A method and system for self-tuning control parameters of a magnetic bearing system

Abstract

The application discloses a self-tuning method and system for control parameters of a magnetic suspension bearing system, and the method comprises the following steps: acquiring a rotor displacement signal of the magnetic suspension bearing system; achieving preliminary suspension of the rotor according to the rotor displacement signal, with a steady-state position and a response speed as indexes; further optimizing control parameters with a peak value of a sensitivity function as an optimization target; and achieving self-tuning of the control parameters by using a breadth-first pruning search tree algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of magnetic bearing self-tuning, and more specifically, relates to a magnetic bearing system control parameter self-tuning method and system. Background Art

[0002] Active magnetic bearing systems use controllable electromagnetic forces to suspend the rotor between the stator's magnetic poles. This system offers advantages such as zero friction, no lubrication, no pollution, high speed, and long life. Currently, active magnetic bearings, as an advanced mechatronic product, have found applications in a range of industrial applications, including gyroscopes, high-speed motors, bearingless motors, aircraft engines, artificial heart pumps, and distributed power generation systems.

[0003] Existing magnetic bearing control systems typically require a complex process of parameter tuning, including model linearization, simulation analysis, and on-site commissioning. This results in a long design cycle and numerous steps. Therefore, it is necessary to implement online self-tuning of the control parameters of magnetic bearing systems using machine learning algorithms. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a method for self-tuning control parameters of a magnetic bearing system, comprising:

[0005] Obtaining a rotor displacement signal of a magnetic bearing system;

[0006] According to the rotor displacement signal, the rotor is initially suspended with steady-state position and response speed as indicators;

[0007] Taking the peak value of the sensitivity function as the optimization target, the control parameters are further optimized, and the breadth-first pruning search tree algorithm is used to realize the self-tuning of the control parameters.

[0008] Further, including:

[0009] The plane formed by the combination of control parameters of the PID controller with a fixed Ki value is infinitely subdivided into a grid, wherein the plane includes: a Kp-Kd plane, Kp is the proportional link coefficient of the PID controller, and Kd is the differential link coefficient;

[0010] Test and score the parameters represented by the grid nodes;

[0011] The pruning algorithm is used to select the node with the highest score, and the pruned node is used as the parent node to derive child nodes in the surrounding areas for a more detailed exploration of the parameter plane.

[0012] Further, including:

[0013] Set the search range and step size of the PID controller control parameters;

[0014] The range of the PID controller control parameter is subdivided into multiple grid points according to a step size, and all grid nodes are added to a search queue;

[0015] A first control parameter combination is selected from the search queue for testing, and it is determined whether the selected control parameter combination can make the rotor reach a stable suspension state;

[0016] For the control parameter combination that passes the stability test, it is recorded as a stable node, and the control parameter combination is removed from the search queue.

[0017] Further, comprising:

[0018] For each degree of freedom, the control parameter combination that passes the stability test is taken as the geometric center of the closed space formed on the parameter plane of the respective degree of freedom, and the control parameter combination corresponding to the geometric center is taken as the control parameter combination of the root node of the search tree.

[0019] The control parameter combination of the root node is subjected to a frequency sweep test of the sensitivity function, and the score of the control parameter combination of the root node and the amplitude of the sensitivity function are recorded.

[0020] Further, comprising:

[0021] The PID controller adds an excitation signal of different frequencies as a rotor displacement instruction in a closed-loop control system, and outputs a displacement error signal.

[0022] Based on the swept frequency and the corresponding sensitivity test results, the detection results of each degree of freedom are plotted.

[0023] The application also provides a magnetic suspension bearing system control parameter self-tuning system, comprising:

[0024] An acquisition signal module is configured to acquire a rotor displacement signal of a magnetic suspension bearing system.

[0025] A preliminary suspension module is configured to realize preliminary suspension of the rotor according to the rotor displacement signal, with a steady-state position and a response speed as indexes.

[0026] A self-tuning module is configured to further optimize the control parameters with a sensitivity function peak value as an optimization target, and to realize self-tuning of the control parameters by using a breadth-first pruning search tree algorithm.

[0027] Further, comprising:

[0028] A plane formed by a control parameter combination of a fixed Ki value PID controller is infinitely subdivided into a grid, wherein the plane includes a Kp-Kd plane, Kp is a proportional link coefficient of the PID controller, and Kd is a differential link coefficient.

[0029] Test and score the parameters represented by the grid nodes;

[0030] Select the node with the highest score by using the pruning algorithm, and use the pruned node as the parent node to derive child nodes in all directions to explore the parameter plane more meticulously.

[0031] Further comprising:

[0032] Set the search range and step size of the PID controller control parameters;

[0033] Subdivide the range of the PID controller control parameters into multiple grid points according to the step size, and add all the grid nodes to the search queue;

[0034] Select the first control parameter combination from the search queue for testing, and determine whether the selected control parameter combination can make the rotor reach a stable suspension state;

[0035] For the control parameter combination that passes the stability test, record it as a stable node and remove it from the search queue.

[0036] Further comprising:

[0037] For each degree of freedom, take the geometric center of the closed space formed by the control parameter combinations that pass the stability test in the respective degree of freedom plane, and use the control parameter combination corresponding to the geometric center as the control parameter combination of the root node of the search tree.

[0038] Perform frequency sweep testing of the sensitivity function on the control parameter combination of the root node, and record the score and sensitivity function amplitude of the control parameter combination of the root node.

[0039] Further comprising:

[0040] The PID controller adds excitation signals of different frequencies as rotor displacement instructions in the closed-loop control system, and outputs displacement error signals.

[0041] Based on the frequency sweep and the corresponding sensitivity test results, draw the detection results of each degree of freedom.

[0042] Overall, compared with the prior art, the above technical solutions conceived by the present application have the following beneficial effects:

[0043] Through the above technical solutions, the present application can accurately and efficiently self-tune the control parameters of the magnetic suspension bearing system based on the sensitivity function. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 is the first stage flow chart of the PID parameter optimization algorithm;

[0045] Figure 2 is a simplified control block diagram of a rigid magnetic suspension rotor closed-loop system;

[0046] Figure 3 is a frequency sweep test principle block diagram of a magnetic suspension rotor (single degree of freedom) system;

[0047] Figure 4 is an example diagram of a sensitivity function scan result;

[0048] Figure 5 is a flow chart of a third stage of a PID parameter optimization algorithm;

[0049] Figure 6 is a system structure diagram of embodiment 2. DETAILED DESCRIPTION

[0050] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in combination with the drawings in the specification and specific embodiments.

[0051] The method provided by the application can be implemented in a terminal environment, which can include one or more of the following components: a processor, a storage medium and a display screen. The storage medium stores at least one instruction, which is loaded and executed by the processor to implement the method described in the following embodiments.

[0052] The processor can include one or more processing cores. The processor connects various parts in the entire terminal through various interfaces and lines, executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the storage medium, and calling data stored in the storage medium.

[0053] The storage medium can include random access memory (RAM) and read-only memory (ROM). The storage medium can be used to store instructions, programs, codes, code sets or instructions.

[0054] The display screen is used to display the user interface of each application program.

[0055] In addition, those skilled in the art can understand that the structure of the terminal described above does not constitute a limitation on the terminal, and the terminal can include more or fewer components, or combine certain components, or different component arrangements. For example, the terminal also includes radio frequency circuit, input unit, sensor, audio circuit, power supply and other components, which will not be described here.

[0056] Embodiment 1

[0057] The core mechanism of the present application is to use a search tree to divide the parameter plane composed of PID controller control parameter combinations into small grids, test and score the parameters represented by the grid nodes, and select the best nodes using a pruning algorithm. The pruned nodes are used as parent nodes to derive child nodes in all directions to explore the parameter plane in more detail. p -K d The parameter plane is divided into small grids, and the parameters represented by the grid nodes are tested and scored, and the best nodes are selected using a pruning algorithm. p Kp is the proportional term coefficient of the PID controller, d Kd is the derivative term coefficient) is infinitely subdivided into small grids, and the parameters represented by the grid nodes are tested and scored, and the best nodes are selected using a pruning algorithm. The pruned nodes are used as parent nodes to derive child nodes in all directions to explore the parameter plane in more detail.

[0058] First stage: initial parameter (root node) search, the flowchart of this stage is shown in Figure 1

[0059] In the initial parameter search stage (first stage), our goal is to explore the initial stable parameter combination within the predetermined PID controller parameter range through a breadth-first search algorithm. The entire process is divided into several key steps, each focusing on how to effectively cover the parameter space and find stable parameters.

[0060] Based on the physical characteristics of the magnetic bearing rotor, we divide the radial four degrees of freedom (X and Y degrees of freedom of the DE and NDE ends of the rotor) into two degrees of freedom (X and Y degrees of freedom) for each degree of freedom combination at both ends. The two degrees of freedom (X and Y degrees of freedom) are shared by the control parameters and test results in this stage.

[0061] Each degree of freedom combination needs to follow the following steps to alternate parameter search.

[0062] 1) First, we need to set the search range and step size of the PID parameters. These ranges and steps determine the starting point of our search and the amplitude of each parameter adjustment to ensure that the entire parameter space can be quickly and effectively covered. They can be adjusted according to actual conditions.

[0063] 2) Next, divide the parameter range into multiple grid points according to the step size, and add all these grid points to the search queue. This queue will be used for subsequent breadth-first search.

[0064] 3) Before entering the formal search, we need to check if there are any untested parameter combinations in the search queue. If the search queue is empty, it means that all possible parameter combinations have been tested at the current step size. At this time, the step size needs to be reduced and the search queue needs to be regenerated (i.e. repeat step 2) to continue the search in more detail.

[0065] ​4) When there are still untested parameter combinations in the search queue, we will select the first parameter combination from the queue to test. At this time, we need to check whether the selected parameter combination has been tested. If the parameter combination has been tested, it will be removed from the search queue, and step 4 will be restarted to continue selecting the next parameter combination for testing.

[0066] 5) For untested parameter combinations, we will perform stability testing to determine whether the parameter combination can make the rotor reach a stable suspension state (judgment criteria are described below). If the test result shows that the parameter combination can stabilize the system, we will record it as a stable node and remove the parameter combination from the search queue. If it is unstable, it will be recorded in the search results, and its parent node information will be recorded as K p = 0, K d = 0, score infinite, score drop rate infinitesimal, and the frequency response results of the sensitivity function are all recorded as infinite. After deleting this set of parameters from the search queue, return to step 4.

[0067] 6) After recording the stable node, we again check whether there are untested parameter combinations in the search queue. If the queue is empty, return to step 3 to continue reducing the step size and regenerating the search queue. When there are still untested parameter combinations in the search queue, we need to find the geometric center of the closed space formed by all stable nodes in the parameter plane for the current degree of freedom. This center point will ensure that the control system is relatively stable and will not be close to collapse, thus helping us to reduce the interference of other degree of freedom lift tests.

[0068] 7) Next, check whether the number of stable nodes found meets the predetermined requirements. If the number is sufficient, we can end the first stage of the search. If the number is insufficient, continue step 3 to search until sufficient stable nodes are found.

[0069] 8) After finding sufficient stable nodes, the first stage of the search task is completed.

[0070] Static suspension judgment criteria

[0071] Stability is the primary consideration for algorithm scoring. The system needs to remain stable under different PID parameter combinations to ensure its reliability in practical applications.

[0072] We take the active magnetic bearing with a central active space radius of 150 units as an example.

[0073] For the rotor lift time domain position feedback signal returned by the lower computer with a length of 2 seconds (1 KHz sampling frequency), we can use the following points to determine whether it can be stably suspended:

[0074] a) the highest point of the lift-off waveform (overshoot) is between 140 and 170;

[0075] b) the average of the last 50 samples of the lift-off waveform (steady state error) is between 140 and 160;

[0076] c) the peak to peak value of the last 50 samples of the lift-off waveform (ring) is less than 10.

[0077] If the system shows unstable behavior, such as sustained oscillation or divergence, under a set of PID parameters, this set of parameters will be directly determined as unstable, and both the frequency sweep result and the parameter score will be recorded as infinity (Inf).

[0078] Second stage: root node stability and performance index calculation

[0079] For each set of stable parameters of a degree of freedom obtained in the first stage, we take the geometric center of the closed space formed by them in the parameter plane of the respective degree of freedom, and take the parameters corresponding to this geometric center as the root node parameters of the search tree.

[0080] It should be noted that from this stage, we will no longer explore in the form of degree of freedom combinations, i.e. each degree of freedom's parameters are explored, tested, scored and the results are stored separately.

[0081] For the root node of the search tree of each degree of freedom, we need to do the frequency sweep test of the sensitivity function (definition see below) and record the root node parameters, scores (score evaluation index see below) and sensitivity function amplitude. Since the root node has no parent node, its parent node K p = 0, K d = 0, score is infinity, score drop rate is infinitesimal.

[0082] As Figure 2 shown in the figure, in this closed-loop architecture, the controlled magnetic levitation rotor G includes the magnetic bearing-rigid rotor, power amplifier, displacement sensor link, and considers the effects of three independent inputs: position reference command r, control object input disturbance d and control object output disturbance (or noise) n. The output characteristics of the system are represented by the closed-loop transfer function T0(s), and the parameter variations of the forward channel are represented by the controlled object transfer function G(s). When only considering the position reference command r input, the closed-loop transfer function T0(s) is expressed as:

[0083] T o (s) = G(s)K(s)(I + G(s)K(s)) -1

[0084] T o (s) is the output sensitivity of T o (s) to G(s) and can be written as:

[0085]

[0086] When the sensitivity |S o (jω) is smaller, the magnetic bearing system has the following good control characteristics:

[0087] (1) The system has higher robustness to the perturbation of the magnetic bearing model parameters;

[0088] (2) The tracking error is smaller, and the control precision of the rotor position is higher;

[0089] (3) The system has better suppression ability to the output noise or disturbance.

[0090] Overall, S o can be used as a key performance indicator of the closed-loop system, and directly reflects the response quality of the system. On the other hand, the correlation between the sensitivity peak and the phase margin and the gain margin of the closed-loop system also shows that the smaller the peak value of the sensitivity function is, the higher the stability margin of the closed-loop system is.

[0091] Figure 3 In the formula, the variable superscript k=1, 2; 3, 4. They respectively represent the relevant signals or transfer functions of the X and Y degrees of freedom of the DE end and NDE end radial magnetic bearings (i.e. bearings A and B) of the rotor, and x represents a series of variables related to the rotor displacement. The controller is in Figure 3 The closed-loop control system shown in the figure is added with an excitation signal of different frequencies and an amplitude A as the rotor displacement instruction and outputs a displacement error signal The test range of the frequency f h is 5-300 Hz. Based on the frequency of the sweep and the corresponding sensitivity test results, the detection results of each degree of freedom are plotted, Figure 4 is an example of the sensitivity function sweep results.

[0092] In order to make the sensitivity function more accurate, the present application optimizes the sensitivity function, as follows:

[0093]

[0094] In the formula, G ij is the transfer function from the jth input variable u j to the ith output variable, s is a complex frequency variable in Laplace transform, u is an input variable of the magnetic suspension bearing system, a ijn is the nth coefficient of the polynomial of the numerator of the transfer function from the jth input variable u j to the ith output variable, and b ijm is the nth coefficient of the polynomial of the denominator of the transfer function from the jth input variable u jmth coefficient of the denominator polynomial of the transfer function to the ith output variable, u i is the ith input variable, and t is time.

[0095] The random disturbance and the uncertainty of system parameters are introduced and described by stochastic process and probability model:

[0096] H = H0+ σH1

[0097]

[0098] where H is the stochastic process and probability model, H0is the baseline value of the magnetic suspension bearing system, σ is the disturbance term of the magnetic suspension bearing system, σ is the standard deviation, indicating the amplitude of the disturbance, and it obeys the normal distribution

[0099]

[0100] where u(t) is the input variable of the magnetic suspension bearing system at time t, K is the gain matrix, r(t) is the reference signal vector at time t, y(t) is the output signal vector at time t, is the jth first nonlinear function of the ith reference signal vector, β ij is the second weight of the jth nonlinear function of the ith reference signal vector, ψ ij is the jth second nonlinear function of the ith reference signal vector, r i is the ith reference signal vector at the integral time variable τ, y i is the ith output signal vector at the integral time variable τ, n is the number of reference signal vectors, m is the number of nonlinear functions, and α ij is the first weight of the jth nonlinear function of the ith reference signal vector.

[0101] The above first and second nonlinear functions are polynomial functions, exponential functions, logarithmic functions, sine functions and cosine functions.

[0102] System overall score criteria:

[0103] In order to comprehensively consider the parameter performance, we designed a heuristic function for multiple optimization objectives to evaluate the performance of the parameters:

[0104] For the frequency response function amplitude or maximum amplitude A at a certain frequency or frequency band, we want to optimize the amplitude as T (the optimization target T should be a realistic target or a value close to the feasible range to prevent the score from being too large after scoring, thereby ignoring other optimization targets), and the weight is W (the weights of multiple targets only need to meet the proportional relationship). In such cases, we can use the following formula to evaluate the system:

[0105] Parameter score =∑W x |T-A|

[0106] In actual testing, we have two optimization targets:

[0107] (1) The frequency response function amplitude of the low frequency band (5 Hz) is -5 dB, and the weight is 2;

[0108] (2) The maximum amplitude of the frequency response function in the middle frequency band (5-300 Hz) is 0 dB, and the weight is 8.

[0109] Thus we can get the heuristic function for the above two optimization targets:

[0110] Parameter score = 2 x |(-5)-A 5Hz + 8 x |0-A 5～300Hz |

[0111] Third stage: Pruning breadth-first search tree

[0112] The main task of the third stage is to efficiently explore the parameter space through the root node data recorded in the second stage to further optimize the performance of the PID controller. In this stage, we not only need to ensure the stability of the system in each degree of freedom, but also need to consider the performance of the sensitivity function of each degree of freedom. The following are the detailed steps of the third stage:

[0113] 1) Initialize search: Before starting the third stage, we first reset the step size of parameter search. As the optimization deepens, gradually reduce the step size and continuously improve the accuracy of the search.

[0114] 2) Prepare parent node queue: Sort all parameter combinations with score results recorded by the expected score calculated from the score drop rate from best to worst, and select several parameter combinations with the best expected score to join the parent node queue.

[0115] 3) Improve the parent node queue: Determine whether the number of elements in the parent node queue is sufficient. If not, select the parameter combination with the smallest score to join the parent node queue. The length of the parent node queue should always be larger than the number of parameter combinations selected by the expected score, to ensure that a certain number of low-score high-quality parameter combinations can be used as parent nodes. The calculation of the score drop rate is shown in formula [3]:

[0116]

[0117] 4) Prepare search queue: For each parameter set in the parent node queue, generate new parameter combinations around it in the parameter plane, and add these new parameter combinations to the search queue.

[0118] 5) Select parameter combination: Select the first parameter set from the search queue to prepare for stability testing and performance evaluation. If this parameter combination has already been tested and recorded, skip this parameter combination to avoid repeated testing.

[0119] 6) Stability test: Perform stability testing on the new parameter combination to determine whether the system can run stably under the new parameter combination without oscillation or divergence. If the test result shows that the system runs stably under the new parameter combination, record this parameter combination as a new stable node and save the test result. Otherwise, remove this parameter combination from the search queue.

[0120] 7) Sensitivity function frequency sweep test: For parameter combinations that pass the stability test, perform frequency response testing to record the system's frequency response and performance score for subsequent analysis and optimization.

[0121] 8) Determine if there are elements left in the search queue: Determine if there are any untested parameter combinations left in the search queue. If so, continue to select new parameter combinations from the search queue for testing and verification, i.e., repeat steps 5 to 7.

[0122] 9) Switch freedom: If all parameter combinations have been tested, switch to the next freedom and repeat steps 2 to 8 until all parameters of all freedoms have been optimized and verified, and start building the next layer of search tree.

[0123] 10) Determine if the maximum search level has been reached: Determine if the current search level has reached the pre-set maximum search level. If not, continue parameter search and verification; if so, end the third phase.

[0124] 11) End the third phase: When all parameter combinations of all freedoms have been tested and the search level reaches the pre-set maximum level, the third phase ends. At this time, all test data scores are arranged from small to large to determine the optimal parameter combination of the system in each freedom.

[0125] Through the third phase of pruned breadth-first search tree, we can not only further optimize the performance of the PID controller, but also ensure the stability and reliability of the system under different working conditions. The flowchart of this phase is shown in Figure 5 ​

[0126] Embodiment 2

[0127] As Figure 6 shown, the embodiment of the application further provides a magnetic suspension bearing system control parameter self-tuning system, comprising:

[0128] The signal acquisition module is configured to acquire a rotor displacement signal of the magnetic suspension bearing system.

[0129] The preliminary suspension module is configured to realize preliminary suspension of the rotor according to the rotor displacement signal, with a steady-state position and a response speed as indexes.

[0130] The self-tuning module is configured to further optimize the control parameters with a sensitivity function peak value as an optimization target, and to realize self-tuning of the control parameters by using a breadth-first pruning search tree algorithm.

[0131] Specifically, the method comprises:

[0132] A plane is divided into grids by combining control parameters of the PID controller, wherein the plane comprises a Kp-Kd plane, Kp is a proportional link coefficient of the PID controller, and Kd is a differential link coefficient.

[0133] The nodes of the grids are tested and scored.

[0134] The node with the highest score is selected by using a pruning algorithm, and the node after pruning is used as a parent node to derive child nodes in all directions to explore the parameter plane in more detail.

[0135] Specifically, the method comprises:

[0136] The search range and the step length of the control parameters of the PID controller are set.

[0137] The range of the control parameters of the PID controller is divided into a plurality of grid points according to the step length, and all nodes of the grids are added to a search queue.

[0138] The first control parameter combination is selected from the search queue for testing, and it is determined whether the selected control parameter combination can make the rotor reach a stable suspension state.

[0139] For the control parameter combination that passes the stability test, the control parameter combination is recorded as a stable node, and the control parameter combination is removed from the search queue.

[0140] Specifically, the method comprises:

[0141] For each degree of freedom, the control parameter combination that passes the stability test is taken as a geometric center in a closed space formed on the plane of the respective degree of freedom, and the control parameter combination corresponding to the geometric center is taken as a control parameter combination of a root node of a search tree.

[0142] The frequency sweep test of the sensitivity function is performed on the control parameter combination of the root node, and the score of the control parameter combination of the root node and the amplitude of the sensitivity function are recorded.

[0143] Specifically, the method comprises:

[0144] The PID controller adds the exciting signal of different frequencies as the rotor displacement instruction in the closed-loop control system, and outputs the displacement error signal;

[0145] Based on the frequency sweep and the corresponding sensitivity test results, the detection results of each degree of freedom are plotted.

[0146] The above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0147] In the above-mentioned embodiments of the present application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0148] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the described system embodiments are only schematic. For example, the division of units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between units can be indirect coupling or communication connection through some interface, and can be electrical or other forms.

[0149] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the present embodiment scheme.

[0150] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware, or in the form of software functional unit.

[0151] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only storage medium (ROM, Read-Only Memory), random access storage medium (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, and other media that can store program code.

[0152] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for self-tuning control parameters of a magnetic bearing system, characterized in that: include: Obtaining a rotor displacement signal of a magnetic bearing system; According to the rotor displacement signal, the rotor is initially suspended with steady-state position and response speed as indicators; Taking the peak value of the sensitivity function as the optimization target, the control parameters are further optimized and the breadth-first pruning search tree algorithm is used to realize the self-tuning of the control parameters. The breadth-first pruning search tree algorithm includes: The plane formed by the combination of control parameters of the PID controller with a fixed Ki value is infinitely subdivided into a grid, wherein the plane includes: a Kp-Kd plane, Kp is the proportional link coefficient of the PID controller, and Kd is the differential link coefficient; Test and score the parameters represented by the grid nodes; Use the pruning algorithm to select the node with the highest score, and use the pruned node as the parent node to derive child nodes around it to conduct a more detailed exploration of the parameter plane; Set the search range and step size of the PID controller control parameters; The range of the PID controller control parameters is subdivided into multiple grid points according to the step size, and all grid nodes are added to the search queue; Selecting the first control parameter combination from the search queue for testing and determining whether the selected control parameter combination can enable the rotor to achieve a stable suspension state; For a control parameter combination that passes the stability test, record it as a stable node and remove the control parameter combination from the search queue; For each control parameter combination that passes the stability test for each degree of freedom, the geometric center of the closed space formed on the plane of each degree of freedom is taken, and the control parameter combination corresponding to the geometric center is used as the control parameter combination of the root node of the search tree; Perform a frequency sweep test of the sensitivity function on the control parameter combination of the root node, and record the score of the control parameter combination of the root node and the amplitude of the sensitivity function; The PID controller adds excitation signals of different frequencies as rotor displacement instructions in the closed-loop control system and outputs a displacement error signal; Based on the frequency of the frequency sweep and the corresponding sensitivity test results, the test results of each degree of freedom are plotted.

2. A magnetic bearing system control parameter self-tuning system, characterized in that: include: A signal acquisition module is used to acquire a rotor displacement signal of a magnetic bearing system; A preliminary suspension module, configured to achieve preliminary suspension of the rotor based on the rotor displacement signal, taking steady-state position and response speed as indicators; The self-tuning module is used to further optimize the control parameters with the peak value of the sensitivity function as the optimization target, and adopts the breadth-first pruning search tree algorithm to realize the self-tuning of the control parameters; The breadth-first pruning search tree algorithm includes: The plane formed by the combination of control parameters of the PID controller with a fixed Ki value is infinitely subdivided into a grid, wherein the plane includes: a Kp-Kd plane, Kp is the proportional link coefficient of the PID controller, and Kd is the differential link coefficient; Test and score the parameters represented by the grid nodes; Use the pruning algorithm to select the node with the highest score, and use the pruned node as the parent node to derive child nodes around it to conduct a more detailed exploration of the parameter plane; Set the search range and step size of the PID controller control parameters; The range of the PID controller control parameters is subdivided into multiple grid points according to the step size, and all grid nodes are added to the search queue; Selecting the first control parameter combination from the search queue for testing and determining whether the selected control parameter combination can enable the rotor to achieve a stable suspension state; For a control parameter combination that passes the stability test, record it as a stable node and remove the control parameter combination from the search queue; For each control parameter combination that passes the stability test for each degree of freedom, the geometric center of the closed space formed on the plane of each degree of freedom is taken, and the control parameter combination corresponding to the geometric center is used as the control parameter combination of the root node of the search tree; Perform a frequency sweep test of the sensitivity function on the control parameter combination of the root node, and record the score of the control parameter combination of the root node and the amplitude of the sensitivity function; The PID controller adds excitation signals of different frequencies as rotor displacement instructions in the closed-loop control system and outputs a displacement error signal; Based on the frequency of the frequency sweep and the corresponding sensitivity test results, the test results of each degree of freedom are plotted.

Images