A method for identifying transfer function of a steering engine

Abstract

The application discloses a transfer function identification method of a steering engine, and particularly relates to the following steps: (1) taking the to-be-identified transfer function of the steering engine as a to-be-identified model; (2) taking unknown parameters in the to-be-identified transfer function of the steering engine as identification objects of the to-be-identified model in step (1); (3) adopting a genetic algorithm based on fuzzy logic to identify the identification objects of the to-be-identified model in step (2), so as to obtain an identification result, wherein the identification result is the identified unknown parameters; and (4) substituting the identification result in step (3) into the to-be-identified model in step (1), so as to obtain the to-be-identified model with all known parameters, that is, the to-be-identified transfer function of the steering engine is determined. The application can accurately identify the transfer function of the steering engine, and meanwhile, the identification precision and speed are improved.

Description

Technical Field

[0001] This invention relates to a method for identifying the transfer function of a servo motor, belonging to the field of motor system identification technology. Background Technology

[0002] Lightweight electric servo motors, due to their advantages of small size, light weight, and good dynamic performance, are widely used in the development of military products such as torpedoes, decoy mines, and missiles, and are the most important actuators in weaponry. As a servo mechanism of the control system, the servo motor receives control commands from the missile's onboard control system, moves the adjustment mechanism to generate control torque on the subsystem, controls the missile's control surfaces, and achieves the goal of controlling its air attitude and flight trajectory, enabling rapid maneuverability. The control effect of the actuator directly affects the guidance effect of the entire combat system; therefore, obtaining an accurate servo motor transfer function is crucial for the development of the entire combat system.

[0003] Obtaining the accurate transfer function of a controlled object requires the controlled object to be a white box and a large number of accurate values ​​for physical parameters. However, in obtaining the transfer function of a servo motor, although we can deduce its approximate form, many physical parameters cannot be directly obtained, making the servo system a gray box. Therefore, it is difficult to directly derive the accurate transfer function of the servo system theoretically. Thus, it is necessary to perform system identification on the servo system to obtain the parameters in the transfer function that are difficult to confirm, thereby obtaining the accurate form of the transfer function.

[0004] Genetic Algorithm (GA) is a global, parallel algorithm based on evolutionary theory and the natural law of survival of the fittest. It primarily includes three operations: selection, crossover, and mutation. Its key feature is the manipulation of the encoding of the target object, performing a probabilistic search over a vast space, and executing selection, crossover, and mutation operations during the search process. This allows it to automatically accumulate relevant knowledge to adapt to the search process and find the optimal solution, mimicking the natural law of survival of the fittest. Therefore, it has wide applications in artificial intelligence, machine learning, pattern recognition, and is gradually being applied to system identification. However, as a global search algorithm, the computational cost of genetic algorithms is enormous, resulting in slow convergence speed. In other words, there is a trade-off between the algorithm's globality, identification accuracy, and speed; therefore, improving the algorithm's identification speed and accuracy is a pressing issue that needs to be addressed. Summary of the Invention

[0005] The technical problem solved by the present invention is to overcome the shortcomings of the prior art and provide a method for determining the transfer function of a servo motor, which can accurately identify the transfer function of the servo motor and improve the identification accuracy and speed.

[0006] The technical solution provided by this invention is: a method for identifying the transfer function of a servo motor, comprising the following steps:

[0007] (1) The transfer function of the servo motor to be identified is taken as the model to be identified;

[0008] (2) Take the unknown parameters in the transfer function of the servo motor as the identification object of the model to be identified in step (1);

[0009] (3) A genetic algorithm based on fuzzy logic is used to identify the identification object of the model to be identified in step (2) and the identification result is obtained. The identification result is the identified unknown parameter.

[0010] (4) Substitute the identification result of step (3) into the model to be identified in step (1) to obtain the model to be identified in which all parameters are known, that is, determine the transfer function to be identified of the servo motor.

[0011] Preferably, it also includes step (5), in which the model to be identified, obtained in step (4) with all parameters known, is used for ballistic simulation of a projectile with a servo motor.

[0012] Preferably, it also includes step (6), which performs navigation control on the simulated projectile with servo motor, thereby improving navigation accuracy.

[0013] The preferred transfer function to be identified is specifically: Where θ f For the rudder response position, U in τ is the input voltage of the servo motor, s is a complex variable; K and τ are the parameters to be identified.

[0014] Preferably, K and τ are used as the identification objects.

[0015] Preferably, a sinusoidal sweep frequency signal is input into the servo motor, and the actual servo motor response position is output, with the actual servo motor response position serving as the training sample;

[0016] The sinusoidal sweep frequency signal is input into the model to be identified, and the current servo motor response position is output.

[0017] Determine the error between the current servo response position and the actual servo response position, and establish an objective function based on this error; then establish an individual fitness function based on the established objective function.

[0018] Based on the individual fitness function, a mutation operator based on fuzzy logic is established;

[0019] A genetic algorithm based on fuzzy logic is constructed based on the mutation operator based on fuzzy logic and the selection operator and crossover operator in the traditional genetic algorithm.

[0020] Based on the constructed fuzzy logic-based genetic algorithm, the identification object of the identification model is identified, and the identification result is obtained when the objective function converges.

[0021] The advantages of this invention compared to the prior art are:

[0022] (1) This invention introduces the concept of fuzzy logic into the mutation operator, thus constructing a fuzzy logic-based mutation operator. Combined with the selection and crossover operators in traditional genetic algorithms, a fuzzy logic-based genetic algorithm is constructed. The fuzzy logic-based mutation operator can utilize the concept of membership degree in fuzzy logic to transform the individual fitness in the genetic process into mutation intensity, thereby ensuring that in each generation, individuals with higher fitness have lower mutation intensity, and individuals with lower fitness have higher mutation intensity.

[0023] (2) Based on the mutation operator of fuzzy logic, individuals with higher fitness are better preserved during the mutation operation and inherited by the next generation of population, thereby improving the convergence speed of the objective function during the identification process and improving the identification accuracy. Attached Figure Description

[0024] Figure 1 This is a flowchart of the genetic algorithm identification process based on fuzzy logic in this invention.

[0025] Figure 2 This is a comparison chart of the recognition performance of the fuzzy logic-based genetic algorithm of this invention and the traditional genetic algorithm;

[0026] Figure 3 This is the objective function of the genetic algorithm based on fuzzy logic in this invention after 1000 generations;

[0027] Figure 4 The target function for identifying the servo transfer function in the example using the fuzzy logic-based genetic algorithm of this invention;

[0028] Figure 5 A comparison diagram showing the recognition performance of the bit identification example with that of the original system. Detailed Implementation

[0029] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0030] With the development of guided weapons, target missiles will evolve towards multi-space domain and dual-plane maneuverability, thus significantly altering their structure and model parameters. To ensure personnel safety and conserve research funds, ballistic simulation is essential to verify the correctness of the system design. The accuracy of the transfer function of the actuator on the servo-driven target missile directly impacts the simulation quality of the entire ballistic simulation. This invention proposes a fuzzy logic-based genetic algorithm for identifying the servo motor's transfer function. This algorithm can quickly and accurately obtain the servo motor's transfer function, facilitating better correction of target missile control parameters during ballistic simulation and improving the target missile's tactical performance.

[0031] The present invention provides a method for determining the transfer function to be identified of a servo motor, wherein the preferred embodiment comprises the following steps:

[0032] (1) The transfer function of the servo motor to be identified is taken as the model to be identified; the preferred scheme is as follows:

[0033] In ballistic simulation, when the accurate transfer function of the servo motor cannot be obtained, a first-order inertial element is usually chosen to replace the servo motor's transfer function. The most commonly used and preferred form is... Where θ f For the rudder response position, U in Let be the input voltage of the servo motor, and s be a virtual variable. Therefore, the transfer function to be identified is selected as a first-order form, preferably: Where θ f For the rudder response position, U in Let be the input voltage of the servo motor, and s be a dummy variable; K and τ be the parameters to be identified; s is...; Meanwhile, the preferred selection... As the accurate transfer function of the model to be identified, the modeling method and identification effect of this invention are explained by comparing the form of the transfer function identified by the modeling method used in this invention with that of the accurate transfer function.

[0034] (2) The unknown parameters in the transfer function of the servo motor to be identified are taken as the identification objects of the model to be identified in step (1); the preferred scheme is as follows:

[0035] Using K and τ as the identification objects, determine the maximum value K. max Minimum value K min and the maximum value of τ. max Minimum value τ min Individuals are constructed by using decimal floating-point encoding for K and τ.

[0036] (3) A genetic algorithm based on fuzzy logic is used to identify the object of the model to be identified in step (2) to obtain the identification result, which is the identified unknown parameter; the preferred scheme is as follows:

[0037] Input a sinusoidal sweep frequency signal into the servo motor and output the actual servo motor response position. The actual servo motor response position is used as the training sample.

[0038] The sinusoidal sweep frequency signal is input into the model to be identified, and the current servo motor response position is output.

[0039] Determine the error between the current servo response position and the actual servo response position, and establish an objective function based on this error; then establish an individual fitness function based on the established objective function.

[0040] Based on the individual fitness function, a mutation operator based on fuzzy logic is established;

[0041] A genetic algorithm based on fuzzy logic is constructed based on the mutation operator based on fuzzy logic and the selection operator and crossover operator in the traditional genetic algorithm.

[0042] Based on the constructed fuzzy logic-based genetic algorithm, the identification object of the identification model is identified, and the identification result is obtained when the objective function converges.

[0043] The sinusoidal sweep frequency signal is input into the standard transfer function in step (1), preferably in the form of... Where θ f For the rudder response position, U in Let be the input voltage of the servo motor, and s be a dummy variable. The system output when the sweeping signal is input to the standard transfer function in step (1) constitutes the training samples in the modeling process. Define the i-th sample as inputting a sinusoidal sweeping signal to the standard transfer function, then the training samples will have an output of θ. f i Simultaneously, if the same swept frequency signal is input into the model to be identified at the i-th sampling time, then the output of the model to be identified is... The preferred error is:

[0044]

[0045] Therefore, the preferred objective function is:

[0046]

[0047] Where i is the sampling time and N is the number of samples in the training samples.

[0048] Accordingly, the preferred individual fitness function is:

[0049] f = 1 / J

[0050] Therefore, the smaller the error, the smaller the objective function J, and the larger the individual fitness f, the closer the output of the current model to be identified is to the output of the standard transfer function, and the closer the model to be identified is to the standard transfer function.

[0051] The genetic algorithm based on fuzzy logic mainly includes selection, crossover, and fuzzy mutation operations. In this invention, the selection operation uses a proportional selection operator, the crossover operation uses a single-point crossover operator, and a mutation operator based on fuzzy logic is constructed to perform fuzzy mutation operations.

[0052] The optimal selection based on the fuzzy logic mutation operator is defined as follows:

[0053]

[0054]

[0055] Where K and τ are the parameters to be identified, then K(m,g-1) is the identification result of parameter K for the m-th individual and the (g-1)-th generation, and K(m,g) is the identification result of parameter K for the m-th individual and the g-th generation. max K is the maximum value of parameter K. min The minimum value of parameter K; τ(m,g-1) is the identification result of parameter τ for the m-th individual and the (g-1)-th generation, τ(m,g) is the identification result of parameter τ for the m-th individual and the g-th generation, τ max The maximum value of parameter τ, τ min The minimum value of parameter τ; M is the total number of individuals in the sample, G is the maximum number of generations, f u (m,g) represents the variation intensity, and Rand is a random number between [0,1].

[0056] Variation intensity f u (m,g) borrows the concept of fuzzy membership from fuzzy logic, and can define a fuzzy rule as follows: the higher the fitness of an individual, the lower its probability of mutation should be. This fuzzy rule allows individuals with higher fitness to be better preserved in the genetic process of the entire population. Furthermore, since the center and width of the fuzzy set are adjustable, we can use parameter conditions to allow individuals with higher fitness to undergo fewer mutation operations, improving the convergence speed of the entire optimization process and preventing the system from getting trapped in local optima. Mutation intensity f u The preferred definition of (m,g) is:

[0057]

[0058] Among them, f m (m,g) represents the fitness of the m-th individual in the g-th generation, f max σ(g) is the fitness value of the individual with the highest fitness in the g-th generation population, exp is a power function, and the preferred definition of σ(g) is:

[0059] σ(g)=k(f max (g)-fmin (g))

[0060] Among them, f min (g) represents the fitness value of the individual with the lowest fitness in the g-th generation of the population, f max (g) represents the fitness value k of the individual with the highest fitness in the g-th generation population, where k is a constant.

[0061] Pm(m,g) is the coefficient of variation of the m-th individual in the g-th generation, and G is the total number of generations. Before each individual undergoes a mutation operation, a random number Rand is generated, and the coefficient of variation of that individual is calculated. If Rand is greater than the coefficient of variation, a mutation operation based on fuzzy logic is performed; otherwise, no mutation operation is performed.

[0062] Pm(m,g)=0.2-(0.1-0.001)*g / G

[0063] like Figure 1 As shown, after encoding the parameters to be identified, for each generation of individuals, the following operations are performed: individual fitness calculation, selection operation, crossover operation, fuzzy logic-based mutation operation, and objective function calculation. If the objective function value meets the requirements, the identification process can be exited and the identification result can be output; if the objective function requirements are not met, the next generation population is generated, and the objective function calculation, selection, crossover, and fuzzy logic-based mutation operation continue until an individual that meets the objective function requirements is obtained or the maximum number of generations is reached.

[0064] (4) Substitute the identification result from step (3) into the model to be identified from step (1) to obtain the model to be identified in which all parameters are known, thus determining the transfer function to be identified for the servo motor. The preferred scheme is as follows:

[0065] Through step (3), the identification result of K is 1.006 and the identification result of τ is 0.05078. Substituting these into the model to be identified in step (1), the transfer function of the servo motor to be identified can be preferably determined as follows: Where θ f For the rudder response position, U in denoted as the input voltage of the servo motor, and s as a complex variable.

[0066] To better verify the identification effect of the fuzzy logic-based genetic algorithm, the standard transfer function in step (1) was identified using a traditional genetic algorithm. The identification results for K and τ were 1.007 and 0.05807, respectively. Therefore, the transfer function to be identified for the servo motor can be optimally determined as follows:

[0067]

[0068] In terms of identification results, both the genetic algorithm based on fuzzy logic and the traditional genetic algorithm are relatively accurate in identifying the standard transfer function of the model in step (1), but the identification result of the genetic algorithm based on fuzzy logic is closer to the standard transfer function.

[0069] like Figure 2 As shown, the fuzzy logic-based genetic algorithm exhibits a faster objective function descent rate compared to the traditional genetic algorithm, allowing it to reach a steady state more quickly and obtain the identification result. Furthermore, the steady-state objective function value of the fuzzy logic-based genetic algorithm is smaller than that of the traditional genetic algorithm, thus resulting in more accurate identification results.

[0070] The objective function value of the genetic algorithm based on fuzzy logic in the 1000th generation is as follows: Figure 3 As shown, by Figure 3 As can be seen, the objective function converges after 1000 generations without diverging. Therefore, it can be considered that the genetic algorithm based on fuzzy logic converges stably and has strong stability.

[0071] Figures 2 to 4 The results show that the fuzzy logic-based genetic algorithm can identify the transfer function of the servo motor, and compared with the traditional genetic algorithm, it has a faster identification speed and higher identification accuracy. Preferably, it also includes step (5), in which the model to be identified, obtained in step (4) with all parameters known, is used for ballistic simulation of a projectile with a servo motor.

[0072] Preferably, it also includes step (6), which performs navigation control on the simulated projectile with servo motor, thereby improving navigation accuracy.

[0073] Taking a lightweight integrated servo motor as an example, its transfer function is identified. To obtain training samples, the servo motor's response to sinusoidal signals at different frequencies in open-loop mode is required.

[0074] Furthermore, the servo controller no longer collects position feedback signals, converting the closed-loop system into an open-loop system. Then, the controller directly outputs a sinusoidal control signal, making the system input a sinusoidal voltage signal. The sinusoidal signal is in the form of 1.2*sin(2*pi*fNT), where f is the input signal frequency, N is the sampling time (N=0,1,2,3…), and T is the unit sampling time, T=0.5ms. The frequency of the input sine function is changed, with a starting frequency of 0.5Hz, an ending frequency of 8Hz, and a step size of 0.5Hz. For each frequency point, 800 data points are recorded. This yields the training samples for the servo.

[0075] The training sample is identified using the fuzzy logic-based genetic algorithm described in this invention. The model to be identified is selected as a third-order transfer function cascaded with a 1 / s form, where s is a dummy variable. Substituting the identification result into the model to be identified, the transfer function form of the servo motor can be obtained as follows:

[0076]

[0077] Where θ f For the rudder response position, U in Let be the input voltage of the servo motor, and s be a complex variable. The objective function result under the 1000th iteration is as follows: Figure 4 As shown, the identification results are stable and convergent, and there is no divergence.

[0078] To verify the systemic reconstruction effect of the identification results, we selected the output responses of the original servo and the identification system under 0.5Hz and 1Hz sinusoidal inputs for comparative analysis. Figure 5 As shown.

[0079] Because the measurement of the identification samples inherently contains certain errors, and real systems have a large amount of noise and nonlinear factors, the transfer function response error identified from the measured samples will be slightly larger than that from the simulated samples. However, it can still be used as the identification result of the system. The comparison between this transfer function and the output of the identified servo system shows that the method of the present invention can identify the transfer function of the servo system with high identification accuracy.

[0080] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A method for identifying the transfer function of a servo motor, characterized in that... The steps are as follows: (1) The transfer function of the servo motor to be identified is taken as the model to be identified; (2) Take the unknown parameters in the transfer function of the servo motor as the identification object of the model to be identified in step (1); (3) A genetic algorithm based on fuzzy logic is used to identify the identification object of the model to be identified in step (2) and the identification result is obtained. The identification result is the identified unknown parameter. (4) Substitute the identification result of step (3) into the identification model of step (1) to obtain the identification model with all parameters known, that is, determine the identification transfer function of the servo motor. (5) Use the model to be identified, which has known parameters obtained in step (4), for ballistic simulation of a projectile with a servo motor; Input a sinusoidal sweep frequency signal into the servo motor and output the actual servo motor response position. The actual servo motor response position is used as the training sample. The sinusoidal sweep frequency signal is input into the model to be identified, and the current servo motor response position is output. Determine the error between the current servo response position and the actual servo response position, and establish an objective function based on this error; then establish an individual fitness function based on the established objective function. Based on the individual fitness function, a mutation operator based on fuzzy logic is established; A genetic algorithm based on fuzzy logic is constructed based on the mutation operator based on fuzzy logic and the selection operator and crossover operator in the traditional genetic algorithm. Based on the constructed fuzzy logic-based genetic algorithm, the identification objects of the identification model are identified, and the identification result is obtained when the objective function converges. The mutation operator based on fuzzy logic is defined as follows: Where K and τ are the parameters to be identified, K(m,g-1) is the identification result of parameter K for the m-th individual and the (g-1)-th generation, and K(m,g) is the identification result of parameter K for the m-th individual and the g-th generation. max K is the maximum value of parameter K. min The minimum value of parameter K; τ(m,g-1) is the identification result of parameter τ for the m-th individual and the (g-1)-th generation, τ(m,g) is the identification result of parameter τ for the m-th individual and the g-th generation, τ max The maximum value of parameter τ, τ min The minimum value of parameter τ; M is the total number of individuals in the sample, G is the maximum number of generations, and f u (m,g) represents the variation intensity, Rand is a random number between [0,1], and Pm(m,g) is the variation coefficient of the m-th individual in the g-th generation.

2. The method for identifying the transfer function of a servo motor according to claim 1, characterized in that: It also includes step (6), (6) to perform navigation control on the simulated projectile with servo motor.

3. The method for identifying the transfer function of a servo motor according to claim 1, characterized in that: The transfer function to be identified is, specifically: Where θ f For the rudder response position, U in τ is the input voltage of the servo motor, s is a complex variable; K and τ are the parameters to be identified.

4. The method for identifying the transfer function of a servo motor according to claim 3, characterized in that: K and τ are the objects to be identified.

Images