Adaptive pid-based magnetic bearing support characteristic analysis method and system

Abstract

The application provides a magnetic suspension bearing supporting characteristic analysis method and system based on adaptive PID, comprising the following steps: building a PID control system based on a neural network; constructing dynamic stiffness and dynamic damping expression of a magnetic suspension rotor based on PID parameters; solving PID initial parameters according to characteristic parameters of a magnetic suspension bearing rotor dynamics model; dynamically adjusting PID parameters according to the identified magnetic suspension bearing system; and analyzing system supporting characteristics in the system operation process according to the expression. The analysis method and system can realize dynamic adjustment of PID control parameters, and have good decoupling performance, anti-interference performance and vibration suppression effect.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of magnetic suspension bearing control system, and particularly relates to a magnetic suspension bearing supporting characteristic analysis method and system based on adaptive PID. BACKGROUND

[0002] The research on the control algorithm of the magnetic suspension bearing rotor system shows that the essential instability, nonlinearity and multi-degree-of-freedom coupling characteristics and parameter uncertainty make it particularly crucial to select a suitable control algorithm and control parameters to ensure the stable suspension and high-speed rotation of the magnetic suspension rotor.

[0003] However, the algorithms based on traditional control theory, such as PID and integral separation PID, although relatively simple to implement, show their inherent limitations when facing the multi-degree-of-freedom coupling and nonlinear magnetic suspension bearing-rotor system. The PID parameter design generally requires rich engineering experience and repeated debugging, and on the basis of the appropriate proportional, differential and integral coefficients, the parameters of the series trap wave filter, phase lead compensator and the like also need to be carefully debugged. The parameters themselves have strong randomness, and unreasonable control parameters often lead to electromagnet saturation and even rotor instability. As for some intelligent control algorithms, such as the LQG algorithm, because the accuracy of the model is relatively high, if the uncertain factors cannot be considered in the system, the robustness of the control system will be affected. At the same time, the magnetic suspension bearing system is a nonlinear and multi-degree-of-freedom coupling critical stable system, and with the long-time operation of the system, the coupling of the system is aggravated, and the traditional control algorithm has poor performance in resisting multi-source disturbances such as high temperature, thermal bending and time delay. SUMMARY

[0004] In order to solve the above problems existing in the prior art, the application provides a magnetic suspension bearing supporting characteristic analysis method and system based on adaptive PID, which can realize dynamic adjustment of PID control parameters and has good decoupling performance, anti-interference performance and vibration suppression effect.

[0005] In order to achieve the above purpose, the application adopts the following technical solutions:

[0006] The application provides a magnetic suspension bearing supporting characteristic analysis method based on adaptive PID, which comprises the following steps:

[0007] Step 1: build a PID control system based on a neural network;

[0008] Step 2: construct a dynamic stiffness and dynamic damping expression of the magnetic suspension rotor based on the PID parameters;

[0009] Step 3: solve the initial parameters of PID according to the characteristic parameters of the magnetic suspension bearing rotor dynamics model;

[0010] Step 4: Dynamically adjust PID parameters based on the identified magnetic bearing system;

[0011] Step 5: Analyze the system support characteristics during system operation based on the dynamic stiffness and dynamic damping expressions of the magnetic levitation rotor;

[0012] Wherein, in said step 1, a PID control algorithm of a diagonal recursive neural network is adopted to build a PID control system based on a neural network;

[0013] The dynamic stiffness and dynamic damping expressions in step 2 are:

[0014]

[0015] Among them, k e represents dynamic stiffness; c e represents dynamic damping; k x represents the displacement stiffness coefficient; k i represents the current stiffness coefficient; A s Represents the sensor magnification; A p Represents the power amplifier amplification factor; k p Represents the controller proportional coefficient; k ic Represents the controller integral coefficient; k d represents the controller differential coefficient; T s Represents the sensor hysteresis time constant; T p represents the hysteresis time constant of the power amplifier, ω represents the rotation speed of the magnetic bearing rotor, s is the transfer function operator, Re is the real part, and Im is the imaginary part;

[0016] In step 3, solving the PID initial parameters according to the characteristic parameters of the magnetic bearing rotor dynamics model includes: solving the rotor dynamics model under different control parameters according to the dynamic stiffness and dynamic damping expressions in step 2, and selecting the optimal initial parameters under the dynamic model according to the performance indicators of the magnetic bearing system;

[0017] In step 4, dynamically adjusting the PID parameters according to the identified magnetic bearing system includes: obtaining an identified magnetic bearing system according to the result obtained by solving the PID initial parameters in step 3, using the diagonal recursive neural network in step 1 as a network identifier, and dynamically adjusting the PID parameter signal regulation through the difference between the network prediction and the actual output.

[0018] In the above solution, in step 5, the system support characteristics during system operation are analyzed based on the dynamic stiffness and dynamic damping expressions of the magnetic levitation rotor, including identification analysis of dynamic stiffness and dynamic damping.

[0019] In the above scheme, the PID parameter comprises a proportional coefficient, a differential coefficient and an integral coefficient.

[0020] The application also provides a control system constructed based on a self-adaptive PID-based magnetic suspension bearing supporting characteristic analysis method, comprising:

[0021] A control object of the control system, comprising a magnetic suspension bearing;

[0022] A sensor of the control system, used for sampling position data of the magnetic suspension bearing and the rotor;

[0023] A controller of the control system, a PID controller serving as the controller of the control system;

[0024] An actuator of the control system, a power amplifier serving as the actuator of the control system;

[0025] A network identifier of the control system, a diagonal recurrent neural network serving as the network identifier of the control system, used for dynamically adjusting PID parameters.

[0026] The sensor is a displacement sensor.

[0027] Compared with the prior art, the application has the beneficial effects that: the diagonal recurrent neural network is used as the network identifier of the control system, used for dynamically adjusting PID parameters, and has more advantages in processing transient information, and is suitable for solving the nonlinearity and torque coupling dynamic problems of the magnetic suspension bearing system.

[0028] Based on the dynamic stiffness and dynamic damping expression, the initial control parameters of the PID dynamic parameter adjusted magnetic suspension bearing system obtained through supporting characteristic analysis of the magnetic suspension bearing system with dynamic control parameter adjustment are the optimal parameters obtained based on the magnetic suspension rotor dynamics model, so that the dynamics characteristics of the system are more excellent.

[0029] The diagonal recurrent neural network used in the application realizes dynamic adjustment of PID control parameters, and has good decoupling performance, anti-interference performance and vibration suppression effect. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 It is a magnetic suspension bearing control system based on PID dynamic parameter adjustment and performance analysis step schematic diagram provided by the embodiment of the application.

[0031] Figure 2 It is a magnetic suspension bearing control system based on PID dynamic parameter adjustment and performance analysis step schematic diagram provided by the embodiment of the application.

[0032] Figure 3 It is a PID parameter dynamic change curve when the system is disturbed by a sinusoidal signal.

[0033] Figure 4 is a dynamic stiffness change curve when a system is interfered by a sinusoidal signal, provided by an embodiment of the present application;

[0034] Figure 5 is a dynamic damping change curve when a system is interfered by a sinusoidal signal, provided by an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical scheme and effect of the present application clearer and more explicit, and further describe the present application in detail with reference to the drawings and examples, it should be understood that the specific implementation described herein is only used to explain the present application, and is not used to limit the present application.

[0036] As shown in Figure 1 , the present application provides a magnetic suspension bearing supporting characteristic analysis method based on adaptive PID, comprising the following steps:

[0037] Step 1: build a PID control system based on neural network;

[0038] Step 2: construct dynamic stiffness and dynamic damping expression of magnetic suspension rotor based on PID parameters;

[0039] Step 3: solve the initial parameters of PID according to the characteristic parameters of the magnetic suspension bearing rotor dynamics model;

[0040] Step 4: adjust the PID parameters according to the dynamic identification of the magnetic suspension bearing system;

[0041] Step 5: analyze the system supporting characteristics during system operation according to the dynamic stiffness and dynamic damping expression of the magnetic suspension rotor;

[0042] In step 1, the diagonal recurrent neural network PID control algorithm is used to build a PID control system based on neural network;

[0043] The dynamic stiffness and dynamic damping expression in step 2 is:

[0044]

[0045] Wherein, k e represents dynamic stiffness; c e represents dynamic damping; k x represents displacement stiffness coefficient; k i represents current stiffness coefficient; A s represents sensor amplification; A p represents power amplifier amplification; k p represents controller proportional coefficient; k ic represents controller integral coefficient; kd represents controller derivative coefficient; T s represents sensor lag time constant; T p represents power amplifier lag time constant, ω represents rotating speed of magnetic suspension bearing rotor, s is transfer function operator, Re is to take real part, and Im is to take imaginary part;

[0046] In step 3, the PID initial parameters are solved according to the characteristic parameters of the magnetic suspension bearing rotor dynamics model, which includes solving the rotor dynamics model under different control parameters according to the dynamic stiffness and dynamic damping expressions in step 2, and selecting the optimal initial parameters under the dynamics model according to the performance index of the magnetic suspension bearing system.

[0047] In step 4, the PID parameters are dynamically adjusted according to the identified magnetic suspension bearing system, which includes obtaining the identified magnetic suspension bearing system according to the results of solving the PID initial parameters in step 3, using the diagonal recurrent neural network in step 1 as a network identifier, and adjusting the PID parameter signal regulation through the difference between network prediction and actual output.

[0048] In step 5, the system support characteristics during system operation are analyzed according to the dynamic stiffness and dynamic damping expressions of the magnetic suspension rotor, including identification analysis of dynamic stiffness and dynamic damping.

[0049] The PID parameters include proportional coefficient, derivative coefficient and integral coefficient.

[0050] The diagonal recurrent neural network structure has three layers in total, and the hidden layer is a regression layer. Its algorithm is as follows:

[0051]

[0052] In the formula, I is the network input vector, I i (k) is the input of the i-th neuron of the input layer, and the output of the j-th neuron of the network regression layer is X j (k), S j (k) is the input sum of the j-th regression neuron, f(·) is the function of S, O(k) is the output of the network, W D and W O are the weight vectors of the network regression layer and the output layer, and W I is the weight vector of the network input layer.

[0053] In the formula, k represents the number of network iterations; DRNN represents the diagonal recurrent neural network identifier; u(k) and y(k) both represent the input of the identifier, and also represent the input and output of the magnetic suspension bearing system; ym(k) represents the output of the identifier, that is, the P, I and D parameters of the control system. The signal regulation is performed through the difference between network prediction and actual output.

[0054] The output of the network output layer is defined as follows:

[0055]

[0056] The output and input of the network regression layer are defined as follows:

[0057] Xj(k) = f(Sj(k))

[0058]

[0059] The identification error and index are defined as follows:

[0060] em(k) = y(k) - ym(k)

[0061]

[0062] The gradient descent method is used as the learning algorithm:

[0063]

[0064] The regression layer neurons use a double S function:

[0065]

[0066] In the formula, η I , η D , and η o represent the learning rates of the input layer, the regression layer, and the output layer, respectively; and α is an inertia coefficient.

[0067] As shown in Figure 2 , the application further provides a control system constructed based on a magnetic suspension bearing supporting characteristic analysis method of an adaptive PID, comprising:

[0068] A control object of the control system, comprising a magnetic suspension bearing;

[0069] A sensor of the control system, used for sampling position data of the magnetic suspension bearing and a rotor;

[0070] A controller of the control system, a PID controller serving as the controller of the control system;

[0071] An actuator of the control system, a power amplifier serving as the actuator of the control system;

[0072] A network identifier of the control system, an angle recursive neural network serving as the network identifier of the control system, used for dynamically adjusting PID parameters.

[0073] The control system involves diagonal recurrent neural network to realize PID dynamic adjustment of control parameters, wherein, ri (i=1, 2, 3, 4, 5) is a reference input position of the control system; ei (i=1, 2, 3, 4, 5) is an error of the control system;

[0074] ui (i=1, 2, 3, 4, 5) is a control quantity of the control system; yi (i=1, 2, 3, 4, 5) is an output of the control system;

[0075] ymi (i=1, 2, 3, 4, 5) is an output of the DRNN identifier; kpi (i=1, 2, 3, 4, 5) is a proportional coefficient of the controller; kii (i=1, 2, 3, 4, 5) is an integral coefficient of the controller; kdi (i=1, 2, 3, 4, 5) is a differential coefficient of the control system.

[0076] Referring to FIG. 2, during system operation, the displacement sensor monitors the position deviation of the rotor and converts the displacement into a voltage signal, which is further converted into a digital signal through A / D conversion. The controller receives the signals and calculates the displacement deviation based on the expected suspension position of the rotor. Then, PID parameters are obtained through diagonal recurrent neural network identification, so as to generate a control voltage signal. The voltage signal is converted through D / A conversion and delivered to a power amplifier, so as to adjust the current in the electromagnet, and finally realize the suspension of the magnetic suspension rotor system at the balance position.

[0077] In an embodiment of the present application, when subjected to a 0.5A, 50Hz sinusoidal current interference, the step 4 according to the identified results of dynamic adjustment of PID parameters of the magnetic suspension bearing system is as shown in FIG. 4: Figure 3 It is observed that the PID parameters change, kp first increases and then decreases at 0.1s, then fluctuates within 0.06s and finally remains stable, kic increases at 0.1s, then gradually fluctuates and decreases, and finally remains stable, and kd presents an opposite trend to ki.

[0078] For step 5, the system support characteristics during system operation are analyzed according to the dynamic stiffness and dynamic damping expressions of the magnetic suspension rotor; specifically, when subjected to a 0.5A, 50Hz sinusoidal current interference, the support characteristic results of step 5 are as shown in FIG. 5: Figure 4 and Figure 5 After 0.1s, the dynamic stiffness and the dynamic damping both first decrease and then increase along with the trend of the sinusoidal wave and finally remain stable.

[0079] As can be seen from the above embodiment, the diagonal recurrent neural network is used as the network identifier of the control system to dynamically adjust the PID parameters, which has more advantages in processing transient information and is suitable for solving the nonlinear and torque coupling dynamic problems of the magnetic suspension bearing system.

Claims

1. A method for analyzing the supporting characteristics of a magnetic bearing based on adaptive PID, characterized in that, comprising the following steps: Step 1: building a PID control system based on neural network; Step 2: constructing dynamic stiffness and dynamic damping expression of the magnetic bearing rotor based on PID parameters; Step 3: solving the initial parameters of PID according to the characteristic parameters of the magnetic bearing rotor dynamics model; Step 4: dynamically adjusting the PID parameters according to the identified magnetic bearing system; Step 5: analyzing the supporting characteristics of the system during operation according to the dynamic stiffness and dynamic damping expression of the magnetic bearing rotor; In step 1, the diagonal recurrent neural network PID control algorithm is used to build a PID control system based on neural network. The dynamic stiffness and dynamic damping expression in step 2 is: In step 3, solving the initial parameters of PID according to the characteristic parameters of the magnetic bearing rotor dynamics model includes: solving the rotor dynamics model under different control parameters according to the dynamic stiffness and dynamic damping expression in step 2, and selecting the optimal initial parameters under the dynamics model according to the performance indicators of the magnetic bearing system. where k e represents dynamic stiffness; c e represents dynamic damping; k x represents displacement stiffness coefficient; k i represents current stiffness coefficient; A s represents sensor amplification factor; A p represents power amplifier amplification factor; k p represents controller proportional coefficient; k ic represents controller integral coefficient; k d represents controller derivative coefficient; T s represents sensor lag time constant; T p represents power amplifier lag time constant, ω represents rotational speed of magnetic suspension bearing rotor, s is transfer function operator, Re is to take real part, and Im is to take imaginary part; In step 4, dynamically adjusting the PID parameters according to the identified magnetic bearing system includes: obtaining the identified magnetic bearing system according to the results of solving the initial parameters of PID in step 3, using the diagonal recurrent neural network in step 1 as the network identifier, and dynamically adjusting the PID parameter signal adjustment through the difference between network prediction and actual output. In step 5, analyzing the supporting characteristics of the system during operation according to the dynamic stiffness and dynamic damping expression of the magnetic bearing rotor includes identification analysis of dynamic stiffness and dynamic damping.

2. The method for supporting characteristic analysis of a magnetic bearing based on adaptive PID according to claim 1, characterized in that, The PID parameters include proportional coefficient, derivative coefficient and integral coefficient.

3. The method of claim 1, wherein the method is characterized by, The control object of the control system includes the magnetic bearing; 4. The control system constructed according to the method for analyzing the characteristics of the adaptive PID-based magnetic bearing support according to any one of claims 1-3, characterized in that, The sensor of the control system is used to sample the position data of the magnetic bearing and the rotor; The controller of the control system is the PID controller as the controller of the control system; The actuator of the control system is the power amplifier as the actuator of the control system; The network identifier of the control system is the diagonal recurrent neural network as the network identifier of the control system, which is used to dynamically adjust the PID parameters. The sensor is a displacement sensor. ​ 5. The control system constructed based on the adaptive PID based magnetic bearing support characteristic analysis method of claim 4, wherein, ​

Images