Feed system LuGre friction model parameter identification method and system based on improved Gaussian group optimization algorithm, computer readable storage medium and computer program product

Abstract

The invention belongs to the field of system parameter identification, and relates to a feeding system LuGre friction model parameter identification method and system based on an improved Gaussian group optimization algorithm, a computer readable storage medium and a computer program product, and the method comprises the steps: (1) exciting a feeding system by using an identification excitation signal, and collecting operation data of the feeding system; the operation data are the speed, the acceleration and the corresponding load current of the feeding system under the excitation of the identification excitation signal; (2) dividing parameters of the LuGre model into sliding parameters and pre-sliding parameters; (3) constructing a loss function for identifying the sliding parameter and the pre-sliding parameter; and (4) based on the operation data in the step (1), minimizing the loss function of the sliding parameter by using a Gaussian group optimization algorithm, and minimizing the loss function of the pre-sliding parameter by using the Gaussian group optimization algorithm to obtain the sliding parameter and the identification result of the pre-sliding parameter. The method has obvious advantages in identification precision and usability.

Description

Technical Field

[0001] The present invention belongs to the field of system parameter identification, and more specifically, relates to a method, system, computer-readable storage medium and computer program product for identifying parameters of a LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm. Background Art

[0002] Friction is one of the main factors affecting the tracking performance of machine tool feed systems. Establishing an accurate friction model is the key to model-based friction compensation methods. Parametric models are the mainstream friction models in academic research and engineering applications, and are mainly divided into static friction models and dynamic friction models. Among them, dynamic friction models have been widely used in high-precision control scenarios because they can describe dynamic friction characteristics such as pre-slip, friction lag, and friction hysteresis. As one of the famous dynamic friction models, the LuGre model can describe almost all friction characteristics, so it has received widespread attention.

[0003] Although the LuGre model has strong expressive power, it has a large number of parameters that need to be identified. Due to the difficulty in parameter identification, the LuGre model is greatly limited in engineering applications. Existing identification methods generally adopt the strategy of first identifying the static parameters of the model based on steady-state data, and then identifying the dynamic parameters of the model through displacement and friction data in the pre-sliding stage. This identification method has the problems of low identification accuracy and difficulty in engineering application.

[0004] To this end, the present invention proposes a parameter identification method of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm, which has obvious advantages in identification accuracy and ease of use. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a method, system, computer-readable storage medium and computer program product for identifying the parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm, the purpose of which is to solve the problems of difficulty in identifying the existing LuGre model parameters and low identification accuracy.

[0006] To achieve the above object, according to one aspect of the present invention, a method for identifying parameters of a LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm is provided, comprising the following steps:

[0007] (1) using the identification excitation signal to excite the feeding system and collect operating data of the feeding system; the operating data is the speed, acceleration and corresponding load current of the feeding system under the excitation of the identification excitation signal;

[0008] (2) Divide the LuGre model parameters into sliding parameters P slide ={Fs ,F c ,v s ,σ 2} and pre-slip parameter P stick ={σ 0 ,σ 1}; where F c is the Coulomb friction force, F s is the maximum static friction, v s is the Stribeck velocity, σ 2 is the viscous damping coefficient, σ 0 is the bristle stiffness, σ 1 is the bristle damping coefficient;

[0009] (3) Construct the loss function L for identifying sliding parameters and pre-sliding parameters 1 and L 2 ;

[0010] (4) Based on the running data of step (1), we first use the Gaussian swarm optimization algorithm to minimize the loss function L 1 , get the sliding parameter P slide Then substitute the above identification results into the loss function L 2 , using Gaussian swarm optimization algorithm to minimize the loss function L 2 , get the pre-sliding parameter P stick The identification result.

[0011] Furthermore, the identification excitation signal in step (1) is a velocity signal formed by concatenating truncated and translated Gaussian functions, and its mathematical expression is as follows:

[0012]

[0013] Where v(t) is the identification excitation signal at time t, A and ∑ are the amplitude and variance of the Gaussian function, respectively, T v is the period of v(t), v 1 (t), v 2 (t) are two Gaussian functions after truncation and translation, n = 0, 1, 2, ...

[0014] Furthermore, the friction force of the feed system is calculated using the operating data according to formula (4):

[0015]

[0016] Among them, F f is the friction force, K t is the motor torque constant, I is the load current, J is the equivalent total inertia of the feed system, θ is the motor shaft angle, is the second-order derivative of θ, and h is the lead of the screw.

[0017] Furthermore, the model described in step (2) is a LuGre model, and the mathematical expression is as follows:

[0018]

[0019] Where F is the friction force, v is the relative motion speed of the contact surface, z is the bristle deformation, g(v) is used to describe the Stribeck characteristics, and sgn() is the sign function;

[0020] The parameters to be identified in the LuGre model include F c 、F s 、v s , σ 0 , σ 1 and σ 2 Furthermore, through parameter sensitivity analysis, when the system enters the macro sliding state, the friction force and {F s ,F c ,v s ,σ 2} is strongly correlated with {σ 0 ,σ 1} is not strongly correlated, and the model can be equivalent to the Stribeck model; therefore, the friction parameter is divided into the sliding parameter P slide ={F s ,F c ,v s ,σ 2}, pre-sliding parameter P stick ={σ 0 ,σ 1};

[0021] Let φ be the LuGre model, then

[0022] F f =φ(v|P slide ,P stick ) (8)

[0023] When the system enters the sliding state,

[0024] F f_slide =φ slide (v slide |P slide ) (9)

[0025] F f_slide is the friction force in sliding state, φ slide is the LuGre model under sliding state, v slide is the relative movement speed of the contact surface in the sliding state.

[0026] Furthermore, the loss function described in step (4) is constructed as follows:

[0027] Identify the sliding parameter P slide The loss function L 1 :

[0028]

[0029] Where M is {v slide ,F f_slide} data length, To estimate the friction force, the superscript (t) indicates time t;

[0030] Then construct the method for identifying the pre-slip parameter P stick The loss function L 2

[0031]

[0032] Where K is {v,F f} data length, is the total friction force.

[0033] Furthermore, the update strategy of four hyperparameters in the Gaussian swarm optimization algorithm is improved; the four hyperparameters are the individual covariance matrix coefficients P c1 , random update probability P c2 , learning update probability P c3 and mutation probability P c4 ; In each iteration process, adaptive adjustment is performed according to equations (27) to (30):

[0034]

[0035] Among them, max_iter is the maximum number of iterations, c 1 is the covariance factor, c 2 is a random factor, c 3 is the learning factor, c 4 is the variation factor, c 1 、c 2 、c 3 and c 4 are constants, β 1 , β 2 , β 3 and β 4 is a constant, and the superscript q is the current iteration number.

[0036] According to another aspect of the present invention, a feed system LuGre friction model parameter identification system based on an improved Gaussian group optimization algorithm is provided, comprising a memory, a processor and a computer program stored in the memory, the processor executing the computer program to implement the steps of the feed system LuGre friction model parameter identification method based on the improved Gaussian group optimization algorithm as described in any of the preceding items.

[0037] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program / instructions are stored. When the computer program is executed by a processor, the steps of the method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm as described in any of the preceding items are implemented.

[0038] According to another aspect of the present invention, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm as described in any of the preceding items.

[0039] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art.

[0040] (1) The LuGre model parameter identification method based on the improved Gaussian group optimization algorithm of the present invention divides the parameters into sliding parameters and pre-sliding parameters, thereby reducing the complexity of the parameter identification problem. The sliding parameters and pre-sliding parameters are identified in sequence by the improved Gaussian group optimization algorithm, and the identification is automatically realized without manual intervention.

[0041] (2) The improved Gaussian group optimization algorithm used in the LuGre model parameter identification method proposed in the present invention has strong optimization performance and can achieve higher precision identification. Compared with the Gaussian group optimization algorithm, the improved hyperparameter adaptive update strategy can better balance the algorithm exploration and utilization capabilities, avoid falling into the local optimum, and have stronger utilization capabilities in the later iteration, thereby improving the convergence accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of a method for identifying parameters of a LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm in a preferred embodiment of the present invention;

[0043] Figure 2 is an identification excitation signal in a preferred embodiment of the present invention;

[0044] Figure 3 is the friction force of the machine tool feed system in the preferred embodiment of the present invention under the excitation of the identification excitation signal;

[0045] Figure 4This is the prediction effect of the identified LuGre model on the actual friction force in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] like Figure 1 As shown, the feed system LuGre friction model parameter identification method based on the improved Gaussian group optimization algorithm of the present invention mainly includes the following steps:

[0048] (1) Using the identification excitation signal to excite the feeding system and collect the feeding system operation data;

[0049] (2) Divide the model parameters into sliding parameters and pre-sliding parameters;

[0050] (3) Construct loss function;

[0051] (4) Using improved Gaussian group optimization algorithm to identify sliding parameters and pre-sliding parameters;

[0052] Further, such as Figure 2 As shown, the identification excitation signal described in step (1) is a velocity signal composed of truncated and translated Gaussian functions, and its mathematical expression is as follows:

[0053]

[0054] Where v(t) is the identification excitation signal at time t, A and ∑ are the amplitude and variance of the Gaussian function, respectively, T v is the period of v(t), v 1 (t), v 2 (t) are two Gaussian functions after truncation and translation, n = 0, 1, 2, ...

[0055] Furthermore, in order to avoid overfitting, Gaussian signals with different amplitudes and frequencies can be mixed as the final excitation signal for recognition.

[0056] Furthermore, the operating data in step (1) is the speed, acceleration and corresponding load current of the feed system under the excitation of the identification excitation signal. Using the operating data, the friction force of the feed system can be calculated according to formula (4), and the friction force is shown in Figure (3).

[0057]

[0058] Among them, F f is the friction force, K t is the motor torque constant, I is the load current, J is the equivalent total inertia of the feed system, θ is the motor shaft angle, and h is the lead of the screw.

[0059] Furthermore, the model described in step (2) is a LuGre model, and the mathematical expression is as follows:

[0060]

[0061] Where F is the friction force, v is the relative velocity of the contact surface, z is the bristle deformation, g(v) is used to describe the Stribeck characteristics, and sgn() is the sign function.

[0062] The parameters to be identified in the LuGre model include F c 、F s 、v s , σ 0 , σ 1 and σ 2 Furthermore, through parameter sensitivity analysis, it can be seen that when the system enters the macro sliding state, the friction force and {F s ,F c ,v s ,σ 2} is strongly correlated with {σ 0 ,σ 1} is not strongly correlated, and the model can be equivalent to the Stribeck model; therefore, the friction parameter is divided into the sliding parameter P slide ={F s ,F c ,v s ,σ 2}, pre-sliding parameter P stick ={σ 0 ,σ 1};

[0063] Let φ be the LuGre model, then

[0064] F f =φ(v|P slide ,P stick ) (8)

[0065] When the system enters the sliding state,

[0066] F f_slide =φ slide (v slide |P slide ) (9)

[0067] Ff_slide is the friction force in sliding state, φ slide is the LuGre model under sliding state, v slide is the relative motion speed of the contact surface under sliding state. Further, the loss function described in step (3) is constructed as follows:

[0068] First, use a quasi-truncated Gaussian signal to excite the system and obtain the velocity v and friction force F f Then select the data between the maximum value point of the speed curve and the zero point (sliding state data), and record it as {v slide ,F f_slide}. slide and Substitute the friction force model φ into the sliding stage slide , we get the estimated friction force As shown in formula (10).

[0069]

[0070] Then construct the sliding parameter P for identification slide The loss function L 1

[0071]

[0072] Next, all the speed and friction data {v,F f} and sliding parameter identification results Substitute into formula (8) to estimate the total friction force

[0073]

[0074] Then construct the method for identifying the pre-slip parameter P stick The loss function L 2

[0075]

[0076] Where X is {v,F f} data length, M is {v slide ,F f_slide} data length, is the initial friction force in the sliding state.

[0077] Furthermore, the process of identifying the sliding parameters and pre-sliding parameters in step (4) is as follows: using an improved Gaussian swarm optimization algorithm to minimize the loss function L 1 , get the sliding parameter P slide The identification result is then used to minimize the loss function L using the improved Gaussian swarm optimization algorithm. 2, get the pre-sliding parameter P stick The identification result.

[0078] Furthermore, the improved Gaussian swarm optimization algorithm described in step (4) is an improved version of the Gaussian swarm optimization algorithm. The algorithm mainly includes initialization, fitness calculation, individual update, individual mutation and hyperparameter update.

[0079] Furthermore, the Gaussian swarm optimization algorithm takes the Gaussian function as an individual, as shown in formula (14).

[0080]

[0081] Among them, G(X|μ,∑) is the Gaussian function individual, K=(x 1 ,x 2 ,...,x D ) is a D-dimensional random variable, D represents the number of parameters to be identified, x 1 ~x D Represents the corresponding parameter to be identified, μ∈R D is the mean vector, representing a set of possible parameter identification results, is the covariance matrix, T is the matrix transpose operation; R D is a D-dimensional real number space, is a D-dimensional positive definite matrix.

[0082] Furthermore, by linearly iterating multiple individuals in the algorithm population and then normalizing them, a superimposed Gaussian probability function P(X)(SGPF) can be obtained, as shown in formula (15).

[0083]

[0084] z=∑ p ∫w p G p (X) (16)

[0085] Among them, Z is the normalization factor, w p is the weight of the pth individual, G p (X) is the pth individual;

[0086] Furthermore, the specific steps for initializing the Gaussian swarm optimization algorithm are as follows:

[0087] For the identification problem of D parameters, assume that the identification space is Ω. A certain number of Gaussian probability functions are randomly generated as individuals, and all individuals constitute the population. The mean μ of the Gaussian function is p and the covariance matrix Σ p Initialize according to equation (17) and equation (18) respectively.

[0088]

[0089] Where n is the population size, lb and ub are D-dimensional vectors consisting of the lower and upper bounds of Ω, respectively, and r is a D-dimensional random vector with a value of [0,1]. is the Hadamard product, P c1 ∈[0,1] is the covariance matrix coefficient, a j ,b j (j=1,2,...,D) are the lower and upper bounds of the j-th dimension of Ω.

[0090] Furthermore, the specific steps for calculating individual fitness are as follows:

[0091] First, let the mean of the pth individual in the qth iteration be The individual fitness is calculated according to equations (19) and (20).

[0092]

[0093] in, is the objective function value of the p-th individual at the q-th iteration, F obj () is the optimization objective function, Fit() is the fitness function, is the fitness of the p-th individual in the q-th iteration.

[0094] Furthermore, the specific steps for individual update are as follows:

[0095] First, the probability Select an individual from the population, denoted by Then, we get The specific process is shown in equations (21) to (23).

[0096]

[0097]

[0098] in, for The mean vector of for The covariance matrix of sample({G p})withprobability Indicates G p Based on discrete probability distribution Sampling, Gaussion_sample() is Gaussian sampling;

[0099] The above update strategy is called Gaussian-update. In order to improve the exploration and utilization capabilities of the algorithm, the algorithm also uses random-update and learning-update as auxiliary strategies. c2 A D-dimensional vector is randomly selected from Ω with probability ∈[0,1] as the new individual mean.

[0100]

[0101] Among them, uniform_sample() is uniform sampling, and rand() means generating random numbers in [0,1].

[0102] The learning update is to add the mean generated by the Gaussian update to P c3 ∈[0,1], the probability of randomly moving a distance to the current optimal individual.

[0103]

[0104] in, is the current optimal mean, and r is a random number in [0,1].

[0105] Furthermore, the specific steps of individual mutation operation are as follows:

[0106] From the population with mutation probability P c4 ∈[0,1] selects an individual, then randomly selects a dimension k∈{1,2,...,D}, and performs mutation operation on the individual’s covariance matrix according to formula (26).

[0107]

[0108] in, is the element in the kth row and kth column of the covariance matrix, a k and b k are the lower and upper bounds of the kth dimension of Ω, and r is a random number in [0,1].

[0109] Furthermore, the algorithm hyperparameter update is mainly to balance the exploration and utilization capabilities of the algorithm. c1 , random update probability P c2 , learning update probability P c3 and mutation probability P c4 Adaptive adjustments are performed according to equations (27) to (30) respectively.

[0110]

[0111] Among them, max_itre is the maximum number of iterations, c1 is the covariance factor, c 2 is a random factor, c 3 is the learning factor, c 4 is the variation factor, c 1 、c 2 、c 3 and c 4 are constants, β 1 , β 2 , β 3 and β 4 is a constant, and the superscript q is the current iteration number.

[0112] Furthermore, after the algorithm is initialized, fitness calculation, individual update, individual mutation and hyperparameter update will be repeated until the termination condition is met. The termination condition can be set to the number of iterations. The algorithm records the optimal identification result of each iteration during the cycle, that is, the mean μ of the individual with the highest fitness. best .

[0113] In a preferred embodiment, the Gaussian swarm algorithm minimizes the loss function L 1 The process is as follows:

[0114] T1: Set the parameters of the Gaussian swarm algorithm, including: population size n, maximum number of iterations max_iter, covariance factor c 1 , random factor c 2 , learning factor c 3 , mutation factor c 4 , update factor β 1 , β 2 , β 3 and β 4 .

[0115] T2: Set the current iteration step to 1;

[0116] T3: Calculate the individual covariance matrix coefficient P according to equations (27) to (30) c1 , random update probability P c2 , learning update probability P c3 and mutation probability P c4 ;

[0117] T4: Initialize the population according to equations (17) and (18);

[0118] T5: Calculate individual fitness according to formula (20);

[0119] T6: Record the optimal individual mean, i.e. the optimal parameter identification result;

[0120] T7: Generate a random number from 0 to 1. If the random number is less than P c2, update the individual mean according to formula (24), otherwise update the individual mean according to formula (22);

[0121] T8: Generate a random number from 0 to 1. If the random number is less than P c3 , update the individual mean according to formula (25), otherwise execute T9;

[0122] T9: Generate a random number from 0 to 1. If the random number is less than P c4 , update the individual covariance matrix according to formula (26), otherwise execute T10;

[0123] T10: Update the individual covariance matrix coefficient P according to equations (27) to (30) c1 , random update probability P c2 , learning update probability P c3 and mutation probability P c4 ;

[0124] T11: The current iteration step increases by 1;

[0125] T12: If the current iteration number is less than max_iter, execute T5 to T11, otherwise output the optimal individual mean and end the optimization.

[0126] Then, the Gaussian group algorithm is used again to calculate L according to steps T1 to T12. 2 Minimize to identify P stick .

[0127] In order to further demonstrate the correctness and superiority of the feed system LuGre friction model parameter identification method based on the improved Gaussian group optimization algorithm of the present invention, the LuGre model parameter identification experiment was carried out in the simulation environment and on the actual machine tool feed axis. The LuGre model was built in MATLAB, and Table 1 shows the model parameter settings and identification range.

[0128] Table 1 Parameter settings and identification range of LuGre model

[0129]

[0130] Table 2 shows the identification results of the improved Gaussian group optimization algorithm (MGSO) and the Gaussian group optimization algorithm (GSO). It can be seen from the table that the parameter identification method of the LuGre friction model of the feed system based on the improved Gaussian group optimization algorithm can realize the accurate identification of friction parameters, and the identification accuracy is higher than that of the Gaussian group algorithm.

[0131] Table 2 LuGre model simulation identification results and relative error (%)

[0132]

[0133]

[0134] Figure 4 The identification effect of the LuGre friction model parameter identification method of the feed system based on the improved Gaussian group optimization algorithm on the actual machine tool feed axis is demonstrated. As can be seen from the figure, the identified LuGre model can accurately describe the friction characteristics of the system.

[0135] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying parameters of LuGre friction model of feed system based on improved Gaussian group optimization algorithm, characterized in that: The steps include: (1) using the identification excitation signal to excite the feeding system and collect operating data of the feeding system; the operating data is the speed, acceleration and corresponding load current of the feeding system under the excitation of the identification excitation signal; (2) Divide the LuGre model parameters into sliding parameters P slide ={F s ,F c ,v s ,σ2} and the pre-slip parameter P stick ={σ0,σ1}; where F c is the Coulomb friction force, F s is the maximum static friction, v s is the Stribeck velocity, σ2 is the viscous damping coefficient, σ0 is the bristle stiffness, and σ1 is the bristle damping coefficient; (3) constructing loss functions L1 and L2 for identifying sliding parameters and pre-sliding parameters; (4) Based on the running data of step (1), the improved Gaussian swarm optimization algorithm is first used to minimize the loss function L1 to obtain the sliding parameter P slide Then substitute the above identification results into the loss function L2, use the improved Gaussian group optimization algorithm to minimize the loss function L2, and get the pre-sliding parameter P stick The identification result.

2. The method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm according to claim 1, characterized in that: The identification excitation signal described in step (1) is a velocity signal composed of truncated and translated Gaussian functions, and its mathematical expression is as follows: Where v(t) is the identification excitation signal at time t, A and ∑ are the amplitude and variance of the Gaussian function, respectively, T v is the period of v(t), v1(t) and v2(t) are two Gaussian functions after truncation and translation, n=0,1,2,… 3. The method for identifying parameters of LuGre friction model of feed system based on improved Gaussian group optimization algorithm according to claim 1, characterized in that: Using the operating data, the friction force of the feed system is calculated according to formula (4): Among them, F f is the friction force, K t is the motor torque constant, I is the load current, J is the equivalent total inertia of the feed system, θ is the motor shaft angle, is the second-order derivative of θ, and h is the lead of the screw.

4. The method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm according to claim 3, characterized in that: The model described in step (2) is the LuGre model, and the mathematical expression is as follows: Where F is the friction force, v is the relative motion speed of the contact surface, z is the bristle deformation, g(v) is used to describe the Stribeck characteristics, and sgn() is the sign function; The parameters to be identified in the LuGre model include F c 、F s 、v s , σ0, σ1 and σ2.

5. The method for identifying parameters of LuGre friction model of feed system based on improved Gaussian group optimization algorithm according to claim 4, characterized in that: Through parameter sensitivity analysis, when the system enters the macro sliding state, the friction force and {F s ,F c ,v s ,σ2} has a strong correlation, and has a weak correlation with {σ0,σ1}. The model can be equivalent to the Stribeck model; therefore, the friction parameter is divided into the sliding parameter P slide ={F s ,F c ,v s ,σ2}, pre-sliding parameter P stick ={σ0,σ1}; Let φ be the LuGre model, then F f =φ(v|P slide ,P stick ) (8) When the system enters the sliding state, F f_slide is the friction force in sliding state, φ slide is the LuGre model under sliding state, v slide is the relative movement speed of the contact surface in the sliding state.

6. The method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm according to claim 5, characterized in that: The loss function described in step (4) is constructed as follows: Identify the sliding parameter P slide The loss function L1 is: Where M is {v slide ,Ff _slide } data length, To estimate the friction force, the superscript (t) indicates time t; Then construct the method for identifying the pre-slip parameter P stick The loss function L2 is: Where K is {v,F f } data length, is the total friction force.

7. The method for identifying parameters of LuGre friction model of feed system based on improved Gaussian group optimization algorithm according to claim 6, characterized in that: The update strategy of four hyperparameters in the Gaussian swarm optimization algorithm is improved; the four hyperparameters are the individual covariance matrix coefficient P c1 , random update probability P c2 , learning update probability P c3 and mutation probability P c4 ; In each iteration process, adaptive adjustment is performed according to equations (27) to (30): Among them, max_iter is the maximum number of iterations, c1 is the covariance factor, c2 is the random factor, c3 is the learning factor, c4 is the variation factor, c1, c2, c3 and c4 are all constants, β1, β2, β3 and β4 are constants, and the superscript q is the current number of iterations.

8. A feed system LuGre friction model parameter identification system based on an improved Gaussian group optimization algorithm, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the feed system LuGre friction model parameter identification method based on the improved Gaussian group optimization algorithm as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm as described in any one of claims 1 to 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for identifying parameters of the LuGre friction model of a feed system based on an improved Gaussian group optimization algorithm as described in any one of claims 1 to 7 are implemented.

Images