Method, device, system, storage medium and processor for determining displacement of magnetic bearing

By constructing a displacement soft measurement model based on wavelet neural network and using particle swarm optimization algorithm to train the control current of magnetic bearing coil, sensorless radial displacement self-detection was achieved. This solved the problem of magnetic bearing rotor falling due to eddy current sensor damage, reduced system size and cost, and improved suspension safety and stability.

CN113408598BActive Publication Date: 2024-11-08GREE ELECTRIC APPLIANCE INC OF ZHUHAI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110648565.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-06-10
Publication Date
2024-11-08
Estimated Expiration
2041-06-10

AI Technical Summary

Technical Problem

When measuring the radial displacement of magnetic bearing rotors, eddy current sensors are often used. If the eddy current sensor is damaged, it can cause the magnetic bearing rotor to fall, and it also occupies space and increases costs.

Method used

A wavelet neural network-based soft measurement model for displacement is adopted. By collecting sample data of control current and rotor displacement parameters of magnetic bearing coil, the wavelet neural network model is trained to realize sensorless radial displacement self-detection. The model parameters are optimized by particle swarm optimization algorithm to construct a radial displacement prediction module.

Benefits of technology

This technology improves the levitation safety of magnetic bearing rotors, avoids falls caused by damage to eddy current sensors, reduces the size and cost of the magnetic levitation bearing control system, and enhances system stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113408598B_ABST
    Figure CN113408598B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for determining the displacement of a magnetic bearing, a magnetic suspension bearing system, a storage medium and a processor. The method includes: constructing a displacement measurement model based on a wavelet neural network, recorded as a displacement soft measurement model; obtaining the current control current of the magnetic bearing coil when the current displacement parameter of the magnetic bearing rotor needs to be determined; using the current control current of the magnetic bearing coil as an input parameter, inputting the displacement soft measurement model, and determining the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, thereby determining the current displacement parameter of the magnetic bearing rotor. This scheme measures the radial displacement of the magnetic bearing rotor by a soft measurement method, avoids the magnetic bearing rotor from falling due to damage of the eddy current sensor when using the eddy current sensor, and can improve the suspension safety of the magnetic bearing rotor.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of magnetic levitation technology, and specifically relates to a method for determining the displacement of a magnetic bearing (such as a method for determining the displacement of a magnetic bearing of a magnetic bearing system), a device, a magnetic levitation bearing system (i.e., a magnetic bearing system), a storage medium and a processor, and especially relates to a soft measurement method for the displacement of a magnetic levitation bearing, a device, a magnetic levitation bearing system, a storage medium and a processor. Background Art

[0002] Magnetic bearings use magnetic field force to stably suspend the magnetic bearing rotor in the air. They have the characteristics of no mechanical wear, no need for lubrication, and no need for maintenance, and can achieve high-speed operation of the motor. In related solutions, the magnetic bearing system detects the displacement of the magnetic bearing rotor through a displacement sensor, compares it with the given displacement value, and then sends it to the suspension force control system to generate a suspension force control current.

[0003] In related schemes, eddy current sensors are often used to measure the radial displacement of the magnetic bearing rotor. If the eddy current sensor is damaged, the magnetic bearing rotor may fall.

[0004] Since the eddy current sensor takes up space, the size of the magnetic bearing control system becomes larger.

[0005] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The purpose of the present invention is to provide a method and device for determining the displacement of a magnetic bearing, a magnetic suspension bearing system, a storage medium and a processor, so as to solve the problem that when measuring the radial displacement of a magnetic bearing rotor, eddy current sensors are often used. If the eddy current sensors are damaged, the magnetic bearing rotor may fall off. The radial displacement of the magnetic bearing rotor can be measured by using a soft measurement method, so as to avoid the magnetic bearing rotor falling off if the eddy current sensors are damaged, thereby improving the suspension safety of the magnetic bearing rotor.

[0007] The present invention provides a method for determining the displacement of a magnetic bearing, wherein the magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor; the method for determining the displacement of the magnetic bearing comprises: constructing a displacement measurement model based on a wavelet neural network, recorded as a displacement soft measurement model; when it is necessary to determine the current displacement parameter of the magnetic bearing rotor, obtaining the current control current of the magnetic bearing coil; using the current control current of the magnetic bearing coil as an input parameter, inputting the displacement soft measurement model, and determining the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, thereby determining the current displacement parameter of the magnetic bearing rotor.

[0008] In some embodiments, the current control current of the magnetic bearing coil is used as an input parameter and input into the displacement soft measurement model, and the output parameter of the displacement soft measurement model is determined as the current displacement parameter of the magnetic bearing rotor, including: setting the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil.

[0009] In some embodiments, a displacement measurement model based on a wavelet neural network is constructed, denoted as a displacement soft measurement model, including: collecting a set control current of the magnetic bearing coil and a set displacement parameter of the magnetic bearing rotor as sample data; based on the sample data, the wavelet neural network is trained with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model.

[0010] In some embodiments, the number of the magnetic bearing coils is more than one; collecting the set control current of the magnetic bearing coils and the set displacement parameters of the magnetic bearing rotor as sample data includes: selecting the set control current of more than one magnetic bearing coils and the set displacement parameters of the magnetic bearing rotor as feature variables; normalizing the feature variables to form sampling data, which is used as sample data for training and testing.

[0011] In some embodiments, based on the sample data, a wavelet neural network is trained with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model, including: setting a wavelet neural network model; when the number of the magnetic bearing coils is more than one, the number of layers of the wavelet neural network model is more than one; based on a particle swarm optimization algorithm, the parameters of the set wavelet neural network model are optimized to obtain the optimized wavelet neural network model; based on the sample data, the set control current of the magnetic bearing coil is used as an input parameter and the set displacement parameter of the magnetic bearing rotor is used as an output parameter, the optimized wavelet neural network model is trained to obtain the displacement soft measurement model.

[0012] In some embodiments, it also includes: after determining the current displacement parameter of the magnetic bearing rotor, adjusting the current control current of the magnetic bearing coil according to the current displacement parameter of the magnetic bearing rotor to obtain an adjustment value of the current control current of the magnetic bearing coil; and controlling the operation of the magnetic bearing coil according to the adjustment value of the current control current of the magnetic bearing coil.

[0013] Matching the above method, on the other hand, the present invention provides a displacement determination device for a magnetic bearing, wherein the magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor; the displacement determination device for the magnetic bearing comprises: a modeling unit, configured to construct a displacement measurement model based on a wavelet neural network, recorded as a displacement soft measurement model; an acquisition unit, configured to acquire the current control current of the magnetic bearing coil when the current displacement parameter of the magnetic bearing rotor needs to be determined; a control unit, configured to use the current control current of the magnetic bearing coil as an input parameter, input it into the displacement soft measurement model, and determine the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, thereby determining the current displacement parameter of the magnetic bearing rotor.

[0014] In some embodiments, the control unit takes the current control current of the magnetic bearing coil as an input parameter, inputs it into the displacement soft measurement model, and determines the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, including: setting the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil.

[0015] In some embodiments, the modeling unit constructs a displacement measurement model based on a wavelet neural network, recorded as a displacement soft measurement model, including: collecting the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor as sample data; based on the sample data, with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter, training the wavelet neural network to obtain the displacement soft measurement model.

[0016] In some embodiments, the number of the magnetic bearing coils is more than one; the modeling unit collects the set control current of the magnetic bearing coils and the set displacement parameters of the magnetic bearing rotor as sample data, including: selecting the set control current of more than one magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor as characteristic variables; normalizing the characteristic variables to form sampling data, which is used as sample data for training and testing.

[0017] In some embodiments, the modeling unit, based on the sample data, trains a wavelet neural network with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model, including: setting a wavelet neural network model; when the number of the magnetic bearing coils is more than one, the number of layers of the wavelet neural network model is more than one; based on a particle swarm optimization algorithm, optimizing the set parameters of the wavelet neural network model to obtain the optimized wavelet neural network model; based on the sample data, training the optimized wavelet neural network model with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model.

[0018] In some embodiments, it also includes: the control unit is also configured to adjust the current control current of the magnetic bearing coil according to the current displacement parameter of the magnetic bearing rotor after determining the current displacement parameter of the magnetic bearing rotor, so as to obtain the adjustment value of the current control current of the magnetic bearing coil; the control unit is also configured to control the operation of the magnetic bearing coil according to the adjustment value of the current control current of the magnetic bearing coil.

[0019] Matching the above-mentioned device, the present invention further provides a magnetic bearing system on the other hand, including: the displacement determination device of the magnetic bearing described above.

[0020] In accordance with the above method, the present invention provides a storage medium on the other hand, wherein the storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the above-mentioned method for determining the displacement of the magnetic bearing.

[0021] In accordance with the above method, the present invention provides a processor on another aspect, wherein the processor is used to run a program, wherein the above method for determining the displacement of a magnetic bearing is executed when the program is run.

[0022] Therefore, the scheme of the present invention trains the model through the collected sample data, and the trained model can be directly connected in series to the magnetic suspension bearing system to form a radial displacement prediction module, so as to realize the self-detection of the radial displacement of the magnetic suspension bearing rotor; thus, by using the soft measurement method to measure the radial displacement of the magnetic bearing rotor, it is possible to avoid the magnetic bearing rotor falling accident caused by the damage of the eddy current sensor when using the eddy current sensor, and the suspension safety of the magnetic bearing rotor can be improved. At the same time, the scheme of the present invention can also solve the problem that the eddy current sensor is often used when measuring the radial displacement of the magnetic bearing rotor, which leads to the increase in the volume of the magnetic suspension bearing control system, and achieve the effect of measuring the radial displacement of the magnetic bearing rotor by using the soft measurement method, avoiding the increase in the volume of the magnetic suspension bearing control system caused by the use of the eddy current sensor, thereby saving the space occupied by the magnetic suspension bearing control system.

[0023] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the present invention.

[0024] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic flow chart of an embodiment of a method for determining displacement of a magnetic bearing of the present invention;

[0026] Figure 2 A schematic diagram of a flow chart of an embodiment of constructing a displacement measurement model based on a wavelet neural network in the method of the present invention;

[0027] Figure 3 A schematic flow chart of an embodiment of collecting the set control current of the magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor in the method of the present invention;

[0028] Figure 4 A schematic diagram of a flow chart of an embodiment of training a wavelet neural network in the method of the present invention;

[0029] Figure 5 It is a flow chart of an embodiment of controlling the current of the magnetic bearing coil based on the determined current displacement parameter of the magnetic bearing rotor in the method of the present invention;

[0030] Figure 6 It is a structural schematic diagram of an embodiment of a displacement determination device of a magnetic bearing of the present invention;

[0031] Figure 7 A schematic diagram of a process for constructing a wavelet neural network displacement soft-sensing model according to an embodiment;

[0032] Figure 8 A schematic diagram of the structure of an embodiment of a wavelet neural network structure;

[0033] Fig. 9 A flowchart of an embodiment of a process for improving particle swarm optimization of wavelet neural network;

[0034] Fig.10 It is a structural schematic diagram of an embodiment of a displacement measurement circuit based on a wavelet neural network.

[0035] In conjunction with the accompanying drawings, the reference numerals in the embodiments of the present invention are as follows:

[0036] 102 - modeling unit; 106 - acquisition unit; 104 - control unit. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding 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 making creative work are within the scope of protection of the present invention.

[0038] Considering that eddy current sensors are often used when measuring the radial displacement of magnetic bearing rotors, since eddy current sensors need to occupy space, the volume of the magnetic bearing control system becomes larger, and the cost of the magnetic bearing control system increases, and it is inconvenient to install and maintain. In related schemes, a magnetic bearing displacement self-detection method is proposed, which mainly includes a model reference adaptive method and a support vector machine method. For example: in some schemes, there is no need to rely on an accurate mathematical model, which improves the dynamic performance of the system, but the stability, learning ability and generalization ability of the support vector machine mainly depend on the setting of its parameters, and the algorithm has a large amount of calculation, which easily leads to slow network convergence and falling into local minima. For example: in other schemes, a soft measurement scheme is established based on an adaptive BP neural network (i.e., a multi-layer feedforward neural network trained according to an error back propagation algorithm). Since the transfer excitation function of the hidden layer of the BP neural network is a tansig function (i.e., a neural network layer transfer function), it does not have the good time-frequency localization property of the wavelet function, so it is easy to make the BP neural network fall into local optimization, resulting in poor prediction accuracy and robustness of the network.

[0039] According to an embodiment of the present invention, a method for determining the displacement of a magnetic bearing is provided, such as Figure 1 The flowchart of an embodiment of the method of the present invention is shown as follows. The magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor; the displacement determination method of the magnetic bearing comprises: steps S110 to S130.

[0040] At step S110, a displacement measurement model based on a wavelet neural network is constructed, which is recorded as a displacement soft measurement model, i.e., a wavelet neural network displacement prediction model. The displacement soft measurement model is a model obtained by training a wavelet neural network with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter.

[0041] In some implementations, the specific process of constructing a displacement measurement model based on a wavelet neural network in step S110, which is recorded as a displacement soft measurement model, is described in the following exemplary embodiment.

[0042] Combine the following Figure 2 The flowchart of an embodiment of constructing a displacement measurement model based on a wavelet neural network in the method of the present invention further illustrates the specific process of constructing the displacement measurement model based on a wavelet neural network in step S110, including: step S210 and step S220.

[0043] Step S210, collecting the set control current of the magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor as sample data. In the set displacement parameters of the magnetic bearing rotor, the displacement parameters include: radial displacement of the magnetic bearing rotor.

[0044] In some embodiments, the number of the magnetic bearing coil is more than one.

[0045] For the specific process of collecting the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor as sample data in step S210, please refer to the following exemplary description.

[0046] Combine the following Figure 3 The flowchart of an embodiment of collecting the set control current of the magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor in the method of the present invention is shown, which further illustrates the specific process of collecting the set control current of the magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor in step S210, including: step S310 to step S320.

[0047] Step S310, selecting one or more set control currents of the magnetic bearing coils and set displacement parameters of the magnetic bearing rotor as characteristic variables.

[0048] Step S320 normalizes the feature variables to form sampling data, which is used as sample data for training and testing.

[0049] The solution of the present invention is to use the algorithm of "improved particle swarm optimization wavelet neural network" to establish a nonlinear prediction model between the displacement and current of the magnetic bearing, which can realize the displacement self-detection of the magnetic bearing without a displacement sensor. The model is established by using formula (1) to formula (6), which can effectively avoid the problem of the particle swarm falling into the local optimum, reduce the number of iterations, and improve the prediction accuracy of the model.

[0050] Figure 7 The following is a flow chart of an embodiment of constructing a wavelet neural network displacement soft-sensing model. Figure 7 As shown, the bearing displacement soft measurement method includes:

[0051] The first step: select characteristic variables. The characteristic variables of the displacement soft measurement model (i.e., the displacement measurement model based on wavelet neural network) are mainly composed of model input variables and model output variables. FX 、s FY 、s RX 、s RY is the model output variable. Through model analysis, it can be known that the suspension force controls the current i FX 、i FY 、i RX 、i RY It is associated with the bearing displacement, so it is determined as the model input variable.

[0052] Step 2: Sampling data preprocessing. Normalize the input and output signals. The normalization formula is as follows:

[0053] Among them, A K is the normalized input data. K is the original input data. max 、x min are the maximum and minimum values ​​of the original input data respectively. Then filter the normalized data to obtain more accurate input and output data, and select N groups of representative sample data, where N is a positive integer. Each group of sample data contains 4 input variables X i =[i FX 、i FY 、i RX 、i RY ], the corresponding output vector is Y k =[s FX 、s FY 、s RX 、s RY ] to form sampling data for training and testing the model.

[0054] Step S220: Based on the sample data, the wavelet neural network is trained with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model.

[0055] Specifically, the scheme of the present invention proposes a method for soft measurement of magnetic bearing rotor displacement based on a wavelet neural network model optimized by particle swarm optimization. The wavelet neural network model is trained by using collected sample data to obtain a radial displacement prediction module (i.e., a displacement soft measurement model), thereby realizing a sensorless radial displacement self-detection technology, i.e., being able to self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor.

[0056] In some embodiments, in step S220, based on the sample data, the wavelet neural network is trained with the set control current of the magnetic bearing coil as the input parameter and the set displacement parameter of the magnetic bearing rotor as the output parameter to obtain the specific process of the displacement soft measurement model. Please refer to the following exemplary description.

[0057] Combine the following Figure 4 The flowchart of an embodiment of training the wavelet neural network in the method of the present invention further illustrates the specific process of training the wavelet neural network in step S220, including: step S410 to step S430.

[0058] Step S410, setting a wavelet neural network model. When the number of the magnetic bearing coil is more than one, the number of layers of the wavelet neural network model is more than one.

[0059] Step S420, based on the particle swarm optimization algorithm, the parameters of the set wavelet neural network model are optimized to obtain the optimized wavelet neural network model.

[0060] like Figure 7 As shown, the bearing displacement soft measurement method also includes:

[0061] Step 3: Design wavelet neural network. Figure 8 The wavelet neural network model designed is as follows: Figure 8 As shown, there are three layers in total, which is a multi-input and multi-output model. The input layer variable of the network is X i =[i FX 、i FY 、i RX 、i RY ], so the number of input layer nodes is i (i = 1, 2, 3, 4). The output layer variable is Y k =[s FX 、sFY 、s RX 、s RY ], so the number of output layer nodes is k (k = 1, 2, 3, 4). The number of hidden layer nodes is . Figure 8 In the example shown, the design uses the Morlet (wavelet) mother wavelet function ψ(x) as the activation function of the hidden layer, and its formula is as follows:

[0062] The output layer activation function is the Sigmoid function, and its formula is as follows:

[0063] If the i-th input sample of the input layer is set to X i , the kth output value of the output layer is set to Y k , the connection weight between input layer node i and hidden layer node j is set to ω ij , the connection weight between hidden layer node j and output layer node k is set to ω jk The scaling factor and translation factor of the jth hidden layer node are set to a j and b j , then the mathematical model of wavelet neural network is:

[0064]

[0065] Among them, σ and u k It is the activation function of the output layer of the wavelet neural network Since the number of nodes in the output layer is K (k = 1, 2, 3, 4), it can be written as σ(u k ), which represents the output value of the Kth node in the output layer.

[0066] from Figure 8 The structure of the wavelet neural network shown in the figure shows that when the number of nodes in each layer is determined, the connection weights (ω ij ,ω jk ) and the scaling factor (a j ), translation factor (b j ) determines the performance of the wavelet neural network. Therefore, (ω ij ,ω jk ,a j ,b j )It is particularly important to optimize these parameters.

[0067] Step 4: Optimize the parameters of the wavelet neural network. The main function of the particle swarm optimization algorithm (PSO) is to find the global optimal solution. The particle swarm algorithm can be described as follows: Assume that the particle swarm searches in an s-dimensional space, and there are m particles forming a population Z, which can be expressed as Z = (Z 1 , Z 2 , Z 3 , …, Z m ). Let the position of the i-th particle be Z i =(Z i1 , Z i2 , …, Z is ), the velocity of the ith particle is represented by V i =(V i1 , V i2 , …, V is ), the position of each particle represents a potential solution to the problem. First, the objective function is used to calculate the fitness value, and then the performance of the current particle is calculated based on the fitness value of the particle. The particle continuously iterates to find a new solution according to the position update formula, and the individual extreme value is recorded as P id , the total extreme value is recorded as P gd Each particle updates its velocity according to formula (1) and updates its position according to formula (2).

[0068] V id (t+1)=ωV id (t)+η 1 r 1 (P id -Z id (t))+η 2 r 2 (P gd -Z id (t)) (1).

[0069] Z id (t+1)=Z id (t)+V id (t+1) (2).

[0070] Where ω is the inertia weight coefficient. η 1 and η 2 is a non-negative learning factor. i = 1, 2, ..., m. d = 1, 2, ..., s. r1 and r2 are random numbers on [0, 1].

[0071] Improved inertia weight coefficient: The inertia weight coefficient ω is closely related to the global and local search capabilities of the particle swarm. Generally speaking, the larger the value of ω, the stronger the global search capability of the particle swarm, and the smaller the value of ω, the stronger the local search capability of the particle swarm. Therefore, a larger ω is used to enhance the global search capability of the particle swarm in the early stage of iteration, and a smaller ω is used to enhance the local search of the particle swarm in the later stage of iteration, so as to improve the group iteration speed and iteration accuracy.

[0072] Therefore, in the solution of the present invention, formula (3) is used to update the particle swarm inertia weight coefficient.

[0073]

[0074] Among them, ω min and ω max are the minimum and maximum values ​​of the inertia weight change, t is the current iteration number, t max is the maximum number of iterations set. For example: in formula (3) min ,ω max ,t max The values ​​can be 0.2, 0.9, and 500 respectively.

[0075] Improved learning factor: Due to η 1 Determines the contribution rate of individual cognition level of particle swarm, η 2 Determines the contribution rate of the particle group's cognitive level. Therefore, in the early stage of iteration, the particle fitness is large, which can be achieved by a large η 2 To control the particles to develop in the optimal direction of the group. In the later iteration, the fitness of the particles gradually decreases, which can be controlled by a larger η 1 Release the individual cognition of particles until the optimal position is found. In view of this, the solution of the present invention adopts the following formula (4) and formula (5) to update the learning factor.

[0076]

[0077]

[0078] Among them, η 1min , η 2min , η 1max and η 2max , respectively represent the minimum and maximum values ​​of the learning factor. By improving the inertia weight and learning factor of the particle swarm algorithm, not only can the number of iterations of the algorithm be reduced, the calculation time can be reduced, but also the algorithm accuracy can be improved. This improved model can effectively avoid the problem of the particle swarm falling into the local optimum and improve the prediction accuracy of the particle swarm algorithm.

[0079] For example: In formula (4), η 1min , η 1max ,t maxThe values ​​of are 1.7, 2.3, and 500 respectively. In formula (5), η 2min , η 2max ,t max The values ​​of are also 1.7, 2.3, and 500. The selection of these parameters can reduce the number of iterations of the algorithm and reduce the calculation time by adjusting the development direction of the particle swarm at different stages.

[0080] Particle swarm optimization of wavelet neural network is to search the parameters (ω) of wavelet neural network in its search space through each particle in the network back propagation process. ij ,ω jk ,a j ,b j ). The position of the particle in each dimension is the solution we are looking for. We need to establish the dimension of the particle and (ω ij ,ω jk ,a j ,b j ), the weight of the neural network (ω ij ,ω jk ) plus the expansion factor (a) of the wavelet basis function j ) number and translation factor (b j ) is equal to the dimension of the particle. For a three-layer network with a network structure of MnN, the number of network weights is M×n+n×N, and the number of scaling factors and translation factors of the wavelet basis function is n+n, so the dimension of the particle is M×n+n×N+n+n. The mean square error function E is selected as the fitness value function, as shown in formula (6):

[0081]

[0082] Here, M is the number of nodes in the input layer of the wavelet neural network, n is the number of nodes in the hidden layer of the wavelet neural network, and N is the number of nodes in the output layer of the wavelet neural network. The aforementioned N is the number of sample data.

[0083] Step S430, based on the sample data, with the set control current of the magnetic bearing coil as the input parameter and the set displacement parameter of the magnetic bearing rotor as the output parameter, the optimized wavelet neural network model is trained to obtain the displacement soft measurement model.

[0084] Fig. 9 The figure is a flow chart of an embodiment of the process of improving particle swarm optimization of wavelet neural network. The specific process of the improved particle swarm optimization wavelet neural network training algorithm is as follows: Fig. 9 As shown. Assign the value of each dimension of the optimized particle global optimal position to (ω ij ,ω jk ,a j ,b j), for a three-layer network with a network structure of MnN, the first M×n values ​​are generally assigned to the weights between the input layer and the hidden layer, the values ​​between the M×n+1th and M×n+nth are assigned to the scaling factors of the wavelet basis functions, the values ​​between the M×n+n+1th and M×n+2nth are assigned to the translation factors of the wavelet basis functions, and the remaining values ​​are assigned to the weights between the hidden layer and the output layer.

[0085] like Figure 7 As shown, the bearing displacement soft measurement method also includes:

[0086] Step 5: Establish a soft-measurement control system for displacement based on improved particle swarm optimization wavelet neural network. Connect the optimized and trained wavelet neural network radial displacement prediction module to the magnetic bearing control system to achieve online real-time prediction of the radial displacement of the bearing rotor.

[0087] At step S120, when the current displacement parameter of the magnetic bearing rotor needs to be determined, the current control current of the magnetic bearing coil is obtained, such as obtaining the current control current of the magnetic bearing coil sampled by a current sensor.

[0088] At step S130, the current control current of the magnetic bearing coil is used as an input parameter and input into the displacement soft measurement model, and the output parameter of the displacement soft measurement model is determined as the current displacement parameter of the magnetic bearing rotor, so as to determine the current displacement parameter of the magnetic bearing rotor. That is, the current displacement parameter of the magnetic bearing rotor is determined according to the current control circuit of the magnetic bearing coil. Among the current displacement parameters of the magnetic bearing rotor, the displacement parameters include: radial displacement of the magnetic suspension bearing rotor.

[0089] During specific implementation, it is also possible to first construct a correspondence between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor. In the case where the current displacement parameter of the magnetic bearing rotor needs to be determined, the current control current of the magnetic bearing coil is obtained. The set displacement parameter corresponding to the set control current that is the same as the current control current of the magnetic bearing coil in the corresponding relationship is determined as the current displacement parameter of the magnetic bearing rotor corresponding to the current control current of the magnetic bearing coil, that is, the current displacement parameter of the magnetic bearing rotor is determined according to the current control circuit of the magnetic bearing coil. Among the set displacement parameters of the magnetic bearing rotor and the current displacement parameters of the magnetic bearing rotor, the displacement parameters include: radial displacement of the magnetic bearing rotor.

[0090] Wherein, constructing the correspondence between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor may include at least one of the following construction situations:

[0091] The first construction situation: it can be to construct a numerical relationship or table between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor.

[0092] The second construction situation: it can be to construct a displacement measurement model based on wavelet neural network, which is recorded as a displacement soft measurement model, that is, a wavelet neural network displacement prediction model. The displacement soft measurement model is a model that takes the control current of the magnetic bearing coil as an input parameter and the displacement parameter of the magnetic bearing rotor as an output parameter.

[0093] In this way, the scheme of the present invention realizes the sensorless radial displacement self-detection technology by self-detecting the radial displacement of the magnetic suspension bearing rotor, that is, it can self-detect the radial displacement of the magnetic suspension bearing rotor without the need for a mechanical eddy current sensor, solving the problem of the rotor suddenly falling when the mechanical sensor is damaged and the friction with the mechanical part damaging the system, thereby avoiding the magnetic bearing rotor falling accident caused by the eddy current sensor being damaged when the eddy current sensor is used, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it reduces the system cost and volume, improves the stability of the magnetic suspension bearing control system, and facilitates the later maintenance.

[0094] In some embodiments, in step S130, the current control current of the magnetic bearing coil is used as an input parameter and input into the displacement soft measurement model, and the output parameter of the displacement soft measurement model is determined as the current displacement parameter of the magnetic bearing rotor, including: setting the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil.

[0095] Thus, the scheme of the present invention proposes a method for soft measurement of magnetic bearing rotor displacement based on a wavelet neural network model optimized by particle swarm optimization. The wavelet neural network model is trained by the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module (i.e., a displacement soft measurement model), and realizes a sensorless radial displacement self-detection technology, that is, it can self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor, and solves the problem of the rotor suddenly falling when the mechanical sensor is damaged, and the friction with the mechanical part damaging the system. Thus, it can avoid the magnetic bearing rotor falling accident caused by the eddy current sensor being damaged when the eddy current sensor is used, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it also solves the problem of large volume and high cost of the magnetic bearing control system, and can avoid the problem of using eddy current sensors when measuring the radial displacement of the magnetic bearing rotor, which leads to the problem of increasing the volume of the magnetic bearing control system. In addition, for magnetic bearings that are nonlinear and difficult to establish a rotor displacement calculation model, automatic measurement without displacement sensors is realized, and no additional complex circuit is required, which is easy to implement.

[0096] In some implementations, the method further includes: controlling the current of the magnetic bearing coil based on the determined current displacement parameter of the magnetic bearing rotor.

[0097] Combine the following Figure 5 The method of the present invention is shown as a flow chart of an embodiment of controlling the current of the magnetic bearing coil based on the determined current displacement parameters of the magnetic bearing rotor, further illustrating the specific process of controlling the current of the magnetic bearing coil based on the determined current displacement parameters of the magnetic bearing rotor, including: step S510 and step S520.

[0098] Step S510, after the current displacement parameter of the magnetic bearing rotor is determined, the current control current of the magnetic bearing coil is adjusted according to the current displacement parameter of the magnetic bearing rotor to obtain the adjustment value of the current control current of the magnetic bearing coil. Specifically, the current control current of the magnetic bearing coil is adjusted according to the current displacement parameter of the magnetic bearing rotor, including: performing PID control according to the difference between the current displacement parameter of the magnetic bearing rotor and the reference displacement parameter of the magnetic bearing rotor to obtain a first processing result. Then, the difference between the first processing result and the current control current of the magnetic bearing coil is subjected to PWM processing to obtain a second processing result. Then, the second processing result is subjected to power amplification processing to obtain the adjustment value of the current control current of the magnetic bearing coil.

[0099] Step S520: Control the operation of the magnetic bearing coil according to the adjustment value of the current control current of the magnetic bearing coil.

[0100] Fig.10 FIG. 1 is a schematic diagram of the structure of an embodiment of a displacement measurement circuit based on a wavelet neural network. Fig.10 As shown, the displacement measurement circuit based on wavelet neural network includes: for each magnetic bearing coil, such as radial front X direction coil (FX coil), radial front Y direction coil (FY coil), radial rear X direction coil (RX coil) and radial rear Y direction coil (RY coil), a control branch is set. The current signal output by each control branch, that is, the control current of each magnetic bearing coil, is output after passing through the wavelet neural network (WNN) displacement measurement model (that is, the displacement measurement model based on wavelet neural network), and the predicted rotor displacement of the magnetic bearing rotor is output. The difference between each predicted rotor displacement and the given reference position signal of the path is obtained by the first comparator, and after passing through the control branch of the path, the adjustment value of the control current of the magnetic bearing coil of the path is output, and the magnetic bearing coil of the path is controlled according to the adjustment value of the control current of the magnetic bearing coil of the path. Among them, each control branch includes: a PWM control module, a second comparator, a PWM module and a power amplifier. The PWM control module performs PID processing on the difference between each predicted rotor displacement and the given reference position signal of the path, and outputs a first processing result. The difference between the first processing result of each path and the predicted rotor displacement of the path is processed by the PWM module to obtain the second processing result. The second processing result is processed by the power amplifier to obtain the adjustment value of the control current of the magnetic bearing coil of the path.

[0101] exist Fig.10 In the example shown, the displacement measurement model based on wavelet neural network is applied in the magnetic bearing control system. The wavelet neural network displacement prediction model uses the current signal i of the power amplifier (such as a switching power amplifier) FX 、i FY 、i RX 、i RY is the input signal, and the output is the predicted rotor displacement s FX 、s FY 、s RX 、s RY and compare it with the given reference position signal f FX 、f FY 、f RX 、f RY For comparison, the difference is passed through a closed-loop controller and a switching power amplifier, ultimately achieving stable suspension of the magnetic bearing.

[0102] Thus, the scheme of the present invention proposes a method for soft measurement of magnetic bearing rotor displacement based on a wavelet neural network model optimized by particle swarm optimization. The wavelet neural network model is trained by the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module (i.e., a displacement soft measurement model), and realizes a sensorless radial displacement self-detection technology, that is, it can self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor, and solves the problem of the rotor suddenly falling when the mechanical sensor is damaged, and the friction with the mechanical part damaging the system. Thus, it can avoid the magnetic bearing rotor falling accident caused by the eddy current sensor being damaged when the eddy current sensor is used, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it also solves the problem of large volume and high cost of the magnetic bearing control system, and can avoid the problem of using eddy current sensors when measuring the radial displacement of the magnetic bearing rotor, which leads to the problem of increasing the volume of the magnetic bearing control system. In addition, for magnetic bearings that are nonlinear and difficult to establish a rotor displacement calculation model, automatic measurement without displacement sensors is realized, and no additional complex circuit is required, which is easy to implement.

[0103] It should be noted that the wavelet neural network-based displacement soft measurement method in the solution of the present invention can be used not only in magnetic bearing systems, but also in other application scenarios of eddy current sensors.

[0104] After a large number of experimental verifications, the technical solution of this embodiment is adopted, and the model is trained by the collected sample data. The trained model can be directly connected in series to the magnetic suspension bearing system to form a radial displacement prediction module to realize self-detection of the radial displacement of the magnetic suspension bearing rotor. Therefore, by using the soft measurement method to measure the radial displacement of the magnetic bearing rotor, it is possible to avoid the magnetic bearing rotor falling accident caused by the damage of the eddy current sensor when using the eddy current sensor, and the suspension safety of the magnetic bearing rotor can be improved. At the same time, the solution of the present invention can also solve the problem that when measuring the radial displacement of the magnetic bearing rotor, the eddy current sensor is often used, which leads to the increase in the volume of the magnetic suspension bearing control system, so as to achieve the effect of measuring the radial displacement of the magnetic bearing rotor by using the soft measurement method, avoiding the increase in the volume of the magnetic suspension bearing control system caused by the use of the eddy current sensor, thereby saving the space occupied by the magnetic suspension bearing control system.

[0105] According to an embodiment of the present invention, a displacement determination device for a magnetic bearing corresponding to the displacement determination method for a magnetic bearing is also provided. Figure 6 The structural schematic diagram of an embodiment of the device of the present invention is shown. The magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor. The displacement determination device of the magnetic bearing comprises: a modeling unit 102, an acquisition unit 104 and a control unit 106.

[0106] The modeling unit 102 is configured to construct a displacement measurement model based on a wavelet neural network, which is referred to as a displacement soft measurement model, i.e., a wavelet neural network displacement prediction model. The displacement soft measurement model is a model obtained by training the wavelet neural network with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter. The specific functions and processing of the modeling unit 102 refer to step S110.

[0107] In some implementations, the modeling unit 102 constructs a displacement measurement model based on a wavelet neural network, referred to as a displacement soft measurement model, including:

[0108] The modeling unit 102 is further configured to collect the set control current of the magnetic bearing coil and the set displacement parameters of the magnetic bearing rotor as sample data. In the set displacement parameters of the magnetic bearing rotor, the displacement parameters include: radial displacement of the magnetic bearing rotor. The specific functions and processing of the modeling unit 102 are also referred to step S210.

[0109] In some embodiments, the number of the magnetic bearing coil is more than one.

[0110] The modeling unit 102 collects the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor as sample data, including:

[0111] The modeling unit 102 is further configured to select one or more set control currents of the magnetic bearing coils and set displacement parameters of the magnetic bearing rotor as characteristic variables. The specific functions and processing of the modeling unit 102 are also shown in step S310.

[0112] The modeling unit 102 is further configured to perform normalization processing on the feature variables to form sampling data, which is used as sample data for training and testing. The specific functions and processing of the modeling unit 102 are also shown in step S320.

[0113] Figure 7 The following is a flow chart of an embodiment of constructing a wavelet neural network displacement soft-sensing model. Figure 7 As shown, the bearing displacement soft measurement device comprises:

[0114] The first step: select characteristic variables. The characteristic variables of the displacement soft measurement model (i.e., the displacement measurement model based on wavelet neural network) are mainly composed of model input variables and model output variables. FX 、s FY 、s RX 、s RY is the model output variable. Through model analysis, it can be known that the suspension force controls the current iFX 、i FY 、i RX 、i RY It is associated with the bearing displacement, so it is determined as the model input variable.

[0115] Step 2: Sampling data preprocessing. Normalize the input and output signals. The normalization formula is as follows:

[0116] Among them, A K is the normalized input data. K is the original input data. max 、x min are the maximum and minimum values ​​of the original input data respectively. Then filter the normalized data to obtain more accurate input and output data, and select N groups of representative sample data, where N is a positive integer. Each group of sample data contains 4 input variables X i =[i FX 、i FY 、i RX 、i RY ], the corresponding output vector is Y k =[s FX 、s FY 、s RX 、s RY ] to form sampling data for training and testing the model.

[0117] The modeling unit 102 is further configured to train the wavelet neural network based on the sample data, with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter, to obtain the displacement soft measurement model. The specific functions and processing of the modeling unit 102 are also shown in step S220.

[0118] Specifically, the solution of the present invention proposes a device for soft measurement of magnetic bearing rotor displacement based on a wavelet neural network model optimized by particle swarm optimization. The wavelet neural network model is trained by using collected sample data to obtain a radial displacement prediction module (i.e., a displacement soft measurement model), thereby realizing a sensorless radial displacement self-detection technology, i.e., being able to self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor.

[0119] In some embodiments, the modeling unit 102, based on the sample data, takes the set control current of the magnetic bearing coil as an input parameter and takes the set displacement parameter of the magnetic bearing rotor as an output parameter, trains a wavelet neural network to obtain the displacement soft measurement model, including:

[0120] The modeling unit 102 is further configured to set a wavelet neural network model. When the number of the magnetic bearing coils is more than one, the wavelet neural network model has more than one layer. The specific functions and processing of the modeling unit 102 are also shown in step S410.

[0121] The modeling unit 102 is further configured to optimize the set parameters of the wavelet neural network model based on a particle swarm optimization algorithm to obtain the optimized wavelet neural network model. The specific functions and processing of the modeling unit 102 are also shown in step S420.

[0122] like Figure 7 As shown, the bearing displacement soft measurement device also includes:

[0123] Step 3: Design wavelet neural network. Figure 8 The wavelet neural network model designed is as follows: Figure 8 As shown, there are three layers in total, which is a multi-input and multi-output model. The input layer variable of the network is X i =[i FX 、i FY 、i RX 、i RY ], so the number of input layer nodes is i (i = 1, 2, 3, 4). The output layer variable is Y k =[s FX 、s FY 、s RX 、s RY ], so the number of output layer nodes is k (k = 1, 2, 3, 4). The number of hidden layer nodes is . Figure 8 In the example shown, the design uses the Morlet (wavelet) mother wavelet function ψ(x) as the activation function of the hidden layer, and its formula is as follows:

[0124] The output layer activation function is the Sigmoid function, and its formula is as follows:

[0125] If the i-th input sample of the input layer is set to X i , the kth output value of the output layer is set to Y k , the connection weight between input layer node i and hidden layer node j is set to ω ij , the connection weight between hidden layer node j and output layer node k is set to ω jk The scaling factor and translation factor of the jth hidden layer node are set to a j and b j , then the mathematical model of wavelet neural network is:

[0126]

[0127] from Figure 8 The structure of the wavelet neural network shown in the figure shows that when the number of nodes in each layer is determined, the connection weights (ω ij ,ω jk ) and the scaling factor (a j ), translation factor (b j ) determines the performance of the wavelet neural network. Therefore, (ω ij ,ω jk ,a j ,b j )It is particularly important to optimize these parameters.

[0128] Step 4: Optimize the parameters of the wavelet neural network. The main function of the particle swarm optimization algorithm (PSO) is to find the global optimal solution. The particle swarm algorithm can be described as follows: Assume that the particle swarm searches in an s-dimensional space, and there are m particles forming a population Z, which can be expressed as Z = (Z 1 , Z 2 , Z 3 , …, Z m ). Let the position of the i-th particle be Z i =(Z i1 , Z i2 , …, Z is ), the velocity of the ith particle is represented by V i =(V i1 , V i2 , …, V is ), the position of each particle represents a potential solution to the problem. First, the objective function is used to calculate the fitness value, and then the performance of the current particle is calculated based on the fitness value of the particle. The particle continuously iterates to find a new solution according to the position update formula, and the individual extreme value is recorded as P id , the total extreme value is recorded as P gd Each particle updates its velocity according to formula (1) and updates its position according to formula (2).

[0129] V id (t+1)=ωV id (t)+η 1 r 1 (P id -Z id (t))+η 2 r 2 (P gd -Z id (t)) (1).

[0130] Z id (t+1)=Z id (t)+V id (t+1) (2).

[0131] Where ω is the inertia weight coefficient. η 1 and η 2 is a non-negative learning factor. i = 1, 2, ..., m. d = 1, 2, ..., s. r1 and r2 are random numbers on [0, 1].

[0132] Improved inertia weight coefficient: The inertia weight coefficient ω is closely related to the global and local search capabilities of the particle swarm. Generally speaking, the larger the value of ω, the stronger the global search capability of the particle swarm, and the smaller the value of ω, the stronger the local search capability of the particle swarm. Therefore, a larger ω is used to enhance the global search capability of the particle swarm in the early stage of iteration, and a smaller ω is used to enhance the local search of the particle swarm in the later stage of iteration, so as to improve the group iteration speed and iteration accuracy.

[0133] Therefore, in the solution of the present invention, formula (3) is used to update the particle swarm inertia weight coefficient.

[0134]

[0135] Among them, ω min and ω max are the minimum and maximum values ​​of the inertia weight change, t is the current iteration number, t max The maximum number of iterations set.

[0136] Improved learning factor: Due to η 1 Determines the contribution rate of individual cognition level of particle swarm, η 2 Determines the contribution rate of the particle group's cognitive level. Therefore, in the early stage of iteration, the particle fitness is large, which can be achieved by a large η 2 To control the particles to develop in the optimal direction of the group. In the later iteration, the fitness of the particles gradually decreases, which can be controlled by a larger η 1 Release the individual cognition of particles until the optimal position is found. In view of this, the solution of the present invention adopts the following formula (4) and formula (5) to update the learning factor.

[0137]

[0138]

[0139] Among them, η 1min , η 2min , η 1max and η 2max, respectively represent the minimum and maximum values ​​of the learning factor. By improving the inertia weight and learning factor of the particle swarm algorithm, not only can the number of iterations of the algorithm be reduced, the calculation time can be reduced, but also the algorithm accuracy can be improved. This improved model can effectively avoid the problem of the particle swarm falling into the local optimum and improve the prediction accuracy of the particle swarm algorithm.

[0140] Particle swarm optimization of wavelet neural network is to search the parameters (ω) of wavelet neural network in its search space through each particle in the network back propagation process. ij ,ω jk ,a j ,b j ). The position of the particle in each dimension is the solution we are looking for. We need to establish the dimension of the particle and (ω ij ,ω jk ,a j ,b j ), the weight of the neural network (ω ij ,ω jk ) plus the expansion factor (a) of the wavelet basis function j ) number and translation factor (b j ) is equal to the dimension of the particle. For a three-layer network with a network structure of MnN, the number of network weights is M×n+n×N, and the number of scaling factors and translation factors of the wavelet basis function is n+n, so the dimension of the particle is M×n+n×N+n+n. The mean square error function E is selected as the fitness value function, as shown in formula (6):

[0141]

[0142] The modeling unit 102 is further configured to train the optimized wavelet neural network model based on the sample data, with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter, to obtain the displacement soft measurement model. The specific functions and processing of the modeling unit 102 are also shown in step S430.

[0143] Fig. 9 The figure is a flow chart of an embodiment of the process of improving particle swarm optimization of wavelet neural network. The specific process of the improved particle swarm optimization wavelet neural network training algorithm is as follows: Fig. 9 As shown. Assign the value of each dimension of the optimized particle global optimal position to (ω ij ,ω jk ,a j ,b j), for a three-layer network with a network structure of MnN, the first M×n values ​​are generally assigned to the weights between the input layer and the hidden layer, the values ​​between the M×n+1th and M×n+nth are assigned to the scaling factors of the wavelet basis functions, the values ​​between the M×n+n+1th and M×n+2nth are assigned to the translation factors of the wavelet basis functions, and the remaining values ​​are assigned to the weights between the hidden layer and the output layer.

[0144] like Figure 7 As shown, the bearing displacement soft measurement device also includes:

[0145] Step 5: Establish a soft-measurement control system for displacement based on improved particle swarm optimization wavelet neural network. Connect the optimized and trained wavelet neural network radial displacement prediction module to the magnetic bearing control system to achieve online real-time prediction of the radial displacement of the bearing rotor.

[0146] The acquisition unit 104 is configured to acquire the current control current of the magnetic bearing coil when the current displacement parameter of the magnetic bearing rotor needs to be determined, such as acquiring the current control current of the magnetic bearing coil sampled by the current sensor. The specific function and processing of the acquisition unit 104 refer to step S120.

[0147] The control unit 106 is configured to use the current control current of the magnetic bearing coil as an input parameter, input it into the displacement soft measurement model, and determine the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, so as to determine the current displacement parameter of the magnetic bearing rotor. That is, the current displacement parameter of the magnetic bearing rotor is determined according to the current control circuit of the magnetic bearing coil. Among the current displacement parameters of the magnetic bearing rotor, the displacement parameters include: the radial displacement of the magnetic bearing rotor. For the specific functions and processing of the control unit 106, see step S130.

[0148] During specific implementation, it is also possible to first construct a correspondence between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor. In the case where the current displacement parameter of the magnetic bearing rotor needs to be determined, the current control current of the magnetic bearing coil is obtained. The set displacement parameter corresponding to the set control current that is the same as the current control current of the magnetic bearing coil in the corresponding relationship is determined as the current displacement parameter of the magnetic bearing rotor corresponding to the current control current of the magnetic bearing coil, that is, the current displacement parameter of the magnetic bearing rotor is determined according to the current control circuit of the magnetic bearing coil. Among the set displacement parameters of the magnetic bearing rotor and the current displacement parameters of the magnetic bearing rotor, the displacement parameters include: radial displacement of the magnetic bearing rotor.

[0149] Wherein, constructing the correspondence between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor may include at least one of the following construction situations:

[0150] The first construction situation: it can be to construct a numerical relationship or table between the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor.

[0151] The second construction situation: it can be to construct a displacement measurement model based on wavelet neural network, which is recorded as a displacement soft measurement model, that is, a wavelet neural network displacement prediction model. The displacement soft measurement model is a model that takes the control current of the magnetic bearing coil as an input parameter and the displacement parameter of the magnetic bearing rotor as an output parameter.

[0152] In this way, the scheme of the present invention realizes the sensorless radial displacement self-detection technology by self-detecting the radial displacement of the magnetic suspension bearing rotor, that is, it can self-detect the radial displacement of the magnetic suspension bearing rotor without the need for a mechanical eddy current sensor, solving the problem of the rotor suddenly falling when the mechanical sensor is damaged and the friction with the mechanical part damaging the system, thereby avoiding the magnetic bearing rotor falling accident caused by the eddy current sensor being damaged when the eddy current sensor is used, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it reduces the system cost and volume, improves the stability of the magnetic suspension bearing control system, and facilitates the later maintenance.

[0153] In some embodiments, the control unit 106 takes the current control current of the magnetic bearing coil as an input parameter, inputs it into the displacement soft measurement model, and determines the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, including: the control unit 106 is specifically configured to set the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil.

[0154] Thus, the scheme of the present invention proposes a magnetic bearing rotor displacement soft measurement device based on a wavelet neural network model optimized by particle swarm optimization, and the wavelet neural network model is trained by the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module (i.e., a displacement soft measurement model), and realizes a sensorless radial displacement self-detection technology, that is, it can self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor, and solves the problem of the rotor suddenly falling when the mechanical sensor is damaged, and the friction with the mechanical part damaging the system, thereby avoiding the magnetic bearing rotor falling accident caused by the eddy current sensor when the eddy current sensor is damaged, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it also solves the problem of large volume and high cost of the magnetic bearing control system, and can avoid the problem of large volume of the magnetic bearing control system caused by the use of eddy current sensors when measuring the radial displacement of the magnetic bearing rotor. In addition, for magnetic bearings that are nonlinear and difficult to establish a rotor displacement calculation model, automatic measurement without displacement sensors is realized, and no additional complex circuit is required, which is easy to implement.

[0155] In some embodiments, the process further includes: controlling the current of the magnetic bearing coil based on the determined current displacement parameter of the magnetic bearing rotor, specifically as follows:

[0156] The control unit 106 is also configured to adjust the current control current of the magnetic bearing coil according to the current displacement parameter of the magnetic bearing rotor after determining the current displacement parameter of the magnetic bearing rotor, so as to obtain the adjustment value of the current control current of the magnetic bearing coil. The specific functions and processing of the control unit 106 can also be found in step S510. Specifically, according to the current displacement parameter of the magnetic bearing rotor, the current control current of the magnetic bearing coil is adjusted, including: according to the difference between the current displacement parameter of the magnetic bearing rotor and the reference displacement parameter of the magnetic bearing rotor, after PID control, a first processing result is obtained. Then, the difference between the first processing result and the current control current of the magnetic bearing coil is subjected to PWM processing to obtain a second processing result. Then, the second processing result is subjected to power amplification processing to obtain the adjustment value of the current control current of the magnetic bearing coil.

[0157] The control unit 106 is further configured to control the operation of the magnetic bearing coil according to the adjustment value of the current control current of the magnetic bearing coil. The specific functions and processing of the control unit 106 are also shown in step S520.

[0158] Fig.10 FIG. 1 is a schematic diagram of the structure of an embodiment of a displacement measurement circuit based on a wavelet neural network. Fig.10As shown, the displacement measurement circuit based on wavelet neural network includes: for each magnetic bearing coil, such as radial front X direction coil (FX coil), radial front Y direction coil (FY coil), radial rear X direction coil (RX coil) and radial rear Y direction coil (RY coil), a control branch is set. The current signal output by each control branch, that is, the control current of each magnetic bearing coil, is output after passing through the wavelet neural network (WNN) displacement measurement model (that is, the displacement measurement model based on wavelet neural network), and the predicted rotor displacement of the magnetic bearing rotor is output. The difference between each predicted rotor displacement and the given reference position signal of the path is obtained by the first comparator, and after passing through the control branch of the path, the adjustment value of the control current of the magnetic bearing coil of the path is output, and the magnetic bearing coil of the path is controlled according to the adjustment value of the control current of the magnetic bearing coil of the path. Among them, each control branch includes: a PWM control module, a second comparator, a PWM module and a power amplifier. The PWM control module performs PID processing on the difference between each predicted rotor displacement and the given reference position signal of the path, and outputs a first processing result. The difference between the first processing result of each path and the predicted rotor displacement of the path is processed by the PWM module to obtain the second processing result. The second processing result is processed by the power amplifier to obtain the adjustment value of the control current of the magnetic bearing coil of the path.

[0159] exist Fig.10 In the example shown, the displacement measurement model based on wavelet neural network is applied in the magnetic bearing control system. The wavelet neural network displacement prediction model uses the current signal i of the power amplifier (such as a switching power amplifier) FX 、i FY 、i RX 、i RY is the input signal, and the output is the predicted rotor displacement s FX 、s FY 、s RX 、s RY and compare it with the given reference position signal f FX 、f FY 、f RX 、f RY For comparison, the difference is passed through a closed-loop controller and a switching power amplifier, ultimately achieving stable suspension of the magnetic bearing.

[0160] Thus, the scheme of the present invention proposes a magnetic bearing rotor displacement soft measurement device based on a wavelet neural network model optimized by particle swarm optimization, and the wavelet neural network model is trained by the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module (i.e., a displacement soft measurement model), and realizes a sensorless radial displacement self-detection technology, that is, it can self-detect the radial displacement of the magnetic bearing rotor without the need for a mechanical eddy current sensor, and solves the problem of the rotor suddenly falling when the mechanical sensor is damaged, and the friction with the mechanical part damaging the system, thereby avoiding the magnetic bearing rotor falling accident caused by the eddy current sensor when the eddy current sensor is damaged, and can improve the suspension safety of the magnetic bearing rotor. At the same time, it also solves the problem of large volume and high cost of the magnetic bearing control system, and can avoid the problem of large volume of the magnetic bearing control system caused by the use of eddy current sensors when measuring the radial displacement of the magnetic bearing rotor. In addition, for magnetic bearings that are nonlinear and difficult to establish a rotor displacement calculation model, automatic measurement without displacement sensors is realized, and no additional complex circuit is required, which is easy to implement.

[0161] It should be noted that the wavelet neural network-based displacement soft measurement device in the solution of the present invention can be used not only in magnetic bearing systems, but also in other application scenarios of eddy current sensors.

[0162] Since the processing and functions implemented by the device of this embodiment are basically corresponding to the aforementioned Figures 1 to 5 The embodiments, principles and examples of the method shown are as follows. Therefore, for any details not fully described in the description of this embodiment, please refer to the relevant descriptions in the aforementioned embodiments and will not be repeated here.

[0163] After a large number of experimental verifications, the technical solution of the present invention is adopted to train the model through the collected sample data. The trained model can be directly connected in series to the magnetic suspension bearing system to form a radial displacement prediction module, which can realize self-detection of the radial displacement of the magnetic suspension bearing rotor, and solve the problem of sudden falling of the rotor when the mechanical sensor is damaged, and the friction with the mechanical part damaging the system. It can avoid the falling accident of the magnetic bearing rotor caused by the eddy current sensor when the eddy current sensor is damaged, and can improve the suspension safety of the magnetic bearing rotor.

[0164] According to an embodiment of the present invention, a magnetic bearing system corresponding to the displacement determination device of the magnetic bearing is also provided. The magnetic bearing system may include: the displacement determination device of the magnetic bearing described above.

[0165] Since the processing and functions implemented by the magnetic bearing system of this embodiment are basically corresponding to the aforementioned Figure 6The embodiments, principles and examples of the device shown are as follows. Therefore, for matters not fully described in the description of this embodiment, reference can be made to the relevant descriptions in the aforementioned embodiments, and no further elaboration will be given here.

[0166] After a large number of experimental verifications, the technical solution of the present invention is adopted to train the model through the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module, which can realize self-detection of the radial displacement of the magnetic bearing rotor, solve the problems of large size and high cost of the magnetic bearing control system, and can avoid the use of eddy current sensors when measuring the radial displacement of the magnetic bearing rotor, which leads to the problem of increased size of the magnetic bearing control system.

[0167] According to an embodiment of the present invention, a storage medium corresponding to the method for determining the displacement of a magnetic bearing is also provided, wherein the storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the method for determining the displacement of the magnetic bearing described above.

[0168] Since the processing and functions implemented by the storage medium of this embodiment are basically corresponding to the aforementioned Figures 1 to 5 The embodiments, principles and examples of the method shown are as follows. Therefore, for any details not fully described in the description of this embodiment, please refer to the relevant descriptions in the aforementioned embodiments and will not be repeated here.

[0169] After a large number of experimental verifications, the technical solution of the present invention is adopted to train the model through the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module, thereby realizing self-detection of the radial displacement of the magnetic bearing rotor. For magnetic bearings that are nonlinear and difficult to establish a rotor displacement calculation model, automatic measurement without a displacement sensor can be realized, which does not require additional complex circuits and is easy to implement.

[0170] According to an embodiment of the present invention, a processor corresponding to the method for determining the displacement of a magnetic bearing is also provided, and the processor is used to run a program, wherein the method for determining the displacement of a magnetic bearing described above is executed when the program is run.

[0171] Since the processing and functions implemented by the processor of this embodiment are basically corresponding to the aforementioned Figures 1 to 5 The embodiments, principles and examples of the method shown are as follows. Therefore, for any details not fully described in the description of this embodiment, please refer to the relevant descriptions in the aforementioned embodiments and will not be repeated here.

[0172] After a large number of experimental verifications, the technical solution of the present invention is adopted to train the model through the collected sample data. The trained model can be directly connected in series to the magnetic bearing system to form a radial displacement prediction module, which can realize self-detection of the radial displacement of the magnetic bearing rotor, reduce the system cost and volume, improve the stability of the magnetic bearing control system, and facilitate subsequent maintenance.

[0173] In summary, it is easy for those skilled in the art to understand that, under the premise of no conflict, the above-mentioned advantageous methods can be freely combined and superimposed.

[0174] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A method for determining the displacement of a magnetic bearing, characterized in that: The magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor; the displacement determination method of the magnetic bearing comprises: Constructing a displacement measurement model based on a wavelet neural network, denoted as a displacement soft measurement model, comprising: collecting a set control current of the magnetic bearing coil and a set displacement parameter of the magnetic bearing rotor as sample data; based on the sample data, training the wavelet neural network with the set control current of the magnetic bearing coil as an input parameter and the set displacement parameter of the magnetic bearing rotor as an output parameter to obtain the displacement soft measurement model; In case that the current displacement parameter of the magnetic bearing rotor needs to be determined, obtaining the current control current of the magnetic bearing coil; The current control current of the magnetic bearing coil is used as an input parameter and input into the displacement soft measurement model, and the output parameter of the displacement soft measurement model is determined as the current displacement parameter of the magnetic bearing rotor, so as to determine the current displacement parameter of the magnetic bearing rotor; wherein, the current control current of the magnetic bearing coil is used as an input parameter and input into the displacement soft measurement model, and the output parameter of the displacement soft measurement model is determined as the current displacement parameter of the magnetic bearing rotor, including: setting the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil; The displacement soft measurement model is trained by using the collected sample data, and the trained displacement soft measurement model is directly connected in series to the magnetic suspension bearing system to form a radial displacement prediction module, thereby realizing self-detection of the radial displacement of the magnetic suspension bearing rotor.

2. The method for determining the displacement of a magnetic bearing according to claim 1, characterized in that: The number of the magnetic bearing coils is more than one; Collecting the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor as sample data includes: Selecting one or more set control currents of the magnetic bearing coils and set displacement parameters of the magnetic bearing rotor as characteristic variables; The characteristic variables are normalized to form sampling data, which are used as sample data for training and testing.

3. The method for determining the displacement of a magnetic bearing according to claim 1, characterized in that: Based on the sample data, the set control current of the magnetic bearing coil is used as an input parameter, and the set displacement parameter of the magnetic bearing rotor is used as an output parameter, and the wavelet neural network is trained to obtain the displacement soft measurement model, including: Setting a wavelet neural network model; when the number of the magnetic bearing coils is more than one, the number of layers of the wavelet neural network model is more than one layer; Based on the particle swarm optimization algorithm, the parameters of the set wavelet neural network model are optimized to obtain the optimized wavelet neural network model; Based on the sample data, the set control current of the magnetic bearing coil is used as an input parameter, and the set displacement parameter of the magnetic bearing rotor is used as an output parameter, and the optimized wavelet neural network model is trained to obtain the displacement soft measurement model.

4. The method for determining the displacement of a magnetic bearing according to any one of claims 1 to 3, characterized in that: Also includes: After the current displacement parameter of the magnetic bearing rotor is determined, the current control current of the magnetic bearing coil is adjusted according to the current displacement parameter of the magnetic bearing rotor to obtain an adjustment value of the current control current of the magnetic bearing coil; The operation of the magnetic bearing coil is controlled according to the adjustment value of the current control current of the magnetic bearing coil.

5. A displacement determination device for a magnetic bearing, characterized in that: The magnetic bearing comprises: a magnetic bearing coil and a magnetic bearing rotor; the displacement determination device of the magnetic bearing comprises: The modeling unit is configured to construct a displacement measurement model based on a wavelet neural network, which is recorded as a displacement soft measurement model, including: collecting a set control current of the magnetic bearing coil and a set displacement parameter of the magnetic bearing rotor as sample data; based on the sample data, the set control current of the magnetic bearing coil is used as an input parameter, and the set displacement parameter of the magnetic bearing rotor is used as an output parameter, training the wavelet neural network to obtain the displacement soft measurement model; an acquisition unit, configured to acquire a current control current of the magnetic bearing coil when a current displacement parameter of the magnetic bearing rotor needs to be determined; A control unit is configured to take the current control current of the magnetic bearing coil as an input parameter, input it into the displacement soft measurement model, and determine the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, so as to determine the current displacement parameter of the magnetic bearing rotor; wherein, the control unit takes the current control current of the magnetic bearing coil as an input parameter, inputs it into the displacement soft measurement model, and determines the output parameter of the displacement soft measurement model as the current displacement parameter of the magnetic bearing rotor, including: setting the displacement soft measurement model between the sampling end of the current control current of the magnetic bearing coil and the feedback end of the current displacement parameter of the magnetic bearing rotor, so that the displacement soft measurement model can output the current displacement parameter of the magnetic bearing rotor based on the current control current of the magnetic bearing coil; The displacement soft measurement model is trained by using the collected sample data, and the trained displacement soft measurement model is directly connected in series to the magnetic suspension bearing system to form a radial displacement prediction module, thereby realizing self-detection of the radial displacement of the magnetic suspension bearing rotor.

6. The displacement determination device of a magnetic bearing according to claim 5, characterized in that: The number of the magnetic bearing coils is more than one; The modeling unit collects the set control current of the magnetic bearing coil and the set displacement parameter of the magnetic bearing rotor as sample data, including: Selecting one or more set control currents of the magnetic bearing coils and set displacement parameters of the magnetic bearing rotor as characteristic variables; The characteristic variables are normalized to form sampling data, which are used as sample data for training and testing.

7. The displacement determination device of a magnetic bearing according to claim 5, characterized in that: The modeling unit, based on the sample data, takes the set control current of the magnetic bearing coil as an input parameter and takes the set displacement parameter of the magnetic bearing rotor as an output parameter, trains the wavelet neural network to obtain the displacement soft measurement model, including: Setting a wavelet neural network model; when the number of the magnetic bearing coils is more than one, the number of layers of the wavelet neural network model is more than one layer; Based on the particle swarm optimization algorithm, the parameters of the set wavelet neural network model are optimized to obtain the optimized wavelet neural network model; Based on the sample data, the set control current of the magnetic bearing coil is used as an input parameter, and the set displacement parameter of the magnetic bearing rotor is used as an output parameter, and the optimized wavelet neural network model is trained to obtain the displacement soft measurement model.

8. The displacement determination device of a magnetic bearing according to any one of claims 5 to 7, characterized in that: Also includes: The control unit is further configured to adjust the current control current of the magnetic bearing coil according to the current displacement parameter of the magnetic bearing rotor after determining the current displacement parameter of the magnetic bearing rotor, so as to obtain an adjustment value of the current control current of the magnetic bearing coil; The control unit is further configured to control the operation of the magnetic bearing coil according to the adjustment value of the current control current of the magnetic bearing coil.

9. A magnetic bearing system, characterized in that: include: A displacement determination device for a magnetic bearing as claimed in any one of claims 5 to 8.

10. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute the method for determining the displacement of a magnetic bearing according to any one of claims 1 to 4.

11. A processor, characterized in that: The processor is used to run a program, wherein the program executes the displacement determination method of a magnetic bearing according to any one of claims 1 to 4 when running.