Adaptive control method for micro des magneto-locating unit based on improved sliding mode algorithm

Abstract

The application relates to a micro DES magnetic position control unit adaptive control method based on an improved sliding mode algorithm, which solves the technical problem of how to improve the response speed and precision of the electromagnetic drive flexible joint displacement, and comprises a data feedback and processing part and an improved sliding mode control part. In the data processing and feedback, a capacitive dielectric elastomer sensor is detected and filtered, and the current displacement is obtained after mapping. The displacement and expectation are re-input into the controller for circulation. The improved sliding mode control corrects the dielectric elastomer sensor tension by using an RBF neural network, realizes online optimization of weights by using a composite weight update law composed of a pose state and a sliding mode surface, and further improves the robustness and control precision by using delay interference compensation. The application is suitable for executing dynamic tasks in a complex environment.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of electromagnetic driving flexible joints, in particular to a micro DES magnetostrictive position control unit adaptive control method based on an improved sliding mode algorithm. BACKGROUND

[0002] In the technical field of robot joints, flexible joints have the advantages of good deformation capacity, high safety, multiple degrees of freedom, light weight and the like, and can support robots to complete more complex tasks of higher difficulty.

[0003] Among the driving modes of flexible joints, electromagnetic driving has the advantages of high positioning accuracy, fast response speed, compact structure and the like. The electromagnetic driving flexible joint can be referred to the application patent application with the application number CN116787486A. However, the response speed and accuracy of the displacement of the existing electromagnetic driving flexible joint need to be improved. SUMMARY

[0004] The application is to solve the technical problem of how to improve the response speed and accuracy of the displacement of the electromagnetic driving flexible joint, and provides a micro DES magnetostrictive position control unit adaptive control method based on an improved sliding mode algorithm.

[0005] The improved sliding mode algorithm is a sliding mode control algorithm using delay interference compensation and RBF neural network deep optimization.

[0006] In the data processing and feedback process, the controller provides an input to the power supply unit, and the power supply unit outputs a specified voltage or current to the micro DES magnetostrictive position control unit. The micro DES magnetostrictive position control unit is affected by electromagnetic force to generate driving force, drive the capacitive dielectric elastomer sensor to stretch, and cause the capacitance value of the capacitive dielectric elastomer sensor to change. The sensor capacitance is detected and filtered using a capacitance detection device, interference fluctuations are removed, and the initial capacitance obtained in advance is compared. After filtering and mapping, the current displacement is obtained, and the displacement and expectation are re-input to the controller for circulation.

[0007] The application discloses a micro DES magnetostrictive position control unit adaptive control method based on an improved sliding mode algorithm. The micro DES magnetostrictive position control unit includes a dielectric elastomer sensor and a magnetostrictive driver. The magnetostrictive driver includes a stator, a rotor and a spring shell. The coil is fixed above the support base and is fixed opposite to the sleeve to form the stator. The magnetic ring, the magnetic column and the magnetic block are arranged in series according to the same magnetic pole sequence and form the rotor with the magnetic ring fixed sleeve. The rotor can freely move in the sleeve. The dielectric elastomer sensor has an upper part and a lower part. The upper part is wrapped on the magnetic ring fixed sleeve of the magnetostrictive driver, and the lower part is wrapped on the support base of the magnetostrictive driver. The adaptive control method includes the following steps:

[0008] Step 1, measure the capacitance value of the dielectric elastomer sensor multiple times with the LCR bridge instrument, average the multiple measurement results to calculate the initial capacitance value C0;

[0009] Step 2, construct the RBF neural network and the improved sliding mode controller; the improved sliding mode controller is constructed using a nonlinear sliding surface as the basis and a nonlinear discrete terminal sliding surface:

[0010] s(k) = c1e1(k) + c2e2(k) + c3sig α (e1(k))

[0011] In the formula, s(k) represents the constructed sliding surface, e1(k), e2(k) represent displacement error and speed error respectively, and c1, c2, c3 are constants;

[0012] Step 3, set the desired displacement;

[0013] Step 4, input the desired displacement to the improved sliding mode controller;

[0014] Step 5, the improved sliding mode controller outputs a signal to the power supply unit;

[0015] Step 6, the power supply unit supplies power to the coil of the magnetostrictive actuator;

[0016] Step 7, the mover of the magnetostrictive actuator is displaced to stretch the dielectric elastomer sensor;

[0017] Step 8, the LCR bridge instrument detects the capacitance value Cp1 of the dielectric elastomer sensor;

[0018] Step 9, filter the capacitance value Cp1 through the Kalman filter to obtain the capacitance value Cp2;

[0019] Step 10, subtract the initial capacitance value C0 from the capacitance value Cp2 to obtain the capacitance change value ΔCp;

[0020] Step 11, filter the capacitance change value ΔCp through the Kalman filter, and then map it to obtain the current displacement x1; the mapping process is performed through the function y = kx, where x represents the capacitance change value, y represents the displacement, and k represents the constant coefficient;

[0021] Step 12, the difference between the desired displacement and the current displacement x1 is the displacement error e1(k), which is input to the improved sliding mode controller;

[0022] Step 13, input the current displacement x1 to the RBF neural network:

[0023]

[0024] In the formula, represents the estimated value of the weight between the hidden layer and the output layer, and is a column vector; H j represents the output of the jth hidden layer neuron, c j and b j respectively represent the center and width of the jth neuron, and x1 represents the current displacement of the input;

[0025] obtain a corrected value of the sensor tension

[0026] Step 14, delay interference compensation, at time k, the measured data is fed back to time k-1 through the state equation of the system, the interference d(k-1) at the previous time is estimated, and it is assumed that the interference changes slowly in a short time, so that d(k-1) is regarded as the estimated value of the interference d(k) at the current time; then, a compensation term is added in the control input;

[0027] Step 15, the corrected value of the sensor tension and the interference estimate value obtained by delay interference compensation are input into the improved sliding mode controller to obtain the output u(k), and then the output u(k) is input into the power supply unit, and then the new displacement value is obtained through data feedback and processing, and the closed-loop control is completed.

[0028] Preferably, the RBF neural network includes an input layer, a hidden layer and an output layer, the centers of the hidden layer neurons are determined by the K-means clustering algorithm, and the specific process is as follows:

[0029]

[0030] In the formula, U j represents the jth cluster, |U j | represents the number of samples in U j , u ji represents the sample in U j , D m represents the maximum distance between the two clustering centers, and n represents the number of hidden layer neurons; first, the dielectric elastomer sensor force-position data obtained through the stretching test are preprocessed, the samples for clustering are the sensor force-position data, the sensor force-position data are subjected to the K-means clustering algorithm to obtain key center points, and then the obtained neuron center points and widths are used to fit the sensor force-position data by using the RBF neural network to obtain the initial weight;

[0031] the compound weight update law is as follows:

[0032] ΔW j (k) = -η * c2T * s(k) * H j (x1(k))

[0033] In the formula, ΔWj (k) represents the weight change value at time k, represents the learning rate, c2 is a constant, and T represents the sampling time of the discrete system;

[0034] The RBF neural network is used for fitting the tension of the dielectric elastomer sensor, a composite weight update law composed of a pose state and a sliding mode surface is used for realizing online optimization of the weight.

[0035] The application has the advantages of quick response, accurate positioning, strong robustness, and is suitable for the micro DES magnetostrictive position control unit to perform dynamic tasks in a complex environment.

[0036] An improved sliding mode control with disturbance compensation and RBF neural network is adopted, a nonlinear sliding mode surface is designed to adapt to the nonlinear characteristics of the electromagnetic system, the RBF neural network is used for correcting the tension of the dielectric elastomer sensor, and a composite weight update law composed of a pose state and a sliding mode surface is used for realizing online optimization of the weight, so as to effectively suppress the chattering phenomenon of the sliding mode algorithm.

[0037] Not only can high-precision displacement control be realized, but also strong adaptive ability is possessed, system dynamic changes and external disturbances can be quickly and effectively responded to, so that the control precision, response speed and overall stability are significantly improved.

[0038] Further features and aspects of the present application will be made clear by the following description of specific embodiments, with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a structural schematic diagram of a magnetostrictive driver;

[0040] Figure 2 is a structural schematic diagram of a micro DES magnetostrictive position control unit;

[0041] Figure 3 is a structural schematic diagram of a dielectric elastomer sensor;

[0042] Figure 4 is a front view of a magnetostrictive driver;

[0043] Figure 5 is Figure 4 is a sectional view in the A-A direction of the magnetostrictive driver;

[0044] Figure 6 is a structural schematic diagram of the composition of a magnetostrictive driver, preparation of a dielectric elastomer sensor and connection of the dielectric elastomer sensor on the magnetostrictive driver;

[0045] Figure 7 is a total flowchart of a control method;

[0046] Figure 8 This is a flowchart of the data feedback and processing of the control method;

[0047] Figure 9 This is a flowchart of the adaptive control process based on the sliding mode algorithm;

[0048] Figure 10 These are the response curves of each algorithm in the step test;

[0049] Figure 11 These are the response curves of each algorithm during transient fault testing;

[0050] Figure 12 These are the response curves of various algorithms under a 1.5g load sinusoidal test.

[0051] Figure 13 These are the response curves of various algorithms under different load sinusoidal tests.

[0052] Explanation of symbols in the diagram:

[0053] 1. Magnetoresistive actuator; 2. Dielectric elastomer sensor; 2-1. Upper part; 2-2. Lower part; 3. Miniature DES magnetoresistive positioning unit; 4. Spherical connector; 5. Magnetic ring retaining sleeve; 6. Coil; 7. Sleeve; 8. Support base; 9. Magnetic ring; 10. Magnetic column; 11. Magnetic block; 12. Buffer spring; 13. Protective layer; 14. Copper foil; 15. Electrode layer; 16. Dielectric layer. Detailed Implementation

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] like Figure 1 As shown, the magnetoacter 1 has a conventional structure. For example... Figure 2 As shown, the dielectric elastomer sensor 2 and the magnetostrictive actuator 1 are connected to form a miniature DES magnetostrictive positioning unit 3. Figure 3 This is a schematic diagram of the structure of dielectric elastomer sensor 2.

[0056] like Figure 4 and 5As shown, the magnetic drive 1 is composed of a stator, a rotor and a spring shell, etc. The coil 6 is fixed above the support base 8, and is fixed opposite to the sleeve 7, constituting the stator. The inner diameter of the magnetic ring 9 is larger than the diameter of the magnetic column 10, facilitating the assembly of the two. The magnetic ring 9, the magnetic column 10 and the magnetic block 11 are arranged in series according to the same magnetic pole sequence, and constitute the rotor together with the magnetic ring fixing sleeve 5. The spherical connecting pair 4 serves as the upper fixed point, and preferably a spherical pair connection, and other connection modes can also be selected. The diameter of the lower circular end face of the spherical connecting pair 4 and the upper part inner diameter of the magnetic ring fixing sleeve 5 constitute a transition fit. The upper part inner diameter of the magnetic ring fixing sleeve 5 is slightly larger than the outer diameter of the magnetic ring 9, and the diameter of the lower part through hole of the magnetic ring fixing sleeve 5 is slightly smaller than the inner diameter of the magnetic ring 9, while constituting a transition fit with the magnetic column 10. The magnetic column 10 is sequentially threaded through the lower part through hole of the magnetic ring fixing sleeve 5, the inner hole of the magnetic ring 9, and contacts the upper end face of the magnetic ring 9 together to the lower end face of the spherical connecting pair 4. Selecting the magnetic ring 9 and the magnetic column 10 can significantly improve the electromagnetic driving force. There is a blocking ring at the top end of the sleeve 7, which has a diameter larger than the diameter of the magnetic column 10, and at the same time, the diameter of the magnetic block 11 is larger than the diameter of the blocking ring at the top end of the sleeve 7, which plays a blocking role to prevent the rotor and the stator from separating. A certain gap is left between the magnetic block 11 and the inner wall of the sleeve 7, so as to ensure that the rotor composed of the spherical connecting pair 4, the magnetic ring fixing sleeve 5, the magnetic ring 9, the magnetic column 10 and the magnetic block 11 can move freely in the sleeve 7. There are heat dissipation grooves on the wall of the sleeve 7 to prevent heat accumulation, and at the same time, the bottom of the sleeve 7 is fixed with a buffer spring 12 with a large stiffness coefficient, which is used to reduce the starting current and prevent the impact of falling when a fault occurs. The outer diameter of the sleeve 7 is slightly larger than the outer diameter of the coil 6, and the inner diameter is slightly smaller than the inner diameter of the coil 6, constituting a clearance fit with the coil, and at the same time, the leading part of the coil 6 is connected with the power supply unit. At the same time, the sleeve 7 and the support base 8 constitute a transition fit. The support base 8 can form a fixed structure with bolts and nuts, and different connection modes can also be selected. There is no rigid constraint between the stator and the rotor similar to the bearing, but the driving is carried out through the interaction of the magnetic field between the coil and the magnetic block, which can not only ensure the connection between the stator and the rotor, but also facilitate maintenance and disassembly.

[0057] The materials used for the magnetic ring 9, the magnetic column 10 and the magnetic block 11 can be permanent magnetic materials, including rare earth permanent magnetic materials (neodymium iron boron Nd2Fe14B), samarium cobalt (SmCo), aluminum nickel cobalt (AlNiCo), ferrite permanent magnetic materials, etc. The spherical connecting pair 4, the magnetic ring fixing sleeve 5, the sleeve 7 and the support base 8 can be made of light curing 3D printing technology or non-magnetic material machining.

[0058] The sensor part adopts a capacitive dielectric elastomer sensor (DES), such as Figure 6The dielectric elastomer sensor 2 comprises a dielectric layer 16, electrode layers 15 on both sides of the dielectric layer, copper foils 14 for connection and lead-out located on the electrode layers, and protective layers 13 arranged on the side of each electrode layer away from the dielectric layer. Specifically, the preparation process is as follows: first, lay a protective layer 13, then add a copper foil 14 on the first protective layer 13, then lay an electrode layer 15, the electrode layer adopts a double H shape, one end of the H bottom is in contact with the first added copper foil; lay a dielectric layer; secondly, add a copper foil 14, the copper foil coincides with the other end of the H bottom; secondly, lay an electrode layer 15, the copper foil 14 is in contact with the other end of the H bottom; lay a protective layer; after the sensor is cut along the mold, it can be packaged and integrated with the driver. The copper foil 14 can be extended by welding other wires. The protective layer 13 takes platinum silicone as the base, and different materials can be added according to different requirements to achieve different protection effects. The electrode layer 15 can adopt carbon black dispersion liquid, carbon nanotube dispersion liquid, conductive silver paste and other materials, and different materials can be selected according to the tensile and conductive properties. The material of the dielectric layer 16 can be different high polymer materials, which can be flexibly adjusted according to different capacitance change intervals.

[0059] The dielectric elastomer sensor 2 is wrapped on the magnetic driver 1, referring to Figure 2 and 3 , the upper part 2-1 of the dielectric elastomer sensor 2 is wrapped on the magnetic ring fixing sleeve 5, and the lower part 2-2 is wrapped on the support base 8. Thus, the micro DES magnetic position control unit 3 is formed.

[0060] The control algorithm disclosed in the present application mainly includes two core modules of data feedback and processing part and improved sliding mode control part. In the data processing and feedback link, first, the controller provides accurate input to the power supply unit, which can be a programmable direct current power supply or a PCB power supply system. The power supply unit outputs a specified voltage or current to the coil 6 according to the input, the coil 6 generates an electromagnetic force which interacts with the permanent magnet to generate a corresponding driving force, the mover moves, the magnetic ring fixing sleeve 5 moves to drive the dielectric elastomer sensor 2 to stretch, causing the area of the dielectric elastomer sensor to change and thus causing the capacitance value of the sensor to change. In order to accurately obtain the capacitance value of the sensor, a high-precision capacitance detection device such as an LCR bridge instrument is used to detect the capacitance of the sensor in real time and transmit it, and software filtering is used to deal with environmental noise and interference fluctuations to ensure the reliability of the data. The software filtering adopts Kalman filtering of one-dimensional static system:

[0061]

[0062] In formula (1), represents the prior state estimation at time k, represents the posterior state estimation at time k-1, that is, the measurement value after filtering at time k-1, P(k) represents the prior error covariance matrix at time k k-1 Q represents the covariance of process noise, K k K(k) represents the Kalman gain at time k, used to weigh the prediction and observation. R represents the observation noise, which can be adjusted to control the filtering effect. z k z represents the measurement value of the sensor.

[0063] Reference Figure 7 , 8 , 9, the disclosed adaptive control method of the micro-DES magnetostrictive position control unit based on the improved sliding mode algorithm mainly includes the following steps:

[0064] Step 1, measure the capacitance value of the dielectric elastomer sensor 2 multiple times with the LCR bridge instrument, take the average of the multiple measurement results to calculate the initial capacitance value C0.

[0065] Step 2, construct the RBF neural network and the improved sliding mode controller.

[0066] During the stretching process of the dielectric elastomer sensor 2 driven by the mover displacement of the magnetostrictive driver 1, the dielectric elastomer sensor will exert an additional pulling force on the magnetostrictive driver. Therefore, the RBF neural network is used to correct the sensor pulling force. The RBF neural network used is a three-layer structure, including an input layer, a hidden layer, and an output layer. The centers of the hidden layer neurons are determined by the K-means clustering algorithm, the clusters and centers are initialized, and the Euclidean distance between the sample and the center point is repeatedly adjusted to finally determine the clusters and centers. The specific process is as follows:

[0067]

[0068] In the formula, U j represents the jth cluster, |U j | represents the number of samples in U j , u ji represents the sample in U j , D m represents the maximum distance between the two clustering centers, and n represents the number of hidden layer neurons. First, the dielectric elastomer sensor force-position data obtained through the stretching test is used to pre-process the neural network, and the samples for clustering are the sensor force-position data. The sensor force-position data is subjected to the K-means clustering algorithm to obtain the key center points, and then the obtained neuron center points and widths are used to fit the sensor force-position data using the RBF neural network to obtain the initial weights.

[0069] Compared with the weight iteration mode based on gradient descent method in the traditional RBF neural network, the application designs a compound weight updating law based on the pose state and the sliding mode surface. The updating law dynamically adjusts the network weight by combining the real-time pose state of the system with the sliding mode surface information, realizes the online optimization of the RBF neural network in the control process. This method not only avoids the gradient problem that the traditional gradient descent method may fall into, but also significantly reduces the sliding mode chattering phenomenon, and better adapts to the control demand of the complex nonlinear system. The specific updating law is as follows:

[0070] ΔW j (k)=-η*c2T*s(k)*H j (x1(k))(3)

[0071] In the formula, ΔW j (k) represents the weight change value at the moment k, η represents the learning rate, c2 is a constant, and T represents the sampling time of the discrete system.

[0072] The improved sliding mode controller is constructed, the nonlinear sliding mode surface is used as the basis, the RBF neural network is used to fit the tension of the dielectric elastomer sensor, the compound weight updating law composed of the pose state and the sliding mode surface is used to realize the online optimization of the weight, and the chattering phenomenon of the sliding mode algorithm is effectively suppressed. The coupled disturbance compensation technology effectively deals with the disturbance in the control process, and further improves the control precision and robustness of the system. A more adaptive nonlinear discrete terminal sliding mode surface is constructed:

[0073] s(k)=c1e1(k)+c2e2(k)+c3sig α (e1(k)) (4)

[0074] In the formula, s(k) represents the constructed sliding mode surface, e1(k), e2(k) represent displacement error and speed error respectively, and c1, c2, c3 are all constants.

[0075] Step 3, set the expected displacement.

[0076] Step 4, input the expected displacement to the improved sliding mode controller.

[0077] Step 5, the improved sliding mode controller outputs a signal to the power supply unit.

[0078] Step 6, the power supply unit supplies power to the coil of the magneto-driven actuator 1.

[0079] Step 7, the mover displacement of the magneto-driven actuator 1 drives the dielectric elastomer sensor 2 to stretch.

[0080] Step 8, the LCR bridge instrument detects the capacitance value Cp1 of the dielectric elastomer sensor.

[0081] Step 9, the capacitance value Cp1 is filtered through the Kalman filter to obtain the capacitance value Cp2.

[0082] Step 10, the capacitance value Cp2 is subtracted from the initial capacitance value C0 to obtain the capacitance change value ΔCp

[0083] Step 11, the capacitance change value ΔCp is filtered through the Kalman filter, and then through the mapping to obtain the current displacement x1. The mapping process is through the function y=kx, where x represents the capacitance change value, y represents the displacement, and k represents the coefficient which is a constant value; let x=ΔCp, then the y value can be calculated, which is the current displacement x1.

[0084] Step 12, the difference between the expected displacement and the current displacement x1 is the displacement error e1(k), and the displacement error e1(k) is input into the improved sliding mode controller.

[0085] Step 13, the RBF neural network used in the present application takes the current displacement of the controlled object as input, and outputs the correction value of the sensor tension

[0086]

[0087] In the formula, represents the estimated value of the weight between the hidden layer and the output layer, and is a column vector; H j represents the output of the jth hidden layer neuron, c j and b j respectively represent the center and width of the jth neuron, and x1 represents the current displacement of the input.

[0088] Step 14, the present application uses delay disturbance compensation technology to further improve the control effect. Delay disturbance compensation technology is a technology used in control systems to offset unknown disturbances, which uses historical data to estimate and compensate for disturbances at the current time. Delay disturbance compensation has the advantages of simple implementation and small calculation amount, and is suitable for the application scenarios of the present application which require high real-time performance. The specific implementation method is:

[0089] At time k, the measured data is fed back to time k-1 through the state equation of the system, the disturbance d(k-1) at the previous time is estimated, and it is assumed that the disturbance changes slowly in a short time, so d(k-1) is regarded as the estimated value of the disturbance d(k) at the current time (k time); then, a compensation term is added to the control input to offset the influence of the disturbance on the system. Delay disturbance compensation technology can effectively reduce the influence of unknown disturbances such as external disturbance and model uncertainty on the system performance, and improve the control accuracy and robustness of the system by estimating and compensating the disturbance in real time.

[0090] Step 15, the correction value of the sensor tension And the delay interference compensation obtained together with the interference estimate input improved sliding mode controller, get output u(k), output u(k) input power unit, then data feedback and processing to get new displacement value, complete a closed loop control.

[0091] In order to ensure the performance of closed loop system, the application analyzes the closed loop and steady state performance of control algorithm by Lyapunov method and scalar dynamic system. The Lyapunov method is an important mathematical tool for analyzing system stability, and its core idea is to judge the stability of the system by constructing an energy function (called Lyapunov function). Lyapunov function V(k) is a scalar function, which satisfies V(k) >= 0. By analyzing the time derivative V'(k) of the function, the stability of the system can be judged: if V'(k) < 0, the system is stable; the selected V(k) is as follows:

[0092]

[0093] The application adopts discrete system, and ΔV(k) can be obtained by eliminating high price items, as follows:

[0094]

[0095] In the formula, δ * Indicates the upper limit of the absolute value of the interference mediated derivative.

[0096] The compound weight updating law of RBF neural network is substituted into, and the robust term σ > δ * That is, the system is stable. Consider the scalar dynamic system:

[0097] z(k+1)=z(k)-l1sig α z(k)-l2z(k)+g(k) (8)

[0098] If l1> 0, 0 < l2 < 1, 0 < alpha < 1, and g(k) < gamma, gamma > 0, the state z(k) is always bounded, and there exists a finite number K * > 0, so that:

[0099]

[0100] In the formula, the function Ψ(alpha) is defined as:

[0101]

[0102] Through analysis, it can be obtained that:

[0103]

[0104] Further substituting into the scalar dynamics system, we have

[0105]

[0106] To achieve the highest accuracy, we should choose a = 2 / 3, and the accuracy of the tracking error e1(k) is O(T 3 ). Compared with the nonlinear sliding mode O(T 2 ), the improved sliding mode algorithm using RBF neural network and delayed disturbance compensation has higher steady-state accuracy.

[0107] The above control method can realize fast response, accurate positioning, strong robustness, and is suitable for micro DES magnetostrictive position control unit to execute dynamic tasks in complex environment.

[0108] To further verify the performance of the proposed algorithm, an experimental platform is built using an LCR bridge instrument, a signal generator, a power amplifier, a laser displacement sensor, and a PC to carry out experimental verification.

[0109] The experimental steps are as follows:

[0110] Step one: power supply system connection. Connect the PC and signal generator using the BNS USB connection line, connect the signal generator and power amplifier using the BNS connection line, and confirm that the power amplifier uses external signals. Connect the power amplifier and the coil of the micro magnetic position control unit.

[0111] Step two: detection system connection. Connect the PC and bridge instrument using the USB connection line, and ensure that the bridge instrument is linked to the sensor of the micro magnetic position control unit. At the same time, connect the PC and laser displacement sensor using the USB connection line, and then use the DC stabilized power supply to power the laser displacement sensor.

[0112] Step three: confirm the amplification ratio. Adjust the signal generator to generate a signal, and adjust the amplification ratio of the power amplifier. Fix the appropriate amplification ratio for subsequent experiments.

[0113] Step four: run the test program. Test the communication between the PC and the signal generator, bridge instrument, and laser displacement sensor to ensure that the PC can control the output voltage and receive feedback data from the bridge instrument and laser displacement sensor.

[0114] Step five: formal experiment. Adjust the sampling time to 0.05s, run the experimental program, start the experiment and record the results.

[0115] The experiment adopts step signal test, instantaneous fault simulation, sinusoidal signal test under different loads, and compares the specific performance differences of four algorithms, namely, the fast terminal sliding mode algorithm optimized by RBF neural network and delay disturbance compensation (D-RBF-FTSM), the fast terminal sliding mode algorithm optimized by RBF neural network only (RBF-FTSM), the fast terminal sliding mode algorithm without optimization (FTSM), and the PID algorithm.

[0116] The step test adopts an 8mm step signal, and the response curves of the algorithms are as shown in Figure 10 In addition to the rise time and the settling time, the maximum error, the average error, and the root mean square error are selected from 15s to 18s for evaluation, and the evaluation criteria are as shown below:

[0117] Max Error = max(|el(k)|)

[0118]

[0119] The index data are shown in Table 1, in which the D-RBF-FTSM has the fastest settling time of 2.65s, the lowest maximum error of 0.0419mm, and the lowest average error of 0.0150mm. It exhibits the best comprehensive performance, followed by the RBF-FTSM.

[0120]

[0121] Table 1 Comparison of performances of algorithms in step test

[0122] In order to further compare the fault tolerance of the algorithms, a simulation fault experiment is performed. On the basis of the 8mm step signal, an instantaneous open-circuit fault is simulated at 17s, and the response of each algorithm is observed. The response of each algorithm is as shown in Figure 11 .

[0123] In the fault response experiment, the recovery time, the overshoot, the average error from 18s to 25s, the root mean square error, and the average error from 21s to 25s are selected as the evaluation criteria.

[0124] The index data are shown in Table 2. Combined with Figure 11 and the average error in the two time periods in Table 2, it can be seen that after the power supply is restored, the D-RBF-FTSM rapidly falls from the overshoot state to the steady state, the average error of the entire 18s to 25s is 0.1463mm, and the average error of the 21s to 25s is reduced to 0.0359mm, which is reduced to the order of magnitude of the above 8mm step test. It exhibits excellent fault tolerance.

[0125]

[0126] Table 2 Comparison of performances of algorithms in instantaneous fault test

[0127] To test the dynamic performance of each algorithm. The experiment sets x r = sin(0.05*T) + 5 signal for testing, while in order to evaluate the robustness of each algorithm, set the empty load, 0.5g load, 1g load, 1.5g load for testing. Select 15s-140s for evaluation, in addition to the maximum error, average error, root mean square error, additional phase difference, delay to evaluate the tracking ability of the algorithm, while taking the rolling standard deviation to measure the chattering degree of the sliding mode algorithm.

[0128]

[0129] Δθ = arg(X[it]) - arg(Y[ie])

[0130]

[0131] The standards of phase difference, delay, and rolling standard deviation are shown above. Where x, y, it, ie represent the expected signal, the feedback signal of the sensor, and the corresponding frequency components of the two, respectively. avg represents the average, K represents different nodes, and in the data sample, every 30 nodes are taken, μ represents the average of the samples between two nodes. Select 1.5g load for specific analysis, the response curve of each algorithm is shown in Figure 12 , and the evaluation index data is shown in Table 3.

[0132]

[0133] Table 3 Performance comparison of each algorithm under 1.5g load sinusoidal test

[0134] Combining Figure 12 With the data in Table 3, it can be seen that the FTSM algorithm has strong robustness, but its Astd error index 0.1054mm is far higher than the other three algorithms, exposing severe chattering phenomenon. On the contrary, RBF-FTSM significantly suppresses chattering, but loses robustness, with a phase difference of -0.2926 rad and a delay of -2.75s, which is much higher than the FTSM algorithm and D-RBF-FTSM algorithm. The D-RBF-FTSM algorithm has better performance than the other three algorithms, especially the average error 0.1642mm, Astd error 0.0548mm, and phase difference -0.0747rad. It shows excellent robustness.

[0135] The response of the remaining three groups of load experiments is shown in Figure 13 , and the evaluation data is shown in Table 4.

[0136]

[0137] Table 4 Performance comparison of each algorithm under different loads

[0138] Experiments prove that the RBF neural network used in the application effectively utilizes the composite weight update rate to suppress the chattering phenomenon of the sliding mode, uses the delay interference compensation to ensure the robustness of the control system, and improves the response speed and control accuracy of the system.

[0139] The above only describes the preferred embodiments of the present application and is not intended to limit the present application, and various modifications and changes can be made to the present application by those skilled in the art.

Claims

1. A micro-DES magnetostrictive position control unit adaptive control method based on an improved sliding mode algorithm, characterized in that, The micro-DES magnetic position control unit comprises a dielectric elastomer sensor and a magnetic driver, the magnetic driver comprises a stator, a mover and a spring shell, a coil is fixed above a support base and is fixed opposite to a sleeve to form the stator, a magnetic ring, a magnetic column and a magnetic block are arranged in series in the same magnetic pole sequence to form the mover, the mover can freely move in the sleeve, the dielectric elastomer sensor is provided with an upper part and a lower part, the upper part is wrapped on the magnetic ring fixing sleeve of the magnetic driver, and the lower part is wrapped on the support base of the magnetic driver, and the adaptive control method comprises the following steps: Step 1, the capacitance value of the dielectric elastomer sensor is measured multiple times by using an LCR bridge instrument, and the initial capacitance value C0 is calculated by taking the average value of the multiple measurement results; Step 2, an RBF neural network and an improved sliding mode controller are constructed; the improved sliding mode controller is constructed by using a nonlinear sliding mode surface as a basis and adopting a nonlinear discrete terminal sliding mode surface: s(k) = c1e1(k) + c2e2(k) + c3sig α (e1(k)) In the formula, s(k) represents the constructed sliding mode surface, e1(k) and e2(k) represent displacement error and speed error respectively, and c1, c2 and c3 are all constants; Step 3, the expected displacement is set; Step 4, the expected displacement is input to the improved sliding mode controller; Step 5, the improved sliding mode controller outputs a signal to the power supply unit; Step 6, the power supply unit supplies power to the coil of the magnetic driver; Step 7, the displacement of the mover of the magnetic driver drives the dielectric elastomer sensor to stretch; Step 8, the LCR bridge instrument detects the capacitance value Cp1 of the dielectric elastomer sensor; Step 9, the capacitance value Cp1 is filtered through a Kalman filter to obtain the capacitance value Cp2; Step 10, the capacitance value Cp2 is subtracted from the initial capacitance value C0 to obtain the capacitance change value ΔCp; Step 11, the capacitance change value ΔCp is filtered through a Kalman filter, and then the current displacement x1 is obtained after mapping; the mapping process is performed through a function y=kx, in which x represents the capacitance change value, y represents the displacement, and k represents the constant coefficient; Step 12, the difference between the expected displacement and the current displacement x1 is the displacement error e1(k), and the displacement error e1(k) is input to the improved sliding mode controller; Step 13, the current displacement x1 is input to the RBF neural network: wherein represents the estimated value of the weight between the hidden layer and the output layer, and is a column vector; H j represents the output of the jth hidden layer neuron, c j and b j respectively represent the center and width of the jth neuron, and x1 represents the current displacement of the input; Obtaining a corrected value of the sensor pull force Step 14, delay interference compensation, at k time, the measurement data is fed back to k-1 time through the state equation of the system, the interference d(k-1) at the previous time is estimated, and it is assumed that the interference changes slowly in a short time, d(k-1) is regarded as the estimated value of the interference d(k) at the current time; subsequently, a compensation term is added in the control input; Step 15, correction value of sensor pulling force The output u(k) is input to the power supply unit, and then the data feedback and processing obtain a new displacement value, completing the closed-loop control.

2. The micro-DES magnetic position controlled unit adaptive control method based on the improved sliding mode algorithm according to claim 1, characterized in that, The RBF neural network comprises an input layer, a hidden layer and an output layer, the hidden layer neuron center is determined through a K-means clustering algorithm, and the specific process is as follows: In the formula, U j represents the jth cluster, |U j represents the number of samples in U j represents the number of samples in U ji represents the sample in U j represents the sample in U m represents the maximum distance between the cluster centers, and n represents the number of hidden layer neurons; first, the dielectric elastomer sensor force position data obtained through the stretching test is used to preprocess the neural network, the samples for clustering are the sensor force position data, the sensor force position data is subjected to K-means clustering algorithm to obtain key center points, then the obtained neuron center points and width are used to fit the sensor force position data using the RBF neural network to obtain the initial weight; A compound weight updating law is performed: ΔW j (k) = -η * c2T * s(k) * H j (x1(k)) In the formula, ΔW j (k) represents the weight change value at time k, η represents the learning rate, c2 is a constant, and T represents the sampling time of the discrete system. While fitting the tension of the dielectric elastomer sensor by using the RBF neural network, the compound weight updating law composed of the pose state and the sliding mode surface is used to realize online optimization of the weight.

Images