the rotor's load-bearing capacity and stability requirements. Optionally, the corresponding stiffness data can also be retrieved from a pre-stored lookup table based on the current operating condition; that is, a lookup table can be pre-built for target stiffness data under different temperature and speed conditions.

[0027] S3: Based on the pre-trained stiffness identification model, the real-time running data is gated, corrected, and reconstructed to obtain target identification data; wherein, the stiffness identification model is constructed based on a prior stiffness model.

[0028] The aforementioned a priori stiffness model can be understood as a standard stiffness model constructed based on physical laws such as electromagnetism, magnetic circuit principles, and thermodynamics (resistance temperature drift law), used to characterize the stiffness degradation mechanism of a magnetic bearing rotor under temperature-speed coupling conditions. For example, the stiffness degradation mechanism includes two aspects: firstly, as the rotational speed increases, eddy currents are generated within the rotor laminations, leading to an increase in equivalent dynamic reluctance and a decrease in effective magnetic flux, thus reducing stiffness; secondly, as the temperature increases, the coil resistance increases, causing a decrease in bias current under constant reference voltage conditions, which in turn leads to a decrease in stiffness.

[0029] In this embodiment, a neural network is introduced based on the prior stiffness model to construct a stiffness identification model. Specifically, the stiffness identification model can be understood as a physical-network hybrid model with embedded neural networks and optimized parameters through offline training. Given real-time operating data, the neural network outputs a correction amount for the stiffness, and the target identification data is determined based on the correction amount and the prior stiffness model. This correction amount reflects the degradation effect of the current temperature-speed coupling environment on the stiffness data. The target identification data includes identified bias current, identified current stiffness, and identified displacement stiffness. Optionally, the neural network is a physically constrained neural network, that is, physical prior knowledge is embedded in the neural network through a corresponding training loss function to ensure that the identified bias current and the identified stiffness satisfy an inherent physical relationship.

[0030] In this embodiment, the gating correction and reconstruction of real-time operating data can be understood as performing three levels of operations—gating, correction, and reconstruction—on the real-time operating data to obtain target identification data. Specifically, the gating operation can be understood as controlling the magnitude of the correction amount to prevent excessive correction from causing abrupt changes in output, ensuring the smoothness and stability of the correction process. The correction operation can be understood as outputting a baseline correction amount under the current operating condition through a neural network, and using this baseline correction amount as the stiffness correction amount after gating. The reconstruction operation can be understood as fusing the output of the prior stiffness model with the correction amount to reconstruct the final identification result, including the identification of bias current, current stiffness, and displacement stiffness.

[0031] In this embodiment, the aforementioned gated correction and reconstruction can still achieve accurate stiffness identification results even under sparse and noisy sample conditions. Since stiffness degradation under temperature-speed coupling conditions is continuous and slow, drastic stiffness jumps should be avoided. The gated mechanism ensures a smooth and controllable correction process, thereby guaranteeing the system stability of the magnetic bearing rotor. Learning the residual (correction amount) between the true stiffness and the target stiffness via a neural network, rather than directly outputting the stiffness, significantly reduces training difficulty and dependence on training data. Furthermore, introducing a prior stiffness model during the correction process allows the model to reflect the basic influence trend of temperature-speed coupling on stiffness even without initial stiffness data as a correction benchmark, thus directly outputting the required target stiffness data and maintaining reasonable output even under speed-temperature combinations outside the training conditions.

[0032] S5: Based on the target stiffness data and the target identification data, construct a stiffness error vector and a Jacobian matrix respectively, and perform weighted cost matching on the stiffness error vector and the Jacobian matrix to obtain the bias current increment.

[0033] The aforementioned stiffness error vector is used to reflect the ideal difference between the current stiffness and displacement stiffness of the current magnetic bearing. It is represented by a two-dimensional column vector consisting of the difference between the identified current stiffness and the target current stiffness, and the difference between the identified displacement stiffness and the target displacement stiffness.

[0034] For example, the above stiffness error vector Represented as: ,in, Indicates identification of current stiffness, Indicates the target current stiffness. Indicates the identification of displacement stiffness. This indicates the target displacement stiffness.

[0035] The Jacobian matrix described above is used to describe the sensitivity of the identification current stiffness and the identification displacement stiffness to the bias current, respectively. Specifically, based on the partial derivatives of the identification current stiffness and the identification displacement stiffness with respect to the bias current, the Jacobian matrix is ​​constructed as follows: , This is indicated to identify the bias current. Indicates identification of current stiffness, Indicates the identification of displacement stiffness. Indicates the current moment.

[0036] In this embodiment, the Jacobian matrix may include physical gene elements of stiffness data and gated correction factors. The physical gene elements are the rotational speed-related magnetic circuit parameters in the prior stiffness model, used to embed the mechanism of rotational speed eddy currents into the Jacobian matrix. This allows the Jacobian matrix to automatically adapt to differences in reference sensitivity at different rotational speeds, ensuring the working condition adaptability of the compensation strategy. The gated correction factor represents the bounded correction amount of the neural network output. Embedding the data-driven residual correction and its change sensitivity of the neural network into the Jacobian matrix can further improve the dynamic sensitivity of the Jacobian matrix for data-driven correction.

[0037] For example, the Jacobian matrix described above can be expressed as: ,in , These represent the physical genes for current stiffness and displacement stiffness, respectively. These represent the gating correction factors for current stiffness and displacement stiffness, respectively.

[0038] For example, the Jacobian matrix described above can be approximated as: Alternatively, the current stiffness can also be expressed as: Displacement stiffness can be identified as .

[0039] In this embodiment, the aforementioned weighted cost matching can be understood as constructing a weighted cost function and solving it based on the stiffness error vector and the Jacobian matrix to obtain the bias current increment. Specifically, the weighted cost function includes a weighted sum of squares term and a regularization penalty term. The weighted sum of squares term is constructed based on the stiffness error vector and the Jacobian matrix, and the regularization penalty term represents the bias current increment, which is the value that needs to be increased or decreased based on the current bias current. Specifically, a positive bias current increment indicates that the bias current needs to be increased to improve stiffness, while a negative bias current increment indicates that the bias current needs to be decreased to avoid overcompensation. In engineering practice, the contribution weights of current stiffness and displacement stiffness to system stability are not the same. For example, displacement stiffness directly determines the restoring force after the rotor deviates from the equilibrium position and has a greater impact on the static stability margin, so it may be assigned a higher weight.

[0040] In one embodiment, the above weighted cost function can be expressed as: ,in, A weight matrix (diagonal matrix) is used. Indicates the incremental penalty coefficient. This is the stiffness error vector. Let be the Jacobian matrix. The closed-form solution for the bias current increment is obtained from the weighted cost function. ,in, .

[0041] S7: Determine the target bias current based on the bias current increment and the current bias current. The target bias current is used to perform outer-loop monitoring control on the magnetic bearing rotor to complete stiffness compensation.

[0042] In this embodiment, the target bias current is obtained by adding the current bias current to the bias current increment. Specifically, the bias current directly affects the operating point of the magnetic bearing; the larger the bias current, the stronger the magnetic field at the operating point, and the greater the current stiffness and displacement stiffness. Therefore, by adjusting the bias current to the target value, the degraded current stiffness and displacement stiffness are effectively restored, and the system regains support characteristics close to ideal.

[0043] In a magnetic bearing system, there is usually an inner loop controller responsible for basic suspension stability control (such as PID control). The outer loop supervisory control refers to sending the target bias current as the command value of the bias current loop to the inner loop controller without changing the structure of the inner loop controller. The inner loop controller calculates the control voltage based on the error between the command value and the measured value. After the control voltage is modulated by PWM, it drives the power amplifier to adjust the average voltage across the coil, so that the actual bias current tracks the command value, thereby achieving online stiffness compensation.

[0044] Based on the above ideas, the technical solution provided in this embodiment of the application determines the target bias current by comprehensively considering the dynamic reluctance change and temperature rise current drift of the stiffness identification model, so as to apply it to the magnetic bearing rotor for real-time and accurate stiffness compensation. Specifically, the real-time operating data is gated and reconstructed by the stiffness identification model to obtain target identification data that ensures smooth and controllable correction. The stiffness identification model embeds a prior stiffness model to characterize the dynamic reluctance change and temperature rise current drift of the magnetic bearing rotor under temperature and speed coupling conditions. Furthermore, a stiffness error vector and a Jacobian matrix for adjusting sensitivity differences are constructed based on the target identification data to quickly determine the bias current increment through weighted cost matching. The target bias current is determined by combining the current bias current and the bias current increment, and it is input as an outer loop supervisory control command into the inner loop controller to complete the online and accurate stiffness compensation of the magnetic bearing rotor without changing the inner loop structure. Therefore, this technical solution achieves high-precision identification and low-delay online compensation of dynamic stiffness under temperature-speed coupling conditions, effectively restoring degraded current stiffness and displacement stiffness, and improving the stability and adaptability of the magnetic bearing rotor system.

[0045] In one embodiment, the aforementioned prior stiffness model includes a current stiffness model and a displacement stiffness model; wherein, the current stiffness model is used to characterize the change of current stiffness with bias current and dynamic reluctance, and the displacement stiffness model is used to characterize the change of displacement stiffness with bias current and dynamic reluctance.

[0046] Specifically, as the rotational speed increases, the eddy current effect in the rotor laminations intensifies, the equivalent dynamic reluctance increases, and the effective magnetic flux in the magnetic circuit weakens. This results in a smaller change in the electromagnetic force generated by the same change in control current, i.e., a decrease in current stiffness. In actual operation, the bias current drifts with temperature, and the dynamic reluctance changes with rotational speed. The dynamic change law of current stiffness can be described by the current stiffness model, providing a basis for online compensation.

[0047] Furthermore, dynamic reluctance also affects displacement stiffness, and the weakening effect of increasing rotational speed on displacement stiffness is more severe than that of current stiffness. A displacement stiffness model can describe the dynamic changes in displacement stiffness, providing a basis for online compensation. Therefore, this study describes the stiffness degradation mechanism under temperature-speed coupling conditions based on bias current and dynamic reluctance, and analyzes the contributions of temperature and rotational speed effects to stiffness data using both current and displacement stiffness models.

[0048] In this embodiment, the current stiffness model and the displacement stiffness model are determined based on the bias current term and the dynamic reluctance term. The bias current term is used to characterize the change of bias current with temperature, and the dynamic reluctance term is used to characterize the change of dynamic reluctance with rotational speed.

[0049] In one embodiment, the aforementioned prior stiffness model further includes a bias current model, wherein the bias current is determined by the bias current model, so that the aforementioned current stiffness model and displacement stiffness model are constructed based on the bias current model. The aforementioned bias current model characterizes the change of bias current with temperature. Specifically, an increase in temperature leads to an increase in coil resistance, and under a constant reference voltage, a decrease in bias current affects the stiffness data.

[0050] In this embodiment, the above bias current model can be expressed as: ,in, Indicates the current bias current. This indicates the temperature coefficient of resistance of copper wire. Indicates the current temperature. This represents the initial reference temperature, typically room temperature or the calibration temperature. Understandably, the above bias current model characterizes the decrease in bias current caused by the increase in coil resistance as temperature rises.

[0051] In one embodiment, the above current stiffness model can be expressed as: ,in, This represents the electromagnetic force generated by the magnetic bearing in the x-direction, where N represents the number of coil turns. Represents the dynamic reluctance term. This indicates the bias current term. , Indicates the geometric parameters of the magnetic circuit (e.g., the axial length or effective magnetic circuit length of the magnetic bearing, the pole arc length or pole pitch of the magnetic bearing). This represents the permeability of free space. Understandably, Indicates the rotational speed The current stiffness is simultaneously controlled by the bias current (affected by temperature) and the dynamic reluctance (affected by rotational speed).

[0052] In this embodiment, the dynamic magnetoresistive force described above is represented as: ,in, Indicates stator reluctance. Indicates air gap reluctance. Indicates the dynamic magnetic reluctance of the rotor laminations. This represents the magnetic circuit correction coefficient related to the current stiffness. The aforementioned rotor lamination dynamic reluctance includes an eddy current term related to the rotational speed; that is, the eddy current term is characterized based on the rotational speed and the eddy current effect coefficient.

[0053] In one embodiment, the above displacement stiffness model can be expressed as: ,in, These are all comprehensive magnetoresistive parameters that include dynamic magnetoresistive parameters. Specifically, , , , ,in, This represents the geometric parameters related to the air gap or magnetic circuit gap.

[0054] The technical solution provided in this embodiment refines the composition and construction method of the prior stiffness model, thereby incorporating the stiffness degradation mechanism caused by the temperature-speed coupling effect into the prior stiffness model. Specifically, the decrease in bias current due to temperature rise is described by a bias current model, and the influence of rotational speed on the magnetic circuit is described by a dynamic reluctance expression including eddy current terms. Based on the bias current and dynamic reluctance, current stiffness and displacement stiffness models are constructed respectively, so that both current stiffness and displacement stiffness are simultaneously controlled by the temperature effect (characterized by the bias current term) and the rotational speed effect (characterized by the dynamic reluctance term). This technical solution provides a priori reference with a physical mechanism for subsequent identification, correction, and online compensation by mathematically representing the stiffness degradation mechanism under temperature-speed coupling conditions, thereby achieving accurate stiffness compensation under temperature-speed coupling conditions.

[0055] In one implementation, please refer to Figure 2 The aforementioned prior stiffness model is constructed based on physical gene elements, which reflect the influence of dynamic reluctance changes caused by rotational speed on stiffness. Based on these physical gene elements, real-time operating data is gated, corrected, and reconstructed to obtain stiffness identification data. This process includes the following steps: S31: Perform network correction estimation on the real-time running data to obtain target correction data; S33: Perform gated mapping on the target correction data to obtain the gated mapping factor; S35: Determine the target identification data based on the gating mapping factor and the physical gene.

[0056] In step S31 above, the network correction estimation can be understood as inputting real-time running data into a pre-trained neural network, and the neural network outputs target correction data after forward computation, i.e., the correction operation. The target correction data includes current stiffness correction, displacement stiffness correction, and bias current correction.

[0057] In one embodiment, the neural network described above employs a multi-layer fully connected feedforward network (e.g., a lightweight MLP) to obtain target correction data in the following manner: The above These are, in order, the current stiffness correction, the displacement stiffness correction, and the bias current correction. For example, the input layer of the aforementioned neural network uses two nodes, each receiving the current rotational speed. and current bias current The hidden layer consists of multiple fully connected layers with tanh activation function, and the output layer uses three nodes to output... .

[0058] In this embodiment, network parameters The network includes weight matrices and bias vectors between each layer, and is trained using backpropagation and gradient descent optimization. The training objective is to minimize the joint loss function, thus simultaneously satisfying data fitting accuracy and physical constraints. After training, the network parameters are fixed and no longer updated during the online phase. Instead, the network rapidly infers and outputs corrections based on the current input conditions, ensuring the real-time performance of the online application.

[0059] In step S33 above, the gating mapping can be understood as converting unconstrained target correction data into bounded gating mapping factors, i.e., gating operation. These gating mapping factors include a current mapping factor, a current stiffness mapping factor, and a displacement stiffness mapping factor, which are used to correct bias current, correct current stiffness, and correct displacement stiffness, respectively.

[0060] In one embodiment, the above-mentioned gating mapping factor is determined as follows: current mapping factor Current stiffness mapping factor Displacement stiffness mapping factor ,in, These are hyperparameters that control the correction boundary. Since the temperature drift law of the bias current is relatively clear, the limitation on the correction range of the bias current should be more stringent, specifically manifested in... ,as well as Understandably, the hyperbolic tangent function The exponential function is used to compress the target correction data to the interval (-1, 1). Mapping the compressed values ​​to the positive number range preserves the direction information of the correction while strictly limiting the magnitude to a controllable range.

[0061] Based on step S35 above, the target identification data includes identified bias current and stiffness identification data. Determining the target identification data according to the gating mapping factor and physical gene includes: obtaining the model bias current; determining the identified bias current based on the model bias current and the gating mapping factor; and determining the stiffness identification data based on the identified bias current, the gating mapping factor, and the physical gene. The gating mapping factor includes the current mapping factor, the current stiffness mapping factor, and the displacement stiffness mapping factor; the physical gene includes the current stiffness factor and the displacement stiffness factor; and the stiffness identification data includes identified current stiffness and identified displacement stiffness.

[0062] In one embodiment, the aforementioned current factor is determined based on a current stiffness model, and the aforementioned stiffness factor is determined based on a displacement stiffness model. Specifically, the aforementioned current factor is expressed as: The above stiffness factor is expressed as: .

[0063] The aforementioned identification bias current is expressed as follows: Understandably, the above identification of bias current is based on model bias current. and current mapping factor Confirmed. Here, the bias current in the above model is the bias current output by the prior stiffness model (bias current model) based on real-time operating data, expressed as: .

[0064] The aforementioned identification current stiffness is expressed as follows: Understandably, the above-mentioned identification current stiffness is based on the identification bias current. Current stiffness mapping factor and current stiffness factor Sure, The aforementioned identification displacement stiffness is expressed as: Understandably, the above-mentioned displacement stiffness identification is based on the identification bias current. Displacement stiffness mapping factor and displacement stiffness factor Sure.

[0065] The technical solution provided in this embodiment abstracts the influence of rotational speed in the prior stiffness model into physical gene elements and performs gated correction and reconstruction on real-time operating data to obtain target identification data that integrates data-driven correction and physical priors. Specifically, firstly, physical gene elements reflecting the influence of dynamic magnetic reluctance at rotational speed are determined based on prior stiffness. Real-time operating data is input into a pre-trained multi-layer fully connected feedforward neural network, which outputs target correction data after forward computation. Subsequently, gated mapping is performed on the target correction data to convert it into bounded positive gated mapping factors. The boundary parameter of the bias current correction factor is smaller than that of the stiffness correction factor to reflect the relative certainty of the temperature drift law. Finally, the identification bias current, identification current stiffness, and identification displacement stiffness are determined based on the model bias current, physical gene elements, and current mapping factor. This technical solution ensures the physical rationality and accuracy of the output target identification data through the above-mentioned gated correction and reconstruction from operating condition data to target identification data, providing reliable input for subsequent online stiffness compensation.

[0066] In one implementation, the above stiffness identification model incorporates a neural network; please refer to [link / reference needed]. Figure 3 Specifically, the stiffness identification model is trained in the following manner: S61: Acquire measured data under multiple operating conditions, including measured rotation speed, measured temperature and measured bias current; S63: Based on the measured data and the joint loss function, the neural network is trained to obtain a trained stiffness identification model.

[0067] In this embodiment, the stiffness identification model is a physics-network hybrid model, which can be understood as embedding a neural network as a learnable component into a prior stiffness model that serves as a physical prior framework. The architecture of the stiffness identification model can be understood as follows: first, it receives real-time running data through an input layer; then, it outputs target correction data through a neural network; and finally, it performs gating mapping and reconstruction based on a built-in gating mapping module and a physical reconstruction module to obtain target identification data. The gating mapping module and the physical reconstruction module are represented by corresponding mathematical formulas.

[0068] During the training of the stiffness identification model, measured data under different operating conditions are acquired. This measured data typically includes measured rotational speed, measured temperature, and measured bias current, and may also include corresponding control current and rotor displacement, in order to deduce the true stiffness value as a supervisory signal. Since the quality of the measured data directly affects the performance of the trained model, preprocessing such as filtering, denoising, and normalization is usually required. The different operating conditions mentioned above represent operating conditions under different combinations of rotational speed and temperature. The network parameters in the neural network, such as weights and biases, are continuously adjusted through an optimization algorithm until the network parameters converge to a value that minimizes the loss function.

[0069] In this embodiment, the joint loss function includes a data error term and a physical constraint term; wherein the physical constraint term includes a correction magnitude constraint term, a smoothness constraint term, and a bias current consistency constraint term; and the constraint weights of the smoothness constraint term and the bias current consistency constraint term are determined through annealing scheduling. The optimization objective is to find the parameter combination that minimizes the total loss using gradient descent-type algorithms.

[0070] In one embodiment, the joint loss function described above can be expressed as: ,in, This represents the data error term, specifically as follows: , This represents the physical constraint term, specifically as follows: .

[0071] in, These represent the learnable parameters of a neural network, including the weights and biases of each fully connected layer. , Let N represent the identified current stiffness, measured current stiffness, identified displacement stiffness, and measured displacement stiffness of the nth training sample, respectively, where N represents the total number of training samples. This represents the numerical stability constant, used to construct the relative error between current stiffness and displacement stiffness. and These represent the weighting coefficients for the current stiffness data error and the displacement stiffness data error, respectively.

[0072] In this embodiment, the aforementioned physical constraint terms include correction magnitude constraint terms. Smoothness constraint term and bias current consistency constraints .in, , , .

[0073] in, This represents the identification bias current of the nth sample. This represents the model bias current at the nth sample temperature. , , These represent the current stiffness correction, displacement stiffness correction, and bias current correction for the nth sample, respectively. This represents the square of the correction after passing through the hyperbolic tangent function, since The range of is (-1, 1), and the range of its square is [0, 1). When the absolute value of the correction is very large... Approaching The square approaches 1; when the absolute value of the correction approaches zero, Approaching c, the square approximates .

[0074] In this embodiment, the constraint weights of the smoothness constraint and the bias current consistency constraint are determined as follows: , ,in, ,in, This indicates the current training iteration number. Indicates the total number of training rounds. Indicates the preheating ratio. Indicates the cosine climb ratio. and This represents the target weights for the smoothness constraint and the bias current consistency constraint.

[0075] In this embodiment, during the initial training phase (warm-up phase), the physical constraint weights are zero, allowing the network to freely fit the data and quickly reduce data errors. After the warm-up phase (cosine ramp-up phase), the weights smoothly increase from 0 to 1 according to half a cycle of the cosine function, gradually strengthening the physical constraints. This gradual introduction avoids the physical constraints prematurely limiting the network's expressive power in the early training phase, preventing the network from getting stuck in local optima or converging slowly. In the later training phase (stabilization phase), the weights reach the target value and remain constant, allowing the physical constraints to fully play their role in fine-tuning the physical rationality of the network output.

[0076] The technical solution provided in this embodiment uses a joint loss function to drive a neural network to fit measured stiffness through data error terms, ensuring that the neural network learns patterns from real data, which is the primary training objective. Specifically, physical constraint terms ensure that the learned patterns do not violate known physical laws. Three physical constraint terms—correction magnitude, smoothness, and bias current consistency—ensure the output conforms to engineering rationality. Furthermore, by using cosine-gradient weight scheduling to dynamically balance data accuracy and physical reliability at different training stages, the overfitting risk of purely data-driven methods and the insufficient accuracy of purely physical methods are avoided, ultimately resulting in an accurate and reliable stiffness identification model.

[0077] In one implementation, before determining the target bias current based on the bias current increment and the current bias current, the method further includes: updating the bias current increment according to a rate-of-change constraint and upper and lower bound constraints, so as to determine the target bias current based on the updated bias current increment. The dual safety constraints of the rate-of-change constraint and upper and lower bound constraints ensure a smooth and gradual compensation process, rather than abrupt changes, thereby transforming the solved bias current increment into an engineering-executable target bias current.

[0078] The aforementioned rate-of-change constraint can be understood as limiting the maximum allowable range of change in the bias current per unit time. This constraint limits the single-step adjustment range to prevent excessive jumps in the bias current, thus avoiding electromagnetic force step impacts on the rotor that could cause vibration or even instability. It also protects the coil from thermal stress damage and the power amplifier from tracking saturation, ensuring a smooth and gradual compensation process. The aforementioned upper and lower limit constraints can be understood as limiting the absolute range of the bias current itself. These constraints lock the bias current within a safe range, further preventing rotor instability, coil overheating, and magnetic saturation instability by limiting the cumulative total.

[0079] In one embodiment, the above rate of change constraint can be expressed as: ,in, The bias current increment is obtained by optimizing the solution using a weighted cost function through step S5 in the k-th control cycle. This represents the actual bias current increment after being clipped by the rate of change constraint. This indicates the maximum permissible single-step variation in bias current. This is a clipping function used to restrict the input to a specific range. Within the range.

[0080] In one embodiment, the above upper and lower limit constraints can be expressed as: ,in, It is the lower limit of the bias current. It is the upper limit of the bias current. Indicates the first The current bias current setpoint after the end of each control cycle. For the clipping function, the input is limited to... Within the range.

[0081] The technical solution provided in this embodiment ensures a smooth compensation process that remains within the physically feasible region by implementing rate-of-change constraints and upper / lower limit constraints. Specifically, the rate-of-change constraint limits the maximum allowable change in bias current per step, trimming the original increment obtained from the weighted cost matching solution. This prevents excessive single-step jumps in bias current that could cause electromagnetic force to abruptly impact the rotor, leading to vibration or even instability, thus ensuring a smooth and gradual compensation process. The upper / lower limit constraints lock the absolute value range of the bias current within a safe range, further limiting the cumulative total to prevent the rotor from becoming unstable and falling due to a weak magnetic field, overheating and burning out of the coil due to excessive current, or causing control failure due to magnetic circuit saturation. By applying rate-of-change constraints and upper / lower limit constraints to the increment of bias current, this technical solution comprehensively considers the theoretical compensation amount and engineering safety of the increment of bias current, ensuring both accurate stiffness compensation of the target bias current and the effectiveness and safety of the compensation.

[0082] Please see Figure 4 This application also provides a stiffness compensation device for a magnetic bearing rotor, the device comprising: The data acquisition unit 100 is used to acquire real-time operating data and target stiffness data of the magnetic bearing rotor; wherein, the real-time operating data includes at least the current bias current; The data identification unit 200 is used to perform gated correction and reconstruction on the real-time running data according to the pre-trained stiffness identification model to obtain target identification data; wherein, the stiffness identification model is constructed based on a prior stiffness model. The incremental determination unit 300 is used to construct a stiffness error vector and a Jacobian matrix respectively based on the target stiffness data and the target identification data, and to perform weighted cost matching on the stiffness error vector and the Jacobian matrix to obtain the bias current increment. The stiffness compensation unit 400 is used to determine a target bias current based on the bias current increment and the current bias current. The target bias current is used to perform outer-loop monitoring control on the magnetic bearing rotor to complete stiffness compensation.

[0083] in, In one embodiment, the data acquisition unit 100 is specifically used to read preset target stiffness data under the current operating condition from the internal register of the controller or the external configuration file, including target current stiffness and target displacement stiffness, and to collect real-time operating data in real time through sensors or the controller, wherein the real-time operating data includes at least the current bias current. In one embodiment, the data identification unit 200 is specifically used to perform three levels of operations—gating, correction, and reconstruction—on real-time running data according to a pre-trained stiffness identification model to obtain target identification data. The stiffness identification model is constructed based on a prior stiffness model and has a neural network. In one embodiment, the increment determination unit 300 is specifically used to construct a stiffness error vector based on target stiffness data and target identification data. The target identification data includes identified bias current, identified current stiffness and identified displacement stiffness. The unit also constructs a Jacobian matrix based on the target identification data and solves the weighted cost function based on the stiffness error vector and the Jacobian matrix to obtain the bias current increment. In one embodiment, the stiffness compensation unit 400 is specifically used to update the bias current increment according to the rate of change constraint and the upper and lower limit constraints, so as to determine the target bias current according to the updated bias current increment and the current bias current, and to perform outer loop supervision control on the magnetic bearing rotor according to the target bias current to complete stiffness compensation.

[0084] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0085] The stiffness compensation device for a magnetic bearing rotor in this application embodiment is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, or other devices that can provide the above functions.

[0086] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 5 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 5Take a processor 10 as an example.

[0087] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0088] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0089] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0090] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0091] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0092] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0093] The apparatus or unit described in the above embodiments can be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0094] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0095] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer devices. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer devices according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0100] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0101] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0102] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for stiffness compensation of a magnetic bearing rotor, characterized in that, The method includes: Acquire real-time operating data and target stiffness data of the magnetic bearing rotor; wherein the real-time operating data includes at least the current bias current; Based on a pre-trained stiffness identification model, the real-time running data is gated, corrected, and reconstructed to obtain target identification data; wherein, the stiffness identification model is constructed based on a prior stiffness model. Based on the target stiffness data and the target identification data, a stiffness error vector and a Jacobian matrix are constructed respectively. Weighted cost matching is performed on the stiffness error vector and the Jacobian matrix to obtain the bias current increment. Based on the bias current increment and the current bias current, a target bias current is determined. The target bias current is used to perform outer-loop monitoring control on the magnetic bearing rotor to complete stiffness compensation.

2. The method according to claim 1, characterized in that, The prior stiffness model is constructed based on physical genes; the real-time running data is gated and reconstructed to obtain stiffness identification data, including: The real-time running data is subjected to network correction estimation to obtain the target correction data; The target correction data is gated and mapped to obtain a gating mapping factor; The target identification data is determined based on the gating mapping factor and the physical gene.

3. The method according to claim 2, characterized in that, The target identification data includes identification bias current and stiffness identification data; Based on the gating mapping factor and the physical gene, the target identification data is determined to include: Obtain the model bias current, and determine the identification bias current based on the gating mapping factor and the model bias current; The stiffness identification data is determined based on the identified bias current, the gating mapping factor, and the physical gene.

4. The method according to claim 1 or 2, characterized in that, The prior stiffness model includes a current stiffness model and a displacement stiffness model; wherein, the current stiffness model is used to characterize the change of current stiffness with bias current and dynamic reluctance, and the displacement stiffness model is used to characterize the change of displacement stiffness with bias current and dynamic reluctance.

5. The method according to claim 4, characterized in that, The current stiffness model and the displacement stiffness model are determined based on the bias current term and the dynamic magnetoresistive term. The bias current term is used to characterize the change of bias current with temperature, and the dynamic magnetoresistive term is used to characterize the change of dynamic magnetoresistive with rotational speed.

6. The method according to claim 1, characterized in that, The stiffness identification model has a neural network; the stiffness identification model is trained in the following manner: Acquire measured data under multiple operating conditions, including measured rotational speed, measured temperature, and measured bias current; Based on the measured data and the joint loss function, the neural network is trained to obtain a trained stiffness identification model.

7. The method according to claim 6, characterized in that, The joint loss function includes a data error term and a physical constraint term; wherein: The physical constraints include correction magnitude constraints, smoothness constraints, and bias current consistency constraints; wherein the constraint weights of the smoothness constraints and the bias current consistency constraints are determined by annealing scheduling.

8. The method according to claim 1, characterized in that, Before determining the target bias current based on the bias current increment and the current bias current, the method further includes: The bias current increment is updated based on the rate of change constraint and upper and lower limit constraints, so as to determine the target bias current based on the updated bias current increment.

9. A stiffness compensation device for a magnetic bearing rotor, characterized in that, The device includes: A data acquisition unit is used to acquire real-time operating data and target stiffness data of the magnetic bearing rotor; wherein, the real-time operating data includes at least the current bias current; The data identification unit is used to perform gated correction and reconstruction on the real-time running data according to the pre-trained stiffness identification model to obtain target identification data; wherein, the stiffness identification model is constructed based on a prior stiffness model. The incremental determination unit is used to construct a stiffness error vector and a Jacobian matrix based on the target stiffness data and the target identification data, respectively, and to perform weighted cost matching on the stiffness error vector and the Jacobian matrix to obtain the bias current increment. The stiffness compensation unit is used to determine a target bias current based on the bias current increment and the current bias current. The target bias current is used to perform outer-loop monitoring control on the magnetic bearing rotor to complete stiffness compensation.

10. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the stiffness compensation method for the magnetic bearing rotor as described in any one of claims 1 to 8.

Images