Genetic Algorithm-Based Identification Method for Friction Nonlinearity of Voice Coil Motors

Abstract

The present invention discloses a method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm. By establishing a dynamic model of a voice coil motor servo system containing friction, the genetic algorithm is used to identify the parameters of the S-type friction model. By optimizing the parameters of the friction model, the friction characteristics of the actual system are matched. The effectiveness of the proposed method is verified by simulation experiments, and a comparative analysis is conducted with the traditional method. The research results show that the identification method based on the genetic algorithm can accurately establish the friction nonlinear model of the voice coil motor and improve the control performance and robustness of the system. By establishing the dynamic model and optimizing the friction model parameters, the control performance and robustness of the voice coil motor servo system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of voice coil motor servo system control, and particularly to a method for identifying the friction nonlinearity of a voice coil motor by using a genetic algorithm. Background Art

[0002] Due to its characteristics of high speed, high precision, and high reliability, the voice coil motor is widely used in precision control systems. However, the existence of friction forces leads to the nonlinear characteristics of the system, posing challenges to control design.

[0003] Accurately establishing the friction nonlinear model of the voice coil motor is of great significance for improving the system control performance, reducing vibration and noise, and enhancing the robustness of the system. By deeply understanding the friction phenomenon and establishing an accurate friction model, more effective solutions can be provided for application scenarios such as precision positioning, tracking control, and adaptive control. In addition, the research on the friction nonlinearity of the voice coil motor can also provide reference and inspiration for solving other friction-related problems.

[0004] In the past few decades, research work on friction nonlinear modeling and identification has received extensive attention. Traditional methods include modeling methods based on physical models and identification methods based on experimental data. However, traditional methods usually require a large amount of prior knowledge or a large amount of experimental data, and there are limitations in establishing an accurate friction model. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm to improve the system control performance, reduce vibration and noise, and enhance the robustness of the system.

[0006] To solve the above technical problems, the present invention is implemented as follows:

[0007] The method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm includes the following steps:

[0008] Step S1: Establish the dynamic model of the voice coil motor servo system containing friction and analyze the influence of friction on the system performance;

[0009] Step S2: Use the genetic algorithm to identify the parameters of the S-shaped friction model. By optimizing the parameters of the friction model, make it match the friction characteristics of the actual system to adapt to different voice coil motor servo systems.

[0010] For further optimization, the specific steps of step S1 are as follows:

[0011] Step S1.1: Construct the dynamic model of the voice coil motor servo system containing friction

[0012] Applying the working voltage U to the voice coil motor generates the working current I to drive the voice coil motor to work, and a voltage drop U will be generated on the coil resistance R R , an induced voltage drop U will be generated on the coil inductance L L At the same time, the voice coil motor will generate back electromotive force E when working; according to Kirchhoff's law, the voltage balance equation of the voice coil motor is:

[0013]

[0014] Where v is the speed at which the voice coil motor moves in a straight line, i.e., the speed at which the motor moves in a straight line; k m is the electromagnetic force coefficient of the voice coil motor.

[0015] Combined with the motion characteristics of the voice coil motor servo system, the mechanical motion model of the voice coil motor servo system is simplified and equivalent to a single-degree-of-freedom spring-mass-damper mechanical positioning system; according to Newton's second law, the motion differential equation of the reed-suspended voice coil motor-driven optical image stabilization system is obtained as follows:

[0016]

[0017] Where m is the equivalent mass of the voice coil motor, the camera module load and their attached components; k is the equivalent stiffness of the reed and the system connection; c is the equivalent damping coefficient of the voice coil motor and the camera module; y is the axial displacement of the system, and

[0018] Combine E = NBlv = k m v, k m =NB1, the dynamic model of the voice coil motor servo system is obtained:

[0019]

[0020] Where N is the number of turns of the coil, B is the magnetic induction intensity, l is the effective length of the coil in the magnetic field; k m is the electromagnetic force coefficient of the motor.

[0021] Assume that the current loop has been closed in the servo system, and the dynamics of the current loop can be ignored relative to the mechanical dynamics of the system, that is:

[0022] i≈i ref (0.19)

[0023] i ref Indicates a given current.

[0024] Then the simplified dynamic equation of the voice coil motor servo system is:

[0025]

[0026] Step S1.2: Analyze the influence of friction on system performance

[0027] In the analysis of Step S1.1, friction nonlinearity was not considered, and the friction acting on the system was simply characterized by a linear damping c, which was approximately accurate when the servo system was moving unidirectionally at high speed. However, when the speed of the servo system switches frequently, the friction acting on the system is a non - linear function F f (v) of speed. Based on the Coulomb + viscous friction model, we have:

[0028] F f (v) = Fsgn(v)+F v v(0.21)

[0029] where F is the Coulomb friction force, and F v is the viscous friction coefficient;

[0030] At this time, the dynamic equation of the system is:

[0031]

[0032] Let x1 = y, x2 = v, u = I, and the state - space equation of the voice - coil motor servo system is obtained as:

[0033]

[0034] where k0 = k / m, k v = F v / m, b0 = k m / m, f c = F / m.

[0035] For further optimization, in the friction identification of the voice - coil motor, there are problems such as a large amount of test data and a long test time. In addition, due to the often - existing large noise in the actual system and the fact that the speed of the motor cannot be directly measured, the effect of directly using the identified friction model for friction compensation is not ideal. To address the above problems, this application establishes a practical friction model and then uses the genetic algorithm to identify the entire non - linear system containing this friction model.

[0036] Step S2.1: Establish an S - type friction model

[0037] Most of the existing friction models are discontinuous around the point where the velocity v = 0. Although such models are closer to the physical principle of friction, there are some problems in actual use: 1) The friction when the velocity is close to zero is difficult to measure and identify, and even if it is identified, it is not accurate enough; 2) The frictional force is a non-linear function of the velocity. When compensating for friction, only the estimated value of the velocity can be used. The estimation of the velocity is often not accurate, and due to the large noise in the system, the estimation of the velocity often adds a relatively large noise, which aggravates the inaccuracy of the compensation. Furthermore, the noise causes the estimation of the velocity to frequently switch on both sides of zero velocity near zero velocity, so the friction compensation also frequently switches near zero velocity, resulting in serious chattering and noise phenomena.

[0038] Based on these problems, this application uses an S-shaped friction model to replace the sign function type friction model, which can reduce the chattering and noise phenomena attached to the friction compensation while effectively improving the control performance. The voice coil motor servo system equation based on the hyperbolic tangent function form friction model is:

[0039]

[0040] where f f and a are parameters to be identified.

[0041] Step S2.2: Identify the S-shaped friction based on the genetic algorithm

[0042] There is no effective method for measuring and identifying the S-shaped friction model in the experiment. The effective S-shaped friction model depends on continuous adjustment and testing, which requires a lot of time. Moreover, for different voice coil motor servo systems, the parameters of the S-shaped friction model for achieving optimal friction compensation are also completely different. The genetic algorithm has the ability of global optimization, can automatically obtain and guide the search space, and adaptively adjust the search direction. Therefore, the applicant developed a method for identifying the S-shaped friction based on the genetic algorithm, specifically: encoding all individuals to form an initial population, calculating the fitness of each individual, and judging whether it meets the convergence condition of the genetic algorithm according to the fitness. If it meets, the identification result is output; if not, the selection operator, combination crossover operator, and mutation operator are performed to generate a new solution set population, and gradually evolve to generate better and better approximate solutions. The optimal individual in the last generation population is decoded as the approximate optimal solution.

[0043] For further optimization, before designing the three basic operators of the genetic algorithm, an optimization index needs to be determined - in the genetic algorithm, it is called fitness. The fitness needs to meet the conditions of being single-valued, continuous, and non-negative. In this application, the fitness J k is evaluated based on the standard deviation between the output of the S-shaped friction model and the actual output, that is:

[0044]

[0045] y(ii) is the sampling of y at time ii, where:

[0046]

[0047] For further optimization, in step S2.2, the selection operator is specifically:

[0048] The selection operator is to select superior individuals from the population. The purpose of selection is to directly inherit the optimized individuals to the next generation or generate new individuals through pairing and crossover and then inherit them to the next generation. Here, the roulette wheel selection method is adopted. First, a selection probability is designed:

[0049]

[0050] Obviously, the selection probability is defined here as the ratio of the fitness of a certain individual to the total fitness of the population. The higher the fitness of an individual, the higher the probability of being selected, and the sum of the selection probabilities of all individuals is equal to 1.

[0051] However, in the simulation case, the fitness of the population is within the range of [400 - 600], so the calculated p k is all around 1 / n, and there is no discrimination between the well-performing parameters and the poorly-performing parameters in the selection process, resulting in the non-convergence of the genetic algorithm.

[0052] Therefore, this application defines the selection probability as:

[0053]

[0054] where γ represents an adjustable parameter; J k represents the cost of the k-th individual, and n represents the number of individuals. By subtracting min(J j ), the performance of the model under different parameters can be more clearly reflected. At the same time, the adjustable parameter γ is added to avoid the difference being too obvious and the population quickly converging to the local optimal solution.

[0055] Then, the roulette wheel selection method is used to select superior individuals from the population, and the optimized individuals are directly inherited to the next generation or new individuals are generated through pairing and crossover and then inherited to the next generation to obtain the new population after selection. The roulette wheel selection method is a prior art and will not be elaborated here.

[0056] For further optimization, in step S2.2, the crossover operator is specifically as follows: The linear recombination algorithm in real-value recombination is adopted to exchange genes of two randomly selected individuals in the selected population to generate new gene combinations, and it is expected to group beneficial genes together; First, randomly sort the newly selected population; then, for two adjacent individuals in the new population, determine whether to perform crossover according to the crossover probability P c Decide whether to crossover; For the individuals to be crossed, perform crossover according to the following formula;

[0057] NO i = dO i +(1 - d)O i+1 (0.30)

[0058] NO i+1 = dO i+1 +(1 - d)O i (0.31)

[0059] where, O i is the corresponding parameter of the individual before crossover, NO i is the corresponding parameter of the new individual generated by crossover, and d is a random number between [0, 1].

[0060] For further optimization, in step S2.2, the mutation operator introduces additional randomness to the genetic algorithm to prevent the algorithm from falling into a local optimal solution; specifically, for each individual in the population, determine whether mutation occurs according to the mutation probability P m If mutation occurs, the parameter of the individual becomes a random value within the parameter space.

[0061] Compared with the prior art, the present application has the following beneficial effects:

[0062] 1. The present invention provides a method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm. By establishing a dynamic model of a voice coil motor servo system containing friction, using the genetic algorithm to identify the parameters of the S-shaped friction model, and optimizing the parameters of the friction model to match the friction characteristics of the actual system, the effectiveness of the proposed method is verified through simulation experiments and comparative analysis with traditional methods. Through the above simulation experiments, the effectiveness of the method proposed by the present invention is verified, and the results show that the method can improve the control performance and robustness of the voice coil motor servo system.

[0063] 2. As a global optimization algorithm, the genetic algorithm has good search ability and global convergence in parameter identification problems, can effectively handle nonlinear and multimodal problems, and provides technical support for the control and application of voice coil motor servo systems. Further research can apply this method to actual systems, and conduct experimental verification and optimization improvement to further improve the performance and stability of the system.

[0064] 3. The identification method based on genetic algorithm proposed by the present invention can also provide reference and inspiration for the identification and control problems of other nonlinear systems. Brief Description of the Drawings

[0065] Figure 1 is a simplified diagram of the electromechanical coupling system of the voice coil motor driven optical image stabilization system;

[0066] Figure 2 is the flow chart of the genetic algorithm of the present invention;

[0067] Figure 3 is the diagram of the fitness convergence of the genetic algorithm;

[0068] Figure 4 is the comparison between the genetic algorithm identifying the S-shaped friction model and the experimental debugging of the optimal saturation function friction model;

[0069] Figure 5 is the fitness convergence of the genetic algorithm after optimizing the sampling period;

[0070] Figure 6 is the fitness convergence of the genetic algorithm after adjusting the fitness;

[0071] Figure 7 is the comparison between the genetic algorithm identifying the S-shaped friction model and the experimental debugging of the optimal saturation function friction model. Detailed Embodiment

[0072] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0073] The friction nonlinear identification method of the voice coil motor based on genetic algorithm includes the following steps:

[0074] Step S1: Establish the dynamic model of the voice coil motor servo system with friction and analyze the influence of friction on the system performance; specifically including the following steps:

[0075] Step S1.1: Establish the electromechanical coupling system of the voice coil motor servo system as shown in Figure 1 . Apply the working voltage U to the voice coil motor to generate the working current I to drive the voice coil motor to work, and there will be a voltage drop U R on the coil resistance R, and there will be an induced voltage drop U L on the coil inductance L. At the same time, the voice coil motor will generate a back electromotive force E when working; according to Kirchhoff's law, the voltage balance equation of the voice coil motor is obtained as:

[0076]

[0077] Where v is the speed at which the voice coil motor moves in a straight line, i.e., the speed at which the motor moves in a straight line; k m is the electromagnetic force coefficient of the voice coil motor.

[0078] Combined with the motion characteristics of the voice coil motor servo system, the mechanical motion model of the voice coil motor servo system is simplified and equivalent to a single-degree-of-freedom spring-mass-damper mechanical positioning system, such as Figure 1 As shown in the figure on the right. According to Newton's second law, the motion differential equation of the reed-suspended voice coil motor-driven optical image stabilization system is:

[0079]

[0080] Where m is the equivalent mass of the voice coil motor, the camera module load and their attached components; k is the equivalent stiffness of the reed and the system connection; c is the equivalent damping coefficient of the voice coil motor and the camera module; y is the axial displacement of the system, and

[0081] Combine E = NBlv = k m v, k m =NB1, the dynamic model of the voice coil motor servo system is obtained:

[0082]

[0083] Where N is the number of turns of the coil, B is the magnetic induction intensity, l is the effective length of the coil in the magnetic field; k m is the electromagnetic force coefficient of the motor.

[0084] Assume that the current loop has been closed in the servo system, and the dynamics of the current loop can be ignored relative to the mechanical dynamics of the system, that is:

[0085] i≈i ref (0.35)

[0086] i ref Indicates a given current.

[0087] Then the simplified dynamic equation of the voice coil motor servo system is:

[0088]

[0089] Step S1.2: Analyze the effect of friction on system performance

[0090] In the analysis of step S1.1, the friction nonlinearity was not considered, and the friction suffered by the system was simply characterized by the linear damping c, which is approximately accurate when the servo system moves unidirectionally at high speed. However, when the speed of the servo system switches frequently, the friction suffered by the system is a nonlinear function F f (v) of the speed. Based on the Coulomb + viscous friction model, we have:

[0091] F f (v) = Fsgn(v) + F v v(0.37)

[0092] where F is the Coulomb friction force and F v is the viscous friction coefficient;

[0093] At this time, the dynamic equation of the system is:

[0094]

[0095] Let x1 = y, x2 = v, u = I, and the state - space equation of the voice - coil motor servo system is obtained as:

[0096]

[0097] Among them, k0 = k / m, k v = F v / m, b0 = k m / m, f c = F / m.

[0098] Step S2: For the friction identification of the voice - coil motor, there are problems such as a large amount of test data and a long test time. In addition, due to the large noise often existing in the actual system and the inability to directly measure the speed of the motor, the effect of directly using the identified friction model for friction compensation is not ideal. To address the above problems, this application uses a genetic algorithm to identify the parameters of the S - type friction model, and by optimizing the parameters of the friction model, it makes the friction characteristics match the actual system to adapt to different voice - coil motor servo systems.

[0099] Step S2.1: Establish the S - type friction model

[0100] Most existing friction models are discontinuous around the point where the velocity v = 0. Although such models are closer to the physical principle of friction, there are some problems in actual use: 1), the friction with a velocity close to zero is difficult to be measured and identified, and even if it is identified, it is not accurate enough; 2), the frictional force is a non-linear function of velocity. When performing friction compensation, only the estimated value of velocity can be used. The estimation of velocity is often not very accurate, and due to the large noise in the system, the estimation of velocity often adds a relatively large noise, which aggravates the inaccuracy of compensation. Furthermore, the noise causes the estimation of velocity to frequently switch on both sides of zero velocity near zero velocity, so the friction compensation also frequently switches near zero velocity, resulting in serious chattering and noise phenomena.

[0101] Based on these problems, this application adopts an S-shaped friction model to replace the sign function type friction model, which can reduce the chattering and noise phenomena attached to friction compensation while effectively improving the control performance. The voice coil motor servo system equation based on the hyperbolic tangent function form friction model is:

[0102]

[0103] where f f and a are parameters to be identified.

[0104] Step S2.2: Identify S-shaped friction based on the genetic algorithm

[0105] There is no effective method for measuring and identifying the S-shaped friction model in the experiment. The effective S-shaped friction model depends on continuous adjustment and testing, which requires a lot of time. Moreover, for different voice coil motor servo systems, the parameters of the S-shaped friction model for achieving optimal friction compensation are also completely different. The genetic algorithm has the ability of global optimization, can automatically obtain and guide the search space, and adaptively adjust the search direction. Therefore, the applicant has developed a method for identifying S-shaped friction based on the genetic algorithm, as Figure 2 shown. Specifically: encode all individuals to form an initial population, calculate the fitness of each individual, and judge whether it meets the convergence condition of the genetic algorithm according to the fitness. If it meets, output the identification result; if not, perform selection operator, combinatorial crossover operator and mutation operator to generate a new solution set population, and gradually evolve to generate better and better approximate solutions. The optimal individual in the last generation population is decoded as an approximate optimal solution.

[0106] Step S2.2.1: Fitness J kEvaluation: Before designing the three basic operators of the genetic algorithm, an optimization metric needs to be determined first - in the genetic algorithm, it is called fitness. The fitness needs to satisfy the conditions of being single-valued, continuous, and non-negative. In this application, the fitness J is based on the standard deviation between the output of the S-shaped friction model and the actual output. k Evaluation, that is:

[0107]

[0108] y(ii) is the sampling of y at the ii-th moment, where:

[0109]

[0110] Step S2.2.2: Selection operator

[0111] The selection operator is to select superior individuals from the population. The purpose of selection is to directly inherit the optimized individuals to the next generation or generate new individuals through pairing and crossover and then inherit them to the next generation. Here, the Roulette Wheel Selection method is adopted. First, a selection probability is designed:

[0112]

[0113] Obviously, the selection probability here is defined as the ratio of the fitness of a certain individual to the total fitness of the population. The higher the fitness of an individual, the higher the probability of being selected, and the sum of the selection probabilities of all individuals is equal to 1.

[0114] However, in the simulation case, the fitness of the population is within the range of [400 - 600], so the calculated p k is all around 1 / n, and there is no discrimination between the well-performing parameters and the poorly-performing parameters in the selection process, resulting in the non-convergence of the genetic algorithm.

[0115] Therefore, in this application, the selection probability is defined as:

[0116]

[0117] where γ represents an adjustable parameter; J k represents the cost of the k-th individual, and n represents the number of individuals. By subtracting min(J i ), the performance of the model under different parameters can be more clearly reflected. At the same time, the adjustable parameter γ is added to avoid the differences being too obvious and the population converging to the local optimal solution quickly.

[0118] Then, the roulette wheel selection method is used to select superior individuals from the population, and the optimized individuals are directly inherited to the next generation or new individuals are generated through pairing and crossover and then inherited to the next generation to obtain the new population after selection. The roulette wheel selection method is an existing technology and will not be elaborated here.

[0119] Step S2.2.3: Crossover operator

[0120] Using the linear recombination algorithm in real-value recombination, two randomly selected individuals in the selected population exchange genes to generate new gene combinations, and it is expected to combine beneficial genes together; first, randomly sort the selected new population; then for two adjacent individuals in the new population, determine whether to crossover according to the crossover probability P c For the individuals to be crossed, perform crossover according to the following formula;

[0121] NO i = dO i +(1 - d)O i+1 (0.46)

[0122] NO i+1 = dO i+1 +(1 - d)O i (0.47)

[0123] where O i is the corresponding parameter of the individual before crossover, and NO i is the corresponding parameter of the new individual generated by crossover, and a is a random number between [0, 1].

[0124] Step S2.2.4: Mutation operator

[0125] The mutation operator introduces additional randomness to the genetic algorithm to prevent the algorithm from falling into a local optimal solution; specifically, for each individual in the population, determine whether mutation occurs according to the mutation probability P m If mutation occurs, the parameter of the individual becomes a random value within the parameter space.

[0126] Verify the effectiveness of the proposed method through simulation experiments:

[0127] As can be seen from formulas (1.11) and (1.12), the model to be identified is:

[0128]

[0129] The parameters to be identified by the genetic algorithm are Combined with prior information (the parameters k v = 223, k0 = 193500, b0 = 1220500 identified by using system identification and friction identification algorithms, and the identified f f≈80), the identification ranges of the parameters are as follows:

[0130]

[0131] The individuals in the initial population are randomly valued in the parameter identification space.

[0132] Here, the S-shaped friction model to be identified is written as:

[0133]

[0134] Rather than:

[0135]

[0136] This is because, let:

[0137]

[0138] Then we get:

[0139]

[0140] That is, the length of the transition section of the S-shaped friction model is inversely proportional to This will cause to easily converge to the local minimum. Due to the properties of the inverse proportional function, when is uniformly distributed, the length of the transition section is not uniformly distributed. Therefore, it is improved to equation (1.22). During the calculation process of the genetic algorithm, the differential equation is solved using the ode45 solver. According to the above analysis, the sampling time is selected as 4e -4 , the population size is 100, the number of iteration rounds is 100, the probability parameter γ is set to 20, the crossover probability P c is set to 0.9, the mutation probability P m is set to 0.1. The fitness convergence status of the four experiments is as shown in Figure 3 . Among them, Figure 3 in (a) represents the fitness convergence of the first simulation experiment, Figure 3 in (b) represents the fitness convergence of the second simulation experiment, Figure 3 in (c) represents the fitness convergence of the third simulation experiment, Figure 3 in (d) represents the fitness convergence of the fourth simulation experiment.

[0141] Table 1 Comparison of the parameters identified by the genetic algorithm, the nominal parameters, and the system identification parameters

[0142]

[0143]

[0144] The identified parameters are shown in Table 1. A comparison between the identified friction model and the friction model identified by the friction identification method is as Figure 4 shown.

[0145] Furthermore, the previous analysis shows that the sampling period of the measured data does not match the sampling period (2e -4 ) claimed by the host computer. Through testing, it is shown that it is closer to [3e -4 , 4e -4 . Regarding this problem, the genetic algorithm can be used to optimize various parameters. Regardless of whether they are analytic or not, the sampling time Ts is also regarded as a parameter to be identified. Then the parameter identification range of Ts is T s ~[2e -4 , 5e -4 .

[0146] Similarly, four identifications are carried out. The fitness convergence conditions of the four identifications are as Figure 5 shown in Table 2. Among them, Figure 5 in (a) represents the fitness convergence situation of the first identification, Figure 5 in (b) represents the fitness convergence situation of the second identification, Figure 5 in (c) represents the fitness convergence situation of the third identification, Figure 5 in (d) represents the fitness convergence situation of the fourth identification.

[0147] Table 2 Comparison of the parameters identified by the genetic algorithm, the nominal parameters, and the system identification (IV method)

[0148]

[0149] At this time, it is obvious that there is a large difference between the identified friction model and the optimal friction model obtained through debugging in the actual experiment. This is mainly because the trade-off between control performance and noise performance is considered in the actual debugging. However, by using the method of identifying the friction model with the genetic algorithm, it is easy to take the noise performance into account in the performance index, and only need to adjust the fitness index to:

[0150]

[0151] where, △F f represents the change value of friction.

[0152] Generally speaking, the noise after adding friction compensation is mainly caused by the chattering of the friction switching term or the approximate switching term. Therefore, the change amount of friction is added to the fitness index to measure the noise performance. β is an adjustable parameter. The operation results of the genetic algorithm under different β are as Figure 6 shown.

[0153] Table 3 Comparison of Identified Parameters by Genetic Algorithm, Nominal Parameters, and System Identification (IV Method)

[0154]

[0155] The identified parameters are shown in Table 3. The comparison of the identified friction model with the friction models identified by the friction identification methods is as Figure 7 shown.

[0156] Through the above simulation experiments, the effectiveness of the method proposed in the present invention is verified. The results show that the method can improve the control performance and robustness of the voice coil motor servo system. The solution described in the present invention provides an effective identification method based on genetic algorithm for the control and application of voice coil motor servo systems. This method has the potential to be applied to the identification and control problems of other nonlinear systems. Further research can apply this method to actual systems and conduct experimental verification and optimization improvements to further improve the performance and stability of the systems.

[0157] Taking the above ideal embodiments based on the present invention as an inspiration, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A friction nonlinear identification method for voice coil motors based on genetic algorithms, characterized in that, It includes the following steps: Step S1: Establish the dynamic model of a voice coil motor servo system with friction and analyze the influence of friction on the system performance; Step S2: Use the genetic algorithm to identify the parameters of the S-type friction model. By optimizing the parameters of the friction model, make it match the friction characteristics of the actual system to adapt to different voice coil motor servo systems; The specific steps of the said Step S1 include the following steps: Step S1.1: Construct the dynamic model of a voice coil motor servo system with friction Apply the working voltage U to the voice coil motor, generating a working current I to drive the voice coil motor to work, and a voltage drop U will be generated across the coil resistance R. R and an induced voltage drop U will be generated across the coil inductance L. L Meanwhile, a back electromotive force E will be generated when the voice coil motor is working; according to Kirchhoff's law, the voltage balance equation of the voice coil motor is obtained as follows: Among them, v is the speed at which the mover of the voice coil motor cuts the magnetic field lines, that is, the speed at which the mover moves in a straight line; k m is the electromagnetic force coefficient of the voice coil motor; Combined with the motion characteristics of the voice coil motor servo system, simplify and equivalent the mechanical motion model of the voice coil motor servo system to a single-degree-of-freedom spring-mass-damper mechanical positioning system; According to Newton's second law, the motion differential equation of the leaf spring suspension type voice coil motor driven optical image stabilization system is: where m is the equivalent mass of the voice coil motor mover, the camera module load, and their attached components; k is the equivalent stiffness of the reed and the system connection part; c is the equivalent damping coefficient of the voice coil motor mover and the camera module; y is the displacement of the system along the axial direction, and there is Combining \(E = NBlv=k\) m \(v\), \(k\) m \(= NBl\), the dynamic model of the voice coil motor servo system is obtained as follows: where N is the number of turns of the coil, B is the magnetic induction intensity, and l is the effective length of the coil in the magnetic field; k m is the electromagnetic force coefficient of the motor; It is assumed that the current closed-loop has been carried out in the servo system, and the dynamics of the current loop can be ignored relative to the mechanical dynamics of the system, that is: i≈i ref (0.4) i ref represents a given current Then the simplified dynamic equation of the voice coil motor servo system is: Step S1.2: Analyze the influence of friction on the system performance When the speed of the voice coil motor servo system switches frequently, the friction suffered by the system is a non-linear function F f (v) of the speed. Based on the Coulomb + viscous friction model, we have: F f (v) = Fsgn(v) + F v v(0.6) Among them, F is the Coulomb friction force, and F v is the viscous friction coefficient; At this time, the dynamic equation of the system is: Let x1 = y, x2 = v, u = I, and the state space equation of the voice coil motor servo system is obtained as: where \(k_0 = k / m\), \(k\) v = \(F\) v / m, \(b_0 = k\) m / m, \(f\) c = \(F / m\).

2. The friction nonlinear identification method of a voice coil motor based on a genetic algorithm according to claim 1, characterized in that, The specific steps of the said Step S2 include the following steps: Step S2.1: Establish the S-type friction model The voice coil motor servo system equation based on the hyperbolic tangent function form friction model is: where f f and a are parameters to be identified; Step S2.2: Identify the S-type friction based on the genetic algorithm. Specifically: Encode all individuals to form the initial population, calculate the fitness of each individual, and judge whether it meets the convergence condition of the genetic algorithm according to the fitness. If it meets, output the identification result; if not, perform the selection operator, combination crossover operator and mutation operator to generate a new solution set population, and gradually evolve to generate better and better approximate solutions. The optimal individual in the last generation population is decoded as the approximate optimal solution.

3. The friction non-linearity identification method of the voice coil motor based on the genetic algorithm according to claim 2, wherein In the step S2.2, fitness J is evaluated based on the standard deviation between the output of the S-shaped friction model and the actual output, that is: k Evaluation is as follows: y(ii) is the sampling of y at the ii moment, where:

4. The method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm according to claim 3, wherein In the step S2.2, the selection operator is specifically as follows: First, calculate the selection probability p of an individual k , where the selection probability is the ratio of the fitness of an individual to the total fitness of the population. The higher the fitness of an individual, the higher the probability of being selected, and the sum of the selection probabilities of all individuals is equal to 1; where γ represents an adjustable parameter; J k represents the cost of the k-th individual, and n represents the number of individuals; Then, use the roulette wheel selection method to select the superior individuals from the population, and directly inherit the optimized individuals to the next generation or generate new individuals through pairing crossover and then inherit them to the next generation.

5. The method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm according to claim 4, wherein In the step S2.2, the crossover operator is specifically as follows: The linear recombination algorithm in real-value recombination is adopted to exchange genes of two randomly selected individuals in the selected population, generate new gene combinations, and expect to combine beneficial genes together. First, randomly sort the selected new population. Then, for two adjacent individuals in the new population, determine whether to perform crossover according to the crossover probability P c ; For the individuals that perform crossover, perform crossover according to the following formula; NO i = dO i + (1 - d)O i+1 (0.14) NO i+1 = dO i+1 +(1 - d)O i (0.15) Among them, O i is the corresponding parameter of the individual before crossover, NO i is the corresponding parameter of the new individual generated by crossover, and d is a random number between [0, 1].

6. The method for identifying the friction nonlinearity of a voice coil motor based on a genetic algorithm according to claim 5, wherein In the step S2.2, the mutation operator introduces additional randomness into the genetic algorithm to prevent the algorithm from falling into a local optimal solution; specifically, for each individual in the population, according to the mutation probability P m it is judged whether mutation occurs, and the parameters of the individual with mutation become a random value within the parameter space.

Images