A magnetic suspension rotor control method, device, computer equipment and medium

Abstract

The present invention provides a magnetic levitation rotor control method, device, computer equipment and medium. The magnetic levitation rotor control method includes: obtaining the position error data of the magnetic levitation rotor; inputting the position error data into a pre-constructed controller, obtaining the coil current at the current moment through the control law in the controller, the control law calculates the coil current through the first parameter and the second parameter, the first parameter and the second parameter are obtained through a pre-constructed radial basis function neural network model, the pre-constructed radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor, the first parameter and the second parameter are obtained through the calculation of the stiffness coefficient; the magnetic levitation rotor is controlled according to the coil current at the current moment. Through the present invention, the nonlinear effect of the electromagnetic force of the magnetic levitation bearing is taken into account, and the control accuracy and stability of the magnetic levitation rotor are improved.

Description

Technical Field

[0001] The present invention relates to the field of magnetic suspension rotor control, and in particular to a magnetic suspension rotor control method, device, computer equipment and medium. Background Art

[0002] Electromagnetic bearings, also known as active magnetic bearings (AMB), are a new type of contactless bearings with no mechanical contact, no friction and wear, low power loss, low failure rate, high achievable speed, and active control of the dynamic characteristics of the suspended rotor system. They are widely used in high-speed motors, satellite control torque gyroscopes, aircraft engines and other high-speed occasions. However, achieving effective control of the magnetically suspended rotor is a key issue that needs to be urgently solved in its engineering application.

[0003] For magnetically suspended rotors with small external excitations and low control accuracy in the operating environment, they deflect at a small angle near the equilibrium position, and the nonlinear changes in the electromagnetic force stiffness coefficient have little effect on the magnetic force. The existing control method can meet the stable suspension of the rotor; however, for magnetically suspended rotor systems operating under multi-foundation excitation conditions such as offshore floating platforms and space, in addition to internal excitations such as unbalanced mass and internal damping, they are also subject to various external excitations during operation, and are ultimately transmitted to the rotor through the frame foundation and the support system. The coupling torque generated by the rotation of the frame foundation will increase the rotor runout (moving frame effect) and even collide with the protective bearing, directly causing internal and external damage to the rotor. The moving frame effect will cause the rotor to deviate from the equilibrium position at a large angle, causing the magnetic force stiffness coefficient of the electromagnetic bearing to change nonlinearly, reducing the magnetic force control accuracy, and thus reducing the rotor suspension stability. However, the control algorithm of the existing magnetically suspended rotor ignores the influence of the nonlinear change of the electromagnetic force stiffness coefficient of the electromagnetic bearing, and has the technical disadvantages of low control accuracy and poor stability of the magnetically suspended rotor. Summary of the invention

[0004] In order to improve the control accuracy and stability of a magnetically suspended rotor, the present invention proposes a magnetically suspended rotor control method, device, computer equipment and medium taking into account the nonlinear effect of the electromagnetic force of a magnetically suspended bearing.

[0005] In a first aspect, the present invention provides a method for controlling a magnetically suspended rotor, the method comprising:

[0006] Acquire position error data of the magnetically suspended rotor;

[0007] The position error data is input into a pre-built controller, and the coil current at the current moment is obtained through the control law in the controller. The control law calculates the coil current through a first parameter and a second parameter. The first parameter and the second parameter are obtained through a pre-built radial basis function neural network model. The pre-built radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through a nonlinear electromagnetic force mathematical model of the electromagnetic bearing and a suspended rotor dynamics model. The first parameter and the second parameter are obtained through stiffness coefficient calculation;

[0008] The magnetically suspended rotor is controlled according to the coil current at the current moment.

[0009] Considering that when the magnetic levitation rotor is subjected to internal and external excitations during operation, a moving frame effect will occur, causing the stiffness coefficient of the electromagnetic bearing magnetic force to change nonlinearly, and the nonlinear change of the stiffness coefficient of the electromagnetic bearing magnetic force is ignored in the prior art, which reduces the control accuracy and stability of the magnetic levitation rotor. Through the above method, the nonlinear change of the stiffness coefficient of the electromagnetic bearing electromagnetic force is simulated by a radial basis function neural network model, and the parameters in the controller control law are calculated using the stiffness coefficient simulated by the radial basis function neural network model, and then the coil current at the current moment is obtained to control the magnetic levitation rotor, improve the stability of the magnetic levitation rotor, and achieve high-precision control of the magnetic levitation rotor. In addition, since the radial basis function neural network model has the characteristics of distributed parallel processing of information and distributed storage of information, the use of the radial basis function neural network model can effectively improve the fault tolerance of the magnetic levitation rotor control.

[0010] In combination with the first aspect, in a first embodiment of the first aspect, the step of constructing the radial basis function neural network model includes:

[0011] Obtain the mathematical model of nonlinear electromagnetic force of electromagnetic bearing and the dynamic model of suspended rotor;

[0012] According to the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor, a nonlinear mathematical model of the magnetic suspension rotor is obtained, wherein the nonlinear mathematical model of the magnetic suspension rotor includes a first parameter and a second parameter;

[0013] According to the nonlinear mathematical model of the magnetic levitation rotor, a radial basis function neural network model is constructed so that the first parameter and the second parameter output by the radial basis function neural network model satisfy the nonlinear mathematical model of the magnetic levitation rotor.

[0014] Through the above embodiments, the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor are combined to obtain the nonlinear mathematical model of the magnetic suspension rotor. The nonlinear mathematical model of the magnetic suspension rotor reflects the nonlinear change of the stiffness coefficient of the magnetic force of the electromagnetic bearing. The radial basis function neural network model is used to simulate the nonlinear change of the stiffness coefficient in the nonlinear mathematical model of the magnetic suspension rotor, so that the radial basis function neural network model has the ability to nonlinearly map the electromagnetic force of the electromagnetic bearing. Then, the parameters obtained by the radial basis function neural network model fully consider the nonlinearity of the stiffness coefficient of the magnetic suspension rotor, thereby improving the control accuracy of the magnetic suspension rotor.

[0015] In combination with the first aspect or the first embodiment of the first aspect, in a second embodiment of the first aspect, the radial basis function neural network model includes multiple Gaussian kernel functions, the Gaussian kernel function is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing, and the step of the radial basis function neural network model outputting the first parameter and the second parameter includes:

[0016] Obtain multiple Gaussian kernel function values ​​according to the position error data and the multiple Gaussian kernel functions;

[0017] The first parameter and the second parameter are calculated according to the Gaussian kernel function values ​​and the weights corresponding to the Gaussian kernel function values. The weights are determined by using an adaptive algorithm based on the position error data and the coil current at the previous moment.

[0018] Through the above embodiment, the weights corresponding to the Gaussian kernel function values ​​are continuously updated using an adaptive algorithm through position error data and the coil current at the previous moment, so that the radial basis function neural network model can more accurately simulate the nonlinear changes in the stiffness coefficient of the electromagnetic force of the electromagnetic bearing and has strong robust performance.

[0019] In combination with the first aspect, in a third embodiment of the first aspect, the controller includes any one of a proportional-derivative controller and a sliding mode controller.

[0020] In a second aspect, the present invention further provides a magnetic suspension rotor control device, the device comprising:

[0021] An acquisition module, used for acquiring position error data of the magnetic suspension rotor;

[0022] A calculation module, used for inputting position error data into a pre-built controller, obtaining the coil current at the current moment through a control law in the controller, the control law calculating the coil current through a first parameter and a second parameter, the first parameter and the second parameter being obtained through a pre-built radial basis function neural network model, the pre-built radial basis function neural network model simulating the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through a nonlinear electromagnetic force mathematical model of the electromagnetic bearing and a suspended rotor dynamics model, the first parameter and the second parameter being obtained through stiffness coefficient calculation;

[0023] The control module is used to control the magnetic suspension rotor according to the coil current at the current moment.

[0024] Considering that when the magnetic levitation rotor is subjected to internal and external excitations during operation, a moving frame effect will occur, causing the stiffness coefficient of the electromagnetic bearing magnetic force to change nonlinearly, and the prior art ignores the nonlinear change of the stiffness coefficient of the electromagnetic bearing magnetic force, which reduces the control accuracy and stability of the magnetic levitation rotor. Through the above device, the nonlinear change of the stiffness coefficient of the electromagnetic bearing electromagnetic force is simulated by a radial basis function neural network model, and the parameters in the controller control law are calculated using the stiffness coefficient simulated by the radial basis function neural network model, and then the coil current at the current moment is obtained to control the magnetic levitation rotor, improve the stability of the magnetic levitation rotor, and achieve high-precision control of the magnetic levitation rotor. In addition, since the radial basis function neural network model has the characteristics of distributed parallel processing of information and distributed storage of information, the use of the radial basis function neural network model can effectively improve the fault tolerance of the magnetic levitation rotor control.

[0025] In combination with the second aspect, in a first embodiment of the second aspect, the computing module constructs a radial basis function neural network model through the following submodules, including:

[0026] An acquisition submodule is used to acquire a mathematical model of nonlinear electromagnetic force of an electromagnetic bearing and a dynamic model of a suspended rotor;

[0027] A determination submodule is used to obtain a nonlinear mathematical model of a magnetically suspended rotor according to a nonlinear electromagnetic force mathematical model of an electromagnetic bearing and a dynamic model of a suspended rotor, wherein the nonlinear mathematical model of a magnetically suspended rotor includes a first parameter and a second parameter;

[0028] The construction submodule is used to construct a radial basis function neural network model according to the nonlinear mathematical model of the magnetic levitation rotor, so that the first parameter and the second parameter output by the radial basis function neural network model meet the nonlinear mathematical model of the magnetic levitation rotor.

[0029] Through the above embodiments, the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor are combined to obtain the nonlinear mathematical model of the magnetic suspension rotor. The nonlinear mathematical model of the magnetic suspension rotor reflects the nonlinear change of the stiffness coefficient of the magnetic force of the electromagnetic bearing. The radial basis function neural network model is used to simulate the nonlinear change of the stiffness coefficient in the nonlinear mathematical model of the magnetic suspension rotor, so that the radial basis function neural network model has the ability to nonlinearly map the electromagnetic force of the electromagnetic bearing. Then, the parameters obtained by the radial basis function neural network model fully consider the nonlinearity of the stiffness coefficient of the magnetic suspension rotor, thereby improving the control accuracy of the magnetic suspension rotor.

[0030] In combination with the second aspect or the first embodiment of the second aspect, in the second embodiment of the second aspect, in the calculation module, the radial basis function neural network model includes multiple Gaussian kernel functions, and the Gaussian kernel function is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing. The calculation module outputs the first parameter and the second parameter through the following submodules, including:

[0031] A first calculation submodule, used for obtaining multiple Gaussian kernel function values ​​according to the position error data and multiple Gaussian kernel functions;

[0032] The second calculation submodule is used to calculate the first parameter and the second parameter according to each Gaussian kernel function value and the weight corresponding to each Gaussian kernel function value, and the weight is determined by using an adaptive algorithm based on the position error data and the coil current at the previous moment.

[0033] Through the above embodiment, the weights corresponding to the Gaussian kernel function values ​​are continuously updated using an adaptive algorithm through position error data and the coil current at the previous moment, so that the radial basis function neural network model can more accurately simulate the nonlinear changes in the stiffness coefficient of the electromagnetic force of the electromagnetic bearing and has strong robust performance.

[0034] In combination with the second aspect, in a third embodiment of the second aspect, the controller in the calculation module includes any one of a proportional differential controller and a sliding mode controller.

[0035] In a third aspect, the present invention further provides a computer device, comprising a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the steps of the magnetic levitation rotor control method of the first aspect or any embodiment of the first aspect by executing the computer instructions.

[0036] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the magnetically suspended rotor control method of the first aspect or any embodiment of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 is a flow chart of a magnetic suspension rotor control method proposed according to an exemplary embodiment;

[0039] Figure 2 is a schematic diagram of the electromagnetic force on the x-axis exerted on the magnetically suspended rotor in an example;

[0040] Figure 3 is a schematic diagram of the structure of a magnetically suspended rotor in an example;

[0041] Figure 4 is a schematic diagram of the structure of a radial basis function neural network in an example;

[0042] Figure 5 is a block diagram of a magnetically suspended rotor control system in an example;

[0043] Figure 6 is a control block diagram of a magnetically suspended rotor control system in an example;

[0044] Figure 7 It is a structural schematic diagram of a magnetic suspension rotor control device proposed according to an exemplary embodiment;

[0045] Figure 8 It is a schematic diagram of the hardware structure of a computer device proposed according to an exemplary embodiment. DETAILED DESCRIPTION

[0046] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0048] In order to improve the control accuracy and stability of a magnetically suspended rotor, the present invention proposes a magnetically suspended rotor control method, device, computer equipment and medium taking into account the nonlinear effect of the electromagnetic force of a magnetically suspended bearing.

[0049] Figure 1 FIG. 1 is a flow chart of a magnetic suspension rotor control method according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps S101 to S103.

[0050] Step S101: Acquire position error data of the magnetically suspended rotor.

[0051] In an optional embodiment, the position error data of the magnetically suspended rotor is determined by comparing the ideal position and the current position.

[0052] In an optional embodiment, the position error data includes the position error and the derivative of the position error.

[0053] Step S102: input the position error data into a pre-built controller, and obtain the coil current at the current moment through the control law in the controller, the control law calculates the coil current through the first parameter and the second parameter, the first parameter and the second parameter are obtained through a pre-built radial basis function neural network model, the pre-built radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the suspended rotor dynamics model, and the first parameter and the second parameter are obtained by calculating the stiffness coefficient.

[0054] In an optional embodiment, the controller includes any one of a proportional-derivative controller and a sliding mode controller.

[0055] In an optional embodiment, the radial basis function neural network model outputs a first parameter and a second parameter by simulating the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing, so that the relationship between the stiffness coefficient and the first parameter, and the relationship between the stiffness coefficient and the second parameter satisfy the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing.

[0056] In an optional embodiment, the nonlinear variation of the electromagnetic force stiffness coefficient of the electromagnetic bearing can be obtained through a nonlinear electromagnetic force mathematical model of the electromagnetic bearing and a suspended rotor dynamics model.

[0057] Step S103: Control the magnetically suspended rotor according to the coil current at the current moment.

[0058] In an optional embodiment, at different times, the coil current controlling the magnetic suspension rotor is different. The coil current changes with the change of the position error data, so that the position of the magnetic suspension rotor is close to the ideal position.

[0059] Taking into account that when the magnetic levitation rotor is subjected to internal and external excitations during operation, a moving frame effect will occur, causing the stiffness coefficient of the electromagnetic bearing magnetic force to change nonlinearly, and the prior art ignores the nonlinear change of the stiffness coefficient of the electromagnetic bearing magnetic force, which reduces the control accuracy and stability of the magnetic levitation rotor. Through the method provided in an embodiment of the present invention, a radial basis function neural network model is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing, and the stiffness coefficient simulated by the radial basis function neural network model is used to calculate the parameters in the controller control law, thereby obtaining the coil current at the current moment to control the magnetic levitation rotor, thereby improving the stability of the magnetic levitation rotor and achieving high-precision control of the magnetic levitation rotor.

[0060] In addition, since the radial basis function neural network model has the characteristics of distributed parallel processing of information and distributed storage of information, the use of the radial basis function neural network model can effectively improve the fault tolerance of magnetic levitation rotor control.

[0061] In one example, the radial basis function neural network model in step S102 is constructed as follows:

[0062] Firstly, the mathematical model of the nonlinear electromagnetic force of the electromagnetic bearing and the dynamic model of the suspended rotor are obtained.

[0063] In an alternative embodiment, Figure 2 It is a schematic diagram of the electromagnetic force on the magnetically suspended rotor on the x-axis. F is the electromagnetic force on the rotor on the x-axis, i is the coil current, I0 is the equivalent bias current, and m is the rotor mass. The process of electromagnetic bearing differential control of the magnetically suspended rotor can be described through the mathematical model of the nonlinear electromagnetic force of the electromagnetic bearing. The electromagnetic bearing adopts a differential excitation method, which arranges an excitation coil in the upper and lower directions of the X direction. The electromagnetic force on the rotor in the X direction is the difference between the suction forces of the upper and lower electromagnets. The calculation formula is as follows:

[0064]

[0065] Among them, F is the electromagnetic force exerted on the rotor in the X direction, μ0 is the air magnetic permeability, N is the number of turns of the electromagnet coil winding, A is the pole area, i is the coil current, I0 is the equivalent bias current, and s0 is the unidirectional air gap between the stator and rotor at the equilibrium position.

[0066] Expanding according to Taylor's formula, ignoring the less influential part above the 5th order, and retaining it to the 3rd order, we can get:

[0067]

[0068] make

[0069]

[0070]

[0071]

[0072]

[0073]

[0074] Then formula (2) can be written as:

[0075]

[0076] It can be seen from formula (3) that the mathematical model of the nonlinear electromagnetic force of the electromagnetic bearing contains higher-order nonlinear terms in addition to the first-order linear terms. After sorting, we can get:

[0077] F = k i 'i+k s 'x=(k i +Δk i )i+(k s +Δk s )x (4)

[0078] Among them, the force-current stiffness nonlinear term Force-displacement stiffness nonlinearity Force-current stiffness coefficient Force-displacement stiffness coefficient

[0079] In an alternative embodiment, Figure 3 The magnetic suspension rotor is a schematic diagram of the structure of the magnetic suspension rotor. According to the dynamics theory, the magnetic suspension rotor is on the x-axis. Figure 3 The dynamic equation of the four-degree-of-freedom magnetic suspension rotor can be listed as:

[0080]

[0081] Where, m is the rotor mass; J p , J d are the polar moment of inertia and equatorial moment of inertia of the rotor respectively; ω is the rotor speed; x and y are the displacements of the rotor center of mass on the x and y axes respectively; represents the second derivative of x, that is, the acceleration of the rotor center of mass on the x-axis; represents the second derivative of y, that is, the acceleration of the rotor center of mass on the y axis; θ x ,θ y are the angles that the rotor mass center rotates around the x and y axes, respectively. is the angular velocity of the rotor center of mass around x, is the angular acceleration of the rotor center of mass about x, is the angular velocity of the rotor center of mass around y, is the angular acceleration of the rotor mass center around y; l ra is the distance from the magnetic bearing A to the rotor mass center, l rb is the distance from the magnetic bearing B to the rotor mass center; f xa ,f ya are the electromagnetic forces of the magnetic bearing A in the X and Y directions respectively; f xb ,f yb are the electromagnetic forces of the magnetic bearing B in the X and Y directions respectively.

[0082] Then, according to the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor, a nonlinear mathematical model of the magnetic suspension rotor is obtained, and the nonlinear mathematical model of the magnetic suspension rotor includes a first parameter and a second parameter.

[0083] In an optional embodiment, formula (4) is substituted into formula (5), and the uncertainty of interferences such as odd-frequency vibration and unmodeled dynamics of the high-speed magnetic suspension rotor is considered. Therefore, the interference term is added, and the expression of the nonlinear mathematical model of the magnetic suspension rotor can be obtained as follows:

[0084]

[0085] Among them, f dx 、f dy are the components of the interference force in the X and Y directions respectively; i x 、i y They are the control current in the X and Y directions respectively; k ix +Δk ix is the force-current stiffness coefficient in the X direction, Δk ix is the force-current stiffness nonlinear term in the X direction; k sx +Δk sx is the force-displacement stiffness coefficient in the X direction, Δk sx is the force-displacement stiffness nonlinear term in the X direction; k iy +Δk iy Force-current stiffness coefficient in the Y direction, Δk iy is the force-current stiffness nonlinear term in the Y direction, k sx +Δk sx is the force-displacement stiffness coefficient in the Y direction, Δk sy is the force-displacement stiffness nonlinear term in the Y direction.

[0086] Taking the X direction as an example, the nonlinear mathematical model of the magnetic suspension rotor can be expressed as:

[0087]

[0088] Let x1 = x, u=i x, then it can be expressed as a second-order nonlinear uncertain system through nonlinear transformation:

[0089]

[0090] in, represents the first derivative of x, that is, the speed of the rotor center of mass on the x-axis; α is the first parameter, β is the second parameter, And |d(t)|≤D, D is the upper bound of interference.

[0091] Finally, according to the nonlinear mathematical model of the magnetic levitation rotor, a radial basis function neural network model is constructed so that the first parameter and the second parameter output by the radial basis function neural network model satisfy the nonlinear mathematical model of the magnetic levitation rotor.

[0092] In an optional embodiment, the radial basis function neural network (RBFNN) has the advantages of strong learning ability and continuous nonlinear function approximation. The hidden layer in the RBFNN uses an activation function, which has the characteristics of a nonlinear mapping function and can be used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing, so that the output first parameter and the second parameter meet the nonlinear mathematical model of the magnetic suspension rotor. The structure of the RBFNN is as follows: Figure 4 shown.

[0093] In RBFNN, x=[x i ] T is the input of the network, and the hidden layer output of the network is h = [h j ] T , the output of the jth neuron in the hidden layer.

[0094]

[0095] in, is the coordinate vector of the center point of the Gaussian basis function of the jth neuron in the hidden layer, b j =[b1,…,b m ] T , b j is the width of the Gaussian basis function of the jth neuron in the hidden layer.

[0096] The RBFNN weights are w=[w1,…,w m ] T .

[0097] The RBFNN output is y(t)=w T h=w1h1+w2h2+…+w m h m .

[0098] In the embodiment of the present invention, the ideal suspension position of the magnetic suspension rotor in the X direction is defined as x r , then the rotor error e is: e = x r -x, The Y direction is the same as the X direction, so we will not go into details here. The values ​​of the first parameter α and the second parameter β are approximated by RBFNN, and the input of RBFNN is the position error and the derivative of the position error. Then the RBFNN output is:

[0099]

[0100] Among them, h α (x) and h β (x) is the Gaussian kernel function of RBFNN; W and P are the ideal weights of RBFNN; ε α and ε β is the approximation error; α and β are the ideal outputs of RBFNN.

[0101] The actual output of RBFNN is shown in formula (11).

[0102]

[0103] in, and is the actual output of RBFNN; and is the actual weight of RBFNN.

[0104] Through the above embodiments, the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor are combined to obtain the nonlinear mathematical model of the magnetic suspension rotor. The nonlinear mathematical model of the magnetic suspension rotor reflects the nonlinear change of the stiffness coefficient of the magnetic force of the electromagnetic bearing. By training the radial basis function neural network model, the trained radial basis function neural network model can simulate the nonlinear change of the stiffness coefficient in the nonlinear mathematical model of the magnetic suspension rotor, and has the ability to nonlinearly map the electromagnetic force of the electromagnetic bearing. Furthermore, the parameters obtained by the radial basis function neural network model fully consider the nonlinearity of the stiffness coefficient of the magnetic suspension rotor, thereby improving the control accuracy of the magnetic suspension rotor.

[0105] In one example, the radial basis function neural network model includes multiple Gaussian kernel functions, and the Gaussian kernel function is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing. The specific steps of the radial basis function neural network model outputting the first parameter and the second parameter include:

[0106] First, multiple Gaussian kernel function values ​​are obtained according to the position error data and multiple Gaussian kernel functions.

[0107] Then, the first parameter and the second parameter are calculated according to the Gaussian kernel function values ​​and the weights corresponding to the Gaussian kernel function values, and the weights are determined by using an adaptive algorithm according to the position error data and the coil current at the previous moment.

[0108] Through the above embodiment, the weights corresponding to the Gaussian kernel function values ​​are continuously updated using an adaptive algorithm through position error data and the coil current at the previous moment, so that the radial basis function neural network model can more accurately simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing, thereby improving the robustness of the radial basis function neural network model.

[0109] Figure 5 The structure diagram of the magnetic suspension rotor control system is shown in Figure 1. The magnetic suspension rotor control system includes a controller, a power amplifier, and a sensor. Figure 6 The control block diagram of the system is shown in Figure 2. The control law in the controller is determined by the first parameter and the second parameter output by the RBFNN model, and the weight in the RBFNN model is determined by an adaptive algorithm. In the magnetic suspension rotor control system, the sensor collects the position data of the magnetic suspension rotor, and then obtains the position error data, which is input into the controller. The controller outputs the coil current according to the position error data, and the coil current is input into the power amplifier to obtain the amplified coil current, so that the electromagnetic force of the magnetic bearing controls the magnetic suspension rotor.

[0110] In one example, the controller adopts a proportional plus derivative control (PD control). According to the first parameter and the second parameter output by the radial basis function neural network model, the control law of the PD control is designed as:

[0111]

[0112] Where u is the control law. When u is the control law in the X direction, u = i x ; When u is the control law in the Y direction, u=i y Design K = (k p k d ) T Let the polynomial s 2 +k d s+k p = 0 are all in the left half-complex plane, then when t→∞, e(t)→0,

[0113] The following proves that the controller is stable through the Lyapunov function.

[0114] Substituting formula (12) into formula (8), the closed-loop system is obtained:

[0115]

[0116] make

[0117] Then formula (13) can be written as

[0118]

[0119] The optimal weights of the RBF neural network are:

[0120]

[0121] Therefore, the allowable error of the RBF neural network is:

[0122]

[0123] Substituting formula (15) into formula (13), formula (13) can be written as:

[0124]

[0125] The Lyapunov function of the closed-loop system is defined as:

[0126]

[0127] Among them, γ and η are positive constants, the matrix D is a symmetric positive definite matrix, and satisfies the following Lyapunov equation:

[0128] Λ T D+DΛ=-Q

[0129] Among them, Q>0.

[0130] Pick make Then formula (16) can be written as:

[0131]

[0132] To prove that the controller is stable, we need to prove The proof process is as follows:

[0133]

[0134] Substitute M into because We can get:

[0135]

[0136]

[0137] but

[0138]

[0139] Therefore, the adaptive law of the magnetic suspension rotor system can be obtained as:

[0140]

[0141] Substituting into

[0142]

[0143] because Therefore, the approximation error ε can be reduced by designing an RBF neural network. α and ε β Restricted small enough to make

[0144] At the same time, the rotor displacement and the derivative of the rotor displacement are continuous and bounded, so according to the LaSalle invariance principle, the rotor position error and the derivative of the position error are also bounded. The control law, the neural network output, and the adaptive law are also bounded, so the rotor system control quantity is bounded. Therefore, the magnetic suspension rotor controller proposed in the embodiment of the present invention is asymptotically stable.

[0145] In one example, the controller adopts a sliding mode controller. According to the first parameter and the second parameter output by the radial basis function neural network model, the control law of the PD control is designed as:

[0146]

[0147] Where u is the control law. When u is the control law in the X direction, u = i x ; When u is the control law in the Y direction, u=i y The designed sliding surface is The switching parameter is c, c>0, and η is a positive constant.

[0148] The following proves that the controller is stable through the Lyapunov function.

[0149] Derivative the sliding surface function:

[0150]

[0151] The Lyapunov function of the closed-loop system is defined as:

[0152]

[0153] Among them, γ and η are positive constants.

[0154] Taking the derivative of V and substituting formula (21) into it, we can obtain:

[0155]

[0156] Therefore, the adaptive law of the magnetic suspension rotor system can be obtained as:

[0157]

[0158] Substituting formula (24) into formula (23) yields:

[0159]

[0160] Since the radial basis function neural network can be trained, the approximation error ε α and ε β Can be limited to be small enough, if λ≥|ε α +ε β u+d(t)|, we can get There exists λ0>0 and λ≥λ0+|ε α |+|ε β u|+d(t), so that

[0161] At the same time, because V≥0 ,so Bounded, that is, the control law, the neural network output, and the adaptive law are all bounded, and the rotor system control quantity is also bounded. Therefore, the magnetic suspension rotor controller proposed in the embodiment of the present invention is asymptotically stable.

[0162] Based on the same inventive concept, the embodiment of the present invention also provides a magnetic suspension rotor control device, such as Figure 7 As shown, the device comprises:

[0163] The acquisition module 701 is used to acquire the position error data of the magnetic suspension rotor; for details, please refer to the description of step S101 in the above embodiment, which will not be repeated here.

[0164] The calculation module 702 is used to input the position error data into a pre-built controller, and obtain the coil current at the current moment through the control law in the controller. The control law calculates the coil current through the first parameter and the second parameter. The first parameter and the second parameter are obtained through a pre-built radial basis function neural network model. The pre-built radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the suspended rotor dynamics model. The first parameter and the second parameter are obtained by calculating the stiffness coefficient. For details, please refer to the description of step S102 in the above embodiment, which will not be repeated here.

[0165] The control module 703 is used to control the magnetic suspension rotor according to the coil current at the current moment. For details, please refer to the description of step S103 in the above embodiment, which will not be repeated here.

[0166] Considering that when the magnetic levitation rotor is subjected to internal and external excitations during operation, a moving frame effect will occur, causing the stiffness coefficient of the electromagnetic bearing magnetic force to change nonlinearly, and the prior art ignores the nonlinear change of the stiffness coefficient of the electromagnetic bearing magnetic force, which reduces the control accuracy and stability of the magnetic levitation rotor. Through the above device, the nonlinear change of the stiffness coefficient of the electromagnetic bearing electromagnetic force is simulated by a radial basis function neural network model, and the parameters in the controller control law are calculated using the stiffness coefficient simulated by the radial basis function neural network model, and then the coil current at the current moment is obtained to control the magnetic levitation rotor, improve the stability of the magnetic levitation rotor, and achieve high-precision control of the magnetic levitation rotor. In addition, since the radial basis function neural network model has the characteristics of distributed parallel processing of information and distributed storage of information, the use of the radial basis function neural network model can effectively improve the fault tolerance of the magnetic levitation rotor control.

[0167] In one example, the calculation module 702 constructs a radial basis function neural network model through the following submodules, including:

[0168] The acquisition submodule is used to acquire the mathematical model of the nonlinear electromagnetic force of the electromagnetic bearing and the dynamic model of the suspended rotor; for details, please refer to the description in the above embodiment, which will not be repeated here.

[0169] The determination submodule is used to obtain a nonlinear mathematical model of the magnetically suspended rotor according to the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor. The nonlinear mathematical model of the magnetically suspended rotor includes a first parameter and a second parameter. For details, please refer to the description in the above embodiment and will not be repeated here.

[0170] The construction submodule is used to construct a radial basis function neural network model according to the nonlinear mathematical model of the magnetic levitation rotor, so that the first parameter and the second parameter output by the radial basis function neural network model meet the nonlinear mathematical model of the magnetic levitation rotor. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0171] Through the above embodiments, the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor are combined to obtain the nonlinear mathematical model of the magnetic suspension rotor. The nonlinear mathematical model of the magnetic suspension rotor reflects the nonlinear change of the stiffness coefficient of the magnetic force of the electromagnetic bearing. The radial basis function neural network model is used to simulate the nonlinear change of the stiffness coefficient in the nonlinear mathematical model of the magnetic suspension rotor, so that the radial basis function neural network model has the ability to nonlinearly map the electromagnetic force of the electromagnetic bearing. Then, the parameters obtained by the radial basis function neural network model fully consider the nonlinearity of the stiffness coefficient of the magnetic suspension rotor, thereby improving the control accuracy of the magnetic suspension rotor.

[0172] In one example, in the calculation module 702, the radial basis function neural network model includes multiple Gaussian kernel functions, and the Gaussian kernel function is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing. The calculation module 702 outputs the first parameter and the second parameter through the following submodules, including:

[0173] The first calculation submodule is used to obtain multiple Gaussian kernel function values ​​according to the position error data and multiple Gaussian kernel functions; for details, please refer to the description in the above embodiment, which will not be repeated here.

[0174] The second calculation submodule is used to calculate the first parameter and the second parameter according to each Gaussian kernel function value and the weight corresponding to each Gaussian kernel function value, and the weight is determined by using an adaptive algorithm according to the position error data and the coil current at the previous moment. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0175] Through the above embodiment, the weights corresponding to the Gaussian kernel function values ​​are continuously updated using an adaptive algorithm through position error data and the coil current at the previous moment, so that the radial basis function neural network model can more accurately simulate the nonlinear changes in the stiffness coefficient of the electromagnetic force of the electromagnetic bearing and has strong robust performance.

[0176] In one example, the controller in the calculation module 702 includes any one of a proportional differential controller and a sliding mode controller. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0177] The above modules can be implemented in whole or in part by software, hardware or a combination thereof. The above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above modules.

[0178] Figure 8 FIG. 1 is a schematic diagram of a hardware structure of a computer device according to an exemplary embodiment. Figure 8As shown, the device includes one or more processors 810 and a memory 820, and the memory 820 includes a persistent memory, a volatile memory, and a hard disk. Figure 8 A processor 810 is taken as an example. The device may also include: an input device 830 and an output device 840.

[0179] The processor 810, the memory 820, the input device 830 and the output device 840 may be connected via a bus or other means. Figure 8 The example of connecting through bus is taken in the following.

[0180] The processor 810 may be a central processing unit (CPU). The processor 810 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips. A general-purpose processor may be a microprocessor or the processor may be any conventional processor.

[0181] The memory 820 is a non-transient computer-readable storage medium, including a persistent memory, a volatile memory, and a hard disk, and can be used to store non-transient software programs, non-transient computer executable programs, and modules, such as program instructions / modules corresponding to the magnetic levitation rotor control method in the embodiment of the present application. The processor 810 executes various functional applications and data processing of the server by running the non-transient software programs, instructions, and modules stored in the memory 820, that is, implementing any of the above-mentioned magnetic levitation rotor control methods.

[0182] The memory 820 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data required for use, etc. In addition, the memory 820 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 820 may optionally include a memory remotely arranged relative to the processor 810, and these remote memories may be connected to the data processing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0183] The input device 830 can receive input digital or character information and generate signal input related to user settings and function control. The output device 840 can include display devices such as display screens.

[0184] One or more modules are stored in the memory 820, and when executed by one or more processors 810, the execution is as follows: Figure 1 The method shown.

[0185] The above-mentioned product can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to Figure 1 Related description of the illustrated embodiment.

[0186] The embodiment of the present invention further provides a non-transitory computer storage medium, which stores computer executable instructions, and the computer executable instructions can execute the control method in any of the above method embodiments. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0187] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0188] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features applied herein.

Claims

1. A magnetic suspension rotor control method, characterized in that: The method comprises: Acquire position error data of the magnetically suspended rotor; The position error data is input into a pre-built controller, and the coil current at the current moment is obtained through the control law in the controller, the control law calculates the coil current through a first parameter and a second parameter, the first parameter and the second parameter are obtained through a pre-built radial basis function neural network model, the pre-built radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through a nonlinear electromagnetic force mathematical model of the electromagnetic bearing and a suspended rotor dynamics model, the first parameter and the second parameter are obtained by calculating the stiffness coefficient; the controller is a proportional differential controller, and the control law is expressed by a first relational expression, which is: in, For the control law, x r is the ideal suspension position of the magnetic suspension rotor in the X direction, e is the position error of the magnetically suspended rotor, is the derivative of the position error, , , , , is the actual value of the first parameter, is the actual value of the second parameter; Controlling the magnetically suspended rotor according to the coil current at the current moment; The method further comprises: The stability analysis of the controller is performed through the Lyapunov function. The stability analysis process includes: Substituting the first relational expression into the second-order nonlinear uncertainty system of the magnetically suspended rotor, a closed-loop system is obtained. The closed-loop system is represented by a second relational expression. The second relational expression is: in, , is the component of the disturbance force in the X direction, m is the rotor mass, is the ideal value of the first parameter, is the ideal value of the second parameter; make , , the second relation is transformed into the third relation: The optimal weights of the radial basis function neural network are: in, and is the optimal weight of the radial basis function neural network; The error allowed by the radial basis function neural network is expressed by the fourth relation, which is: ， in, and is the approximation error; Substituting the fourth relation into the third relation, the third relation is transformed into the fifth relation: The Lyapunov function of the closed-loop system is: in, and is a positive constant, the matrix D is a symmetric positive definite matrix that satisfies the following Lyapunov equation , ; Pick , , ,make , transform the fifth relation into: Further get the following formula: Will M Substitution ,because , , we can get: The adaptive law of the magnetic suspension rotor system can be obtained as: Substituting the adaptive law into the equation, we get: in, ,like , determine that the controller is stable.

2. The method according to claim 1, characterized in that The steps of constructing the radial basis function neural network model include: Obtain the mathematical model of nonlinear electromagnetic force of electromagnetic bearing and the dynamic model of suspended rotor; According to the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor, a nonlinear mathematical model of the magnetic suspension rotor is obtained, wherein the nonlinear mathematical model of the magnetic suspension rotor includes the first parameter and the second parameter; The radial basis function neural network model is constructed according to the nonlinear mathematical model of the magnetic levitation rotor, so that the first parameter and the second parameter output by the radial basis function neural network model satisfy the nonlinear mathematical model of the magnetic levitation rotor.

3. The method according to claim 1 or 2, characterized in that: The radial basis function neural network model includes a plurality of Gaussian kernel functions, and the Gaussian kernel functions are used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing. The step of outputting the first parameter and the second parameter by the radial basis function neural network model includes: Obtaining multiple Gaussian kernel function values ​​according to the position error data and multiple Gaussian kernel functions; The first parameter and the second parameter are calculated according to each of the Gaussian kernel function values ​​and the weights corresponding to each of the Gaussian kernel function values, and the weights are determined by using an adaptive algorithm based on the position error data and the coil current at a previous moment.

4. A magnetically suspended rotor control device, characterized in that: The device comprises: An acquisition module, used for acquiring position error data of the magnetic suspension rotor; A calculation module is used to input the position error data into a pre-built controller, and obtain the coil current at the current moment through the control law in the controller, wherein the control law calculates the coil current through a first parameter and a second parameter, and the first parameter and the second parameter are obtained through a pre-built radial basis function neural network model, and the pre-built radial basis function neural network model simulates the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing through a nonlinear electromagnetic force mathematical model of the electromagnetic bearing and a suspended rotor dynamics model, and the first parameter and the second parameter are obtained by calculating the stiffness coefficient; the controller is a proportional differential controller, and the control law is expressed by a first relational expression, and the first relational expression is: in, For the control law, x r is the ideal suspension position of the magnetic suspension rotor in the X direction, e is the position error of the magnetically suspended rotor, is the derivative of the position error, , , , , is the actual value of the first parameter, is the actual value of the second parameter; A control module, used for controlling the magnetic suspension rotor according to the coil current at a current moment; The device comprises: The stability analysis of the controller is performed through the Lyapunov function. The stability analysis process includes: Substituting the first relational expression into the second-order nonlinear uncertainty system of the magnetically suspended rotor, a closed-loop system is obtained. The closed-loop system is represented by a second relational expression. The second relational expression is: in, , is the component of the disturbance force in the X direction, m is the rotor mass, is the ideal value of the first parameter, is the ideal value of the second parameter; make , , the second relation is transformed into the third relation: The optimal weights of the radial basis function neural network are: in, and is the optimal weight of the radial basis function neural network; The error allowed by the radial basis function neural network is expressed by the fourth relation, which is: ， in, and is the approximation error; Substituting the fourth relation into the third relation, the third relation is transformed into the fifth relation: The Lyapunov function of the closed-loop system is: in, and is a positive constant, the matrix D is a symmetric positive definite matrix that satisfies the following Lyapunov equation , ; Pick , , ,make , transform the fifth relation into: Further get the following formula: Will M Substitution ,because , , we can get: The adaptive law of the magnetic suspension rotor system can be obtained as: Substituting the adaptive law into the equation, we get: in, ,like , determine that the controller is stable.

5. The device according to claim 4, characterized in that The calculation module constructs the radial basis function neural network model through the following submodules, including: An acquisition submodule is used to acquire a mathematical model of nonlinear electromagnetic force of an electromagnetic bearing and a dynamic model of a suspended rotor; A determination submodule, configured to obtain a nonlinear mathematical model of a magnetically suspended rotor according to the nonlinear electromagnetic force mathematical model of the electromagnetic bearing and the dynamic model of the suspended rotor, wherein the nonlinear mathematical model of the magnetically suspended rotor includes the first parameter and the second parameter; The construction submodule is used to construct the radial basis function neural network model according to the nonlinear mathematical model of the magnetic levitation rotor, so that the first parameter and the second parameter output by the radial basis function neural network model meet the nonlinear mathematical model of the magnetic levitation rotor.

6. The device according to claim 4 or 5, characterized in that In the calculation module, the radial basis function neural network model includes a plurality of Gaussian kernel functions, and the Gaussian kernel function is used to simulate the nonlinear change of the stiffness coefficient of the electromagnetic force of the electromagnetic bearing. The calculation module outputs the first parameter and the second parameter through the following submodules, including: A first calculation submodule, used for obtaining a plurality of Gaussian kernel function values ​​according to the position error data and a plurality of Gaussian kernel functions; The second calculation submodule is used to calculate the first parameter and the second parameter according to each of the Gaussian kernel function values ​​and the weights corresponding to each of the Gaussian kernel function values, wherein the weights are determined by using an adaptive algorithm based on the position error data and the coil current at the previous moment.

7. A computer device, characterized in that: It comprises a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the steps of the magnetic levitation rotor control method according to any one of claims 1 to 3 by executing the computer instructions.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the magnetic levitation rotor control method according to any one of claims 1 to 3 are implemented.

Images