Modeling method and system for load-related giant magnetostrictive actuator

Abstract

The invention discloses a modeling method and system for a load-related giant magnetostrictive actuator. The modeling method comprises the steps that multi-dimensional data of the giant magnetostrictive actuator under different load working conditions are acquired and preprocessed; a threshold function and a weight coefficient function of a Play operator in the classic PI model are improved, a Sigmoid function is introduced into the Play operator, and an initial improved PI model is obtained; and identifying parameters to be determined in the initial improved PI model based on a particle swarm optimization algorithm and the preprocessed multi-dimensional data to obtain a final improved PI model. According to the modeling method of the load-related giant magnetostrictive actuator, the hysteresis nonlinearity of the actuator can be described, the output displacement change generated by the load change can be reflected in real time, and a reliable theoretical basis is provided for accurate control and performance optimization of the giant magnetostrictive actuator under different load conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of giant magnetostrictive actuator modeling, and in particular relates to a modeling method and system for a load-dependent giant magnetostrictive actuator. Background Art

[0002] Giant magnetostrictive actuators, with their high energy density and fast response, have broad application prospects in precision instruments, aerospace, and other fields. However, their output characteristics are significantly affected by the load. When the load changes, the actuator's output performance, such as displacement and force, will change. Traditional modeling methods have difficulty accurately describing such complex load-related nonlinear characteristics. Although the Prandtl-Ishlinskii (PI) model is often used to describe hysteresis nonlinearity, it is insufficient for describing the characteristics of giant magnetostrictive actuators under load and cannot meet the needs of high-precision control and performance optimization of actuators in practical applications.

[0003] Conventional modeling methods for giant magnetostrictive actuators only consider the hysteresis characteristics of the giant magnetostrictive material, but fail to account for the fact that load changes during actual operation, leading to changes in the actuator's output characteristics. Under actual operating conditions, varying loads not only affect the actuator's output displacement and force but also alter the mechanism of its internal magnetostrictive effect, significantly changing the actuator's dynamic response characteristics. However, traditional modeling methods treat the load as a fixed parameter or a simple external disturbance, failing to accurately reflect the actuator's complex nonlinear behavior when the load changes. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a load-dependent modeling method and system for giant magnetostrictive actuators to solve the above problems in the prior art.

[0005] To achieve the above objectives, the present invention provides a load-dependent modeling method for a giant magnetostrictive actuator, comprising:

[0006] Acquire multi-dimensional data of giant magnetostrictive actuators under different load conditions and perform preprocessing;

[0007] The threshold function and weight coefficient function of the Play operator in the classic PI model are improved, and the Sigmoid function is introduced into the Play operator to obtain the initial improved PI model;

[0008] Based on the particle swarm optimization algorithm and the preprocessed multi-dimensional data, the parameters to be determined in the initial improved PI model are identified to obtain the final improved PI model.

[0009] Optionally, set the load increment parameter to increase the load range from no load to the giant magnetostrictive actuator, and collect multi-dimensional data, wherein the input current signal of the actuator is collected by the current sensor, the corresponding output displacement signal is collected by the displacement sensor, and the load force is monitored in real time by the force sensor.

[0010] Optionally, a median filter is used to process the multi-dimensional data under different load conditions and then perform a normalization operation to complete the preprocessing.

[0011] Optionally, the process of obtaining the initial improved PI model includes: constructing a load-related threshold function and a load-related weight coefficient function, improving the threshold function and weight coefficient function of the Play operator in the classic PI model, introducing a Sigmoid function to improve the Play operator, and obtaining the initial improved PI model based on the load-related threshold function, the load-related weight coefficient function and the improved Play operator.

[0012] Optionally, the load-related threshold function and the load-related weight coefficient function are in the form of cubic polynomials.

[0013] Optionally, an input of a Sigmoid function is constructed based on the input current signal, and the threshold function is adjusted so that the threshold changes dynamically according to changes in the input current signal.

[0014] Optionally, the parameters to be determined are the parameters to be determined of a load-related threshold function, a load-related weight coefficient function, and a Sigmoid function; a set of parameters is randomly generated as initial particles, and a random speed is assigned to each particle, the fitness value of each particle under the current parameter value is calculated, and based on the fitness value, the individual extreme value of each particle and the global extreme value of the entire particle group are updated; in each iteration, the particle adjusts its speed and position according to the individual extreme value and the global extreme value until the preset number of iterations is reached or the fitness value change of the global extreme value is less than the set minimum value to obtain the optimal parameter combination.

[0015] The present invention also provides a load-dependent giant magnetostrictive actuator modeling system, comprising:

[0016] A data acquisition module is used to obtain multi-dimensional data of the giant magnetostrictive actuator under different load conditions;

[0017] A preprocessing module, configured to perform a normalization operation on the multi-dimensional data after processing using a median filter;

[0018] The model building module is used to improve the threshold function and weight coefficient function of the Play operator in the classic PI model and introduce the Sigmoid function into the Play operator to obtain the initial improved PI model;

[0019] The parameter identification module is used to identify the parameters to be determined in the initial improved PI model based on the particle swarm optimization algorithm and the preprocessed multi-dimensional data to obtain the final improved PI model.

[0020] The present invention also provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0021] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0022] Compared with the prior art, the present invention has the following advantages and technical effects:

[0023] This paper innovatively incorporates load factors into the Prandtl-Ishlinskii model and makes targeted improvements to the Play operator, enriching the theoretical framework for modeling giant magnetostrictive actuators. The modeling method and parameter optimization strategy proposed in this paper provide new ideas and technical references for the modeling and control of other intelligent material systems with hysteresis characteristics, such as piezoelectric actuators and magnetorheological devices, and promote technological development and theoretical progress in related fields.

[0024] The load-dependent modeling method of the giant magnetostrictive actuator provided by the present invention can not only characterize the hysteretic nonlinearity of the actuator itself, but also reflect the output displacement changes caused by load changes in real time, providing a reliable theoretical basis for the precise control and performance optimization of the giant magnetostrictive actuator under different load conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0026] Figure 1 A technical roadmap for an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the Play operator in an embodiment of the present invention;

[0028] Figure 3 This is a sigmoid function curve diagram of an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of an improved Play operator according to an embodiment of the present invention;

[0030] Figure 5 4 is a parameter identification flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0031] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0032] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0033] Example 1

[0034] like Figure 1 As shown, this embodiment provides a load-dependent modeling method for a giant magnetostrictive actuator, including:

[0035] Step 1: Signal acquisition and preprocessing;

[0036] High-precision current sensors, displacement sensors, and force sensors are used to comprehensively collect multi-dimensional data of the magnetostrictive actuator under different load conditions. Among them, the current sensor is responsible for collecting the input current data of the actuator, the displacement sensor synchronously obtains the corresponding output displacement signal, and the pressure sensor monitors the load force borne by the actuator in real time. To ensure that the collected data can fully reflect the characteristics of the actuator under actual working conditions, the input current signal and load are reasonably set. At the same time, the load range is gradually increased from no-load to the rated load of the actuator, and the load increment is set to 1N (which can be adjusted according to actual conditions). During the collection process, dspace is used to monitor the waveform and value of the collected data in real time. By observing the stability of the waveform and the rationality of the value, the continuity and accuracy of data collection are ensured.

[0037] The collected current and displacement signals are processed using the median filter algorithm and normalization method, which is specifically achieved through the following mathematical formula:

[0038] (1) Median filter algorithm;

[0039] Median filtering is a nonlinear filtering method that removes noise by replacing the value of each point in a signal sequence with the median of all the points in its neighborhood. For a one-dimensional signal sequence x(n), where n = 1, 2, 3, ... N (N is the signal length), and the median filter window size is 2k + 1 (k is a positive integer), the median-filtered signal y(n) can be calculated using the following steps:

[0040] (1) For the nth point in the signal sequence, select the signal value in the window with n as the center and a length of 2k+1, that is,

[0041] x(nk),x(n-k+1),…,x(n),…,x(n+k-1),x(n+k)(1)

[0042] (2) Sort the 2k+1 signal values ​​in the window in ascending order to obtain the sorted sequence x sorted (i), i=1,2,…,2k+1.

[0043] (3) Take the middle value of the sorted sequence as the output value of the point after filtering, that is, y(n) = x sorted (k+1).

[0044] Through median filtering, the impulse noise in the signal can be effectively suppressed, the edge and detail information of the signal can be retained, and the purity of the signal can be improved.

[0045] (2) Normalization method;

[0046] Normalization is the process of uniformly mapping a signal to a specific interval (in this case, the interval [0,1]). Its purpose is to eliminate the influence of differences between different signals and facilitate subsequent analysis and processing. For an input signal x(n), its normalized signal z(n) can be calculated using the following formula:

[0047]

[0048] Where Xmin and Xmax are the minimum and maximum values ​​in the signal x(n), respectively. Using this formula, the minimum value in the original signal is mapped to 0, the maximum value is mapped to 1, and the other values ​​are mapped to the interval [0, 1] according to the corresponding proportions.

[0049] Step 2: Build a load-dependent improved PI model

[0050] (1) Classic PI model

[0051] Let's first review the classic PI model. The Prandtl-Ishlinskii (PI) model is a commonly used model for describing hysteretic nonlinear characteristics. It is based on the superposition of a series of basic hysteresis operators. This model is usually composed of a finite number of linear Play operators with different weight coefficients.

[0052] like Figure 2 The Play operator diagram shown in the figure shows that v represents input, w represents output, and r represents threshold. The mathematical expression of the Play operator can be expressed as:

[0053] w=max(vr,min(v+r,w)) (3)

[0054] From this we can get the expression of the classic PI model can be expressed as:

[0055]

[0056] Where p represents the weight coefficient, P0V(t) represents the linear part of the PI model, which is a linear function, and the right part is the nonlinear part of the PI model, which is composed of the weighted superposition of the Play operator. To facilitate the computer solution, the integral expression (4) can be discretized as:

[0057]

[0058] (2) load-related improved PI model;

[0059] In order to address the limitations of traditional models in practical applications and meet the needs of accurate modeling and control of giant magnetostrictive actuators under complex working conditions, a load-related improved PI model is introduced. By establishing a link between the load and the model parameters, the input-output relationship of the actuator can be more accurately characterized, thereby improving the accuracy and reliability of the model.

[0060] In practical applications, the hysteresis characteristics of magnetostrictive actuators are significantly affected by load. In order to more accurately describe the input-output relationship of the actuator under different load conditions, we improve the classic PI model and change the threshold r in the model to i And the weight coefficient p is expressed as a function of the load L.

[0061] (1) Construction of load-related threshold function;

[0062] In the modeling process of the giant magnetostrictive actuator, in order to accurately characterize the influence of load on its hysteretic characteristics, the threshold of the Play operator in the classic Prandtl-Ishlinskii model is improved to a function of load. A cubic polynomial is used to construct the load-dependent threshold function. First, the function form is determined:

[0063] Set the threshold r of the i-th Play operator i The functional relationship with the load L is in the form of a cubic polynomial:

[0064] r i (L) = a i0 +a i1 L+a i2 L 2 +a i3 L 3 (6)

[0065] Among them, L represents load correlation, which is a positive real number, r i (L) represents the threshold value related to the load L; a i0 、ai1 、a i2 、a i3 are the coefficients to be determined, reflecting the degree and pattern of the load's impact on the threshold; i = 1, 2, 3, ... n, where n is the number of Play operators. Cubic polynomials can flexibly fit the complex nonlinear relationship between load and threshold to a certain extent. Compared to lower-degree polynomials, they have stronger expressive power and, compared to higher-degree polynomials, better balance model complexity and fitting accuracy while avoiding overfitting.

[0066] (2) Construction of load-related weight coefficient function;

[0067] When constructing the improved Prandtl-Ishlinskii model, in addition to constructing the load-related threshold function, it is also necessary to construct the load-related weight coefficient function. Similar to the construction of the load-related threshold function, we use the weight coefficient p in the classic Prandtl-Ishlinskii model i Expressed as a function of load L i (L). By establishing this functional relationship, the effect of load on the hysteresis characteristics of the magnetostrictive actuator can be described more accurately.

[0068] When constructing the improved Prandtl-Ishlinskii model of giant magnetostrictive actuator, a cubic polynomial is used to construct the load-related weight coefficient function, which can better fit the complex nonlinear relationship between load and weight coefficient.

[0069] Set the weight coefficient P of the i-th Play operator i The functional relationship with the load L is in the form of a cubic polynomial:

[0070] p i (L) = b i0 +b i1 L+b i2 L 2 +b i3 L 3 (7)

[0071] Where Pi(L) represents the weight coefficient related to the load L, b i0 、b i1 、b i2 、b i3 are the coefficients to be determined, reflecting the degree and pattern of the load's influence on the weight coefficient; i = 1, 2, 3, ... n, where n is the number of Play operators. Cubic polynomials can flexibly simulate various load-weight coefficient trends and offer an advantage in balancing model complexity and fitting accuracy.

[0072] (3) Improved Play operator;

[0073] In traditional Play operators, the threshold is usually fixed or only related to the load, like the original threshold function r i (L) = a i0 +a i1 L+a i2 L 2 +a i3 L 3 , which is only related to the load L. This means that no matter how the input signal v(t) changes, the threshold value is always a fixed value determined by the load or changes according to a fixed load-related law. However, in practical applications, the characteristics of systems such as giant magnetostrictive actuators often change with different input signals, and the Play operator with a fixed threshold value cannot accurately describe such complex characteristics. After introducing the Sigmoid function, the threshold value of the Play operator can change dynamically according to the input signal v(t), allowing the Play operator to exhibit different behaviors under different input signals. This allows the improved Play operator to more accurately describe the complex characteristics of systems such as giant magnetostrictive actuators, improving the model's ability to fit the actual system and the accuracy of its description.

[0074] In the improved Play operator, the Sigmoid function plays a key role in dynamically adjusting the threshold. The following is a detailed explanation:

[0075] Characteristics of the Sigmoid function;

[0076] The expression of the Sigmoid function is Its function graph shows an S shape, such as Figure 3 As shown, it has the following properties: Range: The range of the function is (0, 1). As x approaches negative infinity, σ(x) approaches 0; as x approaches positive infinity, σ(x) approaches 1. Monotonicity: The sigmoid function is monotonically increasing, meaning that as the input x increases, the function value σ(x) also increases.

[0077] The mechanism of the Sigmoid function in improving the Play operator;

[0078] Dynamically adjust the threshold: In the improved Play operator, the input signal v(t) is used to construct the input of the Sigmoid function, that is, Here, k and b are parameters to be determined. By adjusting the values ​​of k and b, the shape and position of the Sigmoid function can be changed.

[0079] Adjusted threshold: It will change dynamically with the change of input signal v(t).

[0080] Influencing state variable updates: adjusted thresholds Will be applied to the improved Play operator state variable w i In the update rule of (t), the improved sigmoid function is as follows Figure 4 As shown:

[0081]

[0082] The introduction of the Sigmoid function allows the Play operator's threshold to dynamically change based on the input signal v(t), allowing the Play operator to exhibit different behaviors under different input signals. This allows the improved Play operator to more accurately describe the complex characteristics of systems such as giant magnetostrictive actuators, improving the model's ability to fit and accurately describe real-world systems.

[0083] (4) Improved load-related PI model

[0084] Based on the classic Prandtl-Ishlinskii (PI) model, the model is improved by introducing load-related variables. The improved model can more accurately describe the hysteresis characteristics of the giant magnetostrictive actuator under different load conditions.

[0085] Based on the load-related threshold function and weight coefficient function, and the play operator with the sigmoid function as the envelope function, the expression for the output of the improved Prandtl-Ishlinskii model is:

[0086]

[0087] In order to facilitate the use of computers to solve the problem, the integral expression can be discretized as:

[0088]

[0089] Where k = 1, 2, ..., N. i = 1, 2, ..., m, where m is the number of improved Play operators and N is the number of input data.

[0090] Step 3:

[0091] Through the above description, we need to identify the improved Prandtl-Ishlinskii model, which involves many parameters to be determined. The threshold function a i0 ,a i1 ,a i2 ,a i3 , b in the weight coefficient function i0 ,b i1 ,b i2 ,b i3, as well as k and b in the Sigmoid function, a total of 10 parameters. Once these 10 parameters are identified, the expression of the load-dependent PI model is established.

[0092] The following describes in detail how to perform parameter identification. Figure 5 shown.

[0093] When using the particle swarm optimization (PSO) algorithm for parameter identification in the modified Prandtl-Ishlinskii model, the algorithm simulates the foraging behavior of a flock of birds and treats each parameter combination to be identified as a particle in the search space. Particles have two properties: position and velocity. Each particle continuously adjusts its position in the search space to find the optimal parameter combination. Particles adjust their velocity and position based on their own historical optimal position (individual extremum) and the historical optimal position of the entire particle swarm (global extremum). The following is a detailed description of the process:

[0094] (1) Initialization: There are 10 parameters to be identified in the improved Prandtl-Ishlinskii model, such as b in the weight coefficient function. i0 ,b i1 ,b i2 ,b i3 , a in the threshold function i0 ,a i1 ,a i2 ,a i3 , as well as k and b in the Sigmoid function.

[0095] Define a particle swarm with N particles. The position of the i-th particle in the d-dimensional (corresponding to 10 parameters) space is represented by x i =(x i1 ,x i2 ,…,x i10 ), velocity is represented by v i =(v i1 ,v i2 ,…,v i10 ).

[0096] Randomly initialize the positions and velocities of particles:

[0097] Position x id Usually it is randomly generated within the feasible range of the parameter, for example, for parameter k, if its value range is [k min ,k max ], then x id ~U(k min ,k max )(U represents uniform distribution). Speed ​​v id It is also randomly initialized within a certain range, such as v id ~U(vmin ,v max ).

[0098] (2) Calculate fitness: For each particle position (i.e., a set of parameter values), substitute it into the improved Prandtl-Ishlinskii model and calculate the model output y m (k).

[0099] Assume that the actual measured data is y d (k), k=1,2,…,M (M is the number of data samples), define the fitness function f(x i ), commonly used mean square error (MSE):

[0100]

[0101] (3) Update individual extreme values ​​and global extreme values:

[0102] Individual extreme value: For each particle i, record the fitness f(x i ), and compared with the particle's previous optimal fitness (individual extreme value) pbest i The corresponding position p i =(p i1 ,p i2 ,…,p i10 ) comparison. If f(x i )<f(p i ), then update the individual extreme value, that is, p i =x i , pbest i =f(x i ).

[0103] Global extreme value: Compare the individual extreme values ​​of all particles, find the one with the smallest fitness, and set its corresponding position as the global extreme value position g=(g1,g2,…,g 10 ), recorded as gbest = min{pbest1,pbest2,…,pbest N}.

[0104] (4) Update particle velocity and position The velocity update formula of particle i in dimension d is:

[0105] v id (t+1)=ωv id (t)+c1r 1d (t)(p id -x id (t))+c2r 2d (t)(g d -x id (t)) (12)

[0106] Where t represents the number of iterations, w is the inertia weight, which is used to balance the global search and local search capabilities, and is generally between 0.4 and 0.9. C1 and C2 are learning factors, usually C1 = C2 = 1.5-2.0, which adjust the step size of the particle approaching the individual extreme value and the global extreme value respectively. 1d (t) and r 2d (t) is a random number uniformly distributed in the interval [0,1].

[0107] The position update formula of particle i in dimension d is:

[0108] x id (t+1)=x id (t)+v id (t+1) (13)

[0109] (5) Iteration and termination: Repeat steps 2 to 4, i.e., calculate fitness, update individual extreme values ​​and global extreme values, and update particle speed and position, until the termination condition is met, such as reaching the preset maximum number of iterations T. max , or the fitness value of the global extreme value changes less than a certain minimum value ò in several consecutive iterations.

[0110] When the algorithm terminates, the parameter value corresponding to the global extreme position g is the parameter estimate of the improved Prandtl-Ishlinskii model identified by the particle swarm optimization algorithm. The identified parameters are then applied to the improved Play operator described above to obtain the improved PI model. Through these steps, an improved Play operator based on load-related weight coefficients and thresholds was successfully constructed and applied to the Prandtl-Ishlinskii model, which can more accurately describe the hysteresis characteristics of the system under different loads.

[0111] This embodiment further provides a load-dependent giant magnetostrictive actuator modeling system, including:

[0112] A data acquisition module is used to obtain multi-dimensional data of the giant magnetostrictive actuator under different load conditions;

[0113] A preprocessing module, configured to preprocess the multi-dimensional data by using a median filter and then performing a normalization operation on the multi-dimensional data;

[0114] The model building module is used to improve the threshold function and weight coefficient function of the Play operator in the classic PI model and introduce the Sigmoid function into the Play operator to obtain the initial improved PI model;

[0115] The parameter identification module is used to identify the parameters to be determined in the initial improved PI model based on the particle swarm optimization algorithm and the preprocessed multi-dimensional data to obtain the final improved PI model.

[0116] This embodiment further provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0117] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0118] The load-dependent giant magnetostrictive modeling method in this embodiment makes some changes to the envelope function of the classic PI model, replacing it with a load-dependent improved PI model. Specifically, the key parameters of the play operator in the classic PI model, such as the threshold and weight, are improved to become functions of the load, and the model structure and parameters are dynamically adjusted according to the load size. The improved model is more accurate than the classic PI model. It can not only characterize the hysteretic nonlinearity of the drive itself, but also reflect the output displacement changes caused by load changes in real time, providing a reliable theoretical basis for the precise control and performance optimization of the giant magnetostrictive drive under different load conditions.

[0119] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A load-dependent modeling method for a giant magnetostrictive actuator, characterized in that: The following steps are involved: Acquire multi-dimensional data of giant magnetostrictive actuators under different load conditions and perform preprocessing; The threshold function and weight coefficient function of the Play operator in the classic PI model are improved, and the Sigmoid function is introduced into the Play operator to obtain the initial improved PI model; Based on the particle swarm optimization algorithm and the preprocessed multi-dimensional data, the parameters to be determined in the initial improved PI model are identified to obtain the final improved PI model.

2. The load-dependent giant magnetostrictive actuator modeling method according to claim 1, characterized in that: The load increment parameters are set to increase the load range from no load to the giant magnetostrictive actuator, and multi-dimensional data is collected. Among them, the input current signal of the actuator is collected through the current sensor, the corresponding output displacement signal is collected through the displacement sensor, and the load force is monitored in real time through the force sensor.

3. The load-dependent giant magnetostrictive actuator modeling method according to claim 1, characterized in that: Median filtering is used to process the multi-dimensional data under different load conditions and then normalize them to complete the preprocessing.

4. The load-dependent modeling method of giant magnetostrictive actuator according to claim 1, characterized in that: The process of obtaining the initial improved PI model includes: constructing a load-related threshold function and a load-related weight coefficient function, improving the threshold function and weight coefficient function of the Play operator in the classic PI model, introducing the Sigmoid function to improve the Play operator, and obtaining the initial improved PI model based on the load-related threshold function, the load-related weight coefficient function and the improved Play operator.

5. The load-dependent giant magnetostrictive actuator modeling method according to claim 4, characterized in that: The load-related threshold function and the load-related weight coefficient function are in the form of cubic polynomials.

6. The load-dependent giant magnetostrictive actuator modeling method according to claim 4, characterized in that: The input of the Sigmoid function is constructed based on the input current signal, and the threshold function is adjusted so that the threshold changes dynamically according to the change of the input current signal.

7. The load-dependent modeling method of giant magnetostrictive actuator according to claim 6, characterized in that: The parameters to be determined are the load-related threshold function, the load-related weight coefficient function, and the parameters to be determined of the Sigmoid function; a set of parameters is randomly generated as initial particles, and a random speed is assigned to each particle. The fitness value of each particle under the current parameter value is calculated, and based on the fitness value, the individual extreme value of each particle and the global extreme value of the entire particle swarm are updated; In each iteration, particles adjust their speed and position according to individual extreme values ​​and global extreme values ​​until the preset number of iterations is reached or the fitness value of the global extreme value changes less than the set minimum value to obtain the optimal parameter combination.

8. A load-dependent modeling system for a giant magnetostrictive actuator, characterized in that: include: A data acquisition module is used to obtain multi-dimensional data of the giant magnetostrictive actuator under different load conditions; A preprocessing module, configured to perform a normalization operation on the multi-dimensional data after processing using a median filter; The model building module is used to improve the threshold function and weight coefficient function of the Play operator in the classic PI model and introduce the Sigmoid function into the Play operator to obtain the initial improved PI model; The parameter identification module is used to identify the parameters to be determined in the initial improved PI model based on the particle swarm optimization algorithm and the preprocessed multi-dimensional data to obtain the final improved PI model.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images