Servo parameter identification method and system

Abstract

The invention provides a servo system parameter identification method and system. The system comprises a measurement parameter mean filtering module, a regularization parameter range determining module, a regularization parameter solving module and an identification parameter solving module. According to the method, the optimal regularization parameter is dynamically selected in the identification process, and the optimal regularization parameter is dynamically selected to correct the least square method, so that the ill-conditioned problem occurring in the identification process of the least square method is solved, and the identification precision of the servo system parameters is improved.

Description

Technical Field

[0001] The present invention relates to a servo system parameter identification method and system, in particular to a high-precision servo parameter identification method for a permanent magnet synchronous motor, belonging to the technical field of servo motors. Background Art

[0002] Permanent magnet synchronous motors (PMSM for short) are widely used in the control systems of steering gears. Their driving torque is generated by the interaction between the excitation magnetic field of the rotor permanent magnet and the orthogonal armature rotating magnetic field. The servo system of a steering gear generally includes three control closed loops, namely, a current loop, a speed loop, and a position loop. Accurately obtaining the parameters of the three loops of the servo system is of great significance. Although the nominal values of some servo system parameters can be obtained by querying the manufacturer's manual, due to the influence of temperature and manufacturing errors during the working process, there will be certain deviations in the actual parameters. Therefore, it is necessary to identify the servo parameter values during operation.

[0003] Currently, the commonly used servo parameter identification methods are mainly online identification and offline identification. Commonly used online identification methods include autoregressive method, adaptive model reference method, etc. Po-Ngam proposed a new PI gain autoregressive design method and studied the influence of the autoregressive full-order observer of a sensorless permanent magnet synchronous motor on the robustness of stator resistance and inductance parameter changes. The effectiveness of the proposed method was verified through experiments. Ma Hao et al. proposed an adaptive identification method for power electronic circuit parameters based on the theory of hybrid systems, which can be applied to online fault trend judgment and predictive maintenance. However, limited by the computational efficiency of the online identification algorithm and the computational power of the controller, it is difficult to obtain some parameters within one servo cycle, thus reducing the control effect.

[0004] To solve this problem, currently, the parameter identification of the steering gear servo system mainly adopts the offline identification method. Offline identification regards the entire system as a time-invariant system, gives the system an excitation signal, collects and analyzes its response, and then identifies the relevant parameters through an identification algorithm. This method is related not only to the stability of the excitation signal (such as ramp signal, sine signal, step signal, and pseudo-random binary sequence, etc.), but also to the computational accuracy of the identification algorithm. Currently, traditional servo system identification algorithms include the least squares method, Kalman filter, and extended Kalman filter method, etc. However, for traditional identification algorithms, some problems will occur during the identification process, such as the singular ill-condition problem of the least squares method, the non-convergence problem in the recursive process of the Kalman filter or extended Kalman filter algorithm, etc. At the same time, for some intelligent algorithms, there are problems such as long search time and jamming phenomenon, making it difficult to apply in actual engineering.

[0005] For this reason, Hoerl and Kennard first proposed the ridge regression theory in 1970, which introduced a regularization parameter to solve the ill-posed problem in traditional identification algorithms, thereby improving the accuracy and robustness of servo parameter identification. However, in the identification process of the existing ridge regression theory, the selection of the regularization parameter is only based on fixed parameters or empirical values, and the identification accuracy of servo system parameters is not high. Summary of the Invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a servo system parameter identification method and system with dynamic selection of regularization parameters and high identification accuracy.

[0007] The technical solution of the present invention: A servo system parameter identification system includes a measurement parameter mean filtering module, a regularization parameter range determination module, a regularization parameter solving module, and an identification parameter solving module;

[0008] The measurement parameter mean filtering module filters the m-th measurement value of the servo parameter to be identified to obtain a measurement mean value;

[0009] The regularization parameter range determination module corrects the least squares method by introducing a regularization parameter, and based on the principle of mean square error invariance, obtains the value range of the regularization parameter in the m-th identification process;

[0010] The regularization parameter solving module randomly samples the regularization parameter within its value range, solves it according to formulas (3) and (4), and performs curve fitting according to the solution results to determine the regularization parameter corresponding to the point with the maximum curve curvature as the optimal regularization parameter,

[0011]

[0012] Wherein, is the eigenvalue of G m * i = 1, 2,..., n, representing the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is an intermediate variable in the solving process, Φ is the eigenvector of the eigenvalue of is the estimated value of the servo parameter to be identified in the m-th identification, G m * is the transfer function matrix of the servo parameter to be identified in the m-th time, is the measurement mean value filtered by the measurement parameter mean filtering module in the m-th time, β m is the regularization parameter in the m-th identification, I is the identity matrix;

[0013] The described identification parameter solving module uses the recursive least squares method to solve the servo parameters to be identified according to the optimal regularization parameter obtained by the regularization parameter solving module, and iteratively obtains the servo parameters to be identified that meet the accuracy requirements.

[0014] A servo system parameter identification method includes the following steps:

[0015] First step, determine the servo parameters to be identified in the servo system. According to the current loop model of the servo system, determine the transfer function matrix G of the servo parameters to be identified * ;

[0016] Second step, excite the current loop of the servo system to obtain the measured value Y of the servo parameters to be identified for the m-th time m * , filter the measured value Y m * to obtain the measured mean value of the servo parameters to be identified for the m-th time

[0017] Third step, correct the least squares formula through the regularization parameter β m and obtain the value range of the regularization parameter β m during the m-th identification process based on the invariant property of the mean square error;

[0018] Fourth step, determine the optimal regularization parameter β m ,

[0019] A4.1. Randomly sample the range of the regularization parameter β m determined in the third step, and use formulas (3) and (4) to solve for β m where,

[0020]

[0021] where, is the eigenvalue of G m * , i = 1, 2,..., n, representing the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is an intermediate variable in the solution process, Φ is the eigenvector of the eigenvalue (λ i * ), 2 is the estimated value of the servo parameters to be identified for the m-th identification, G m * is the transfer function matrix of the servo parameters to be identified for the m-th time, m is the measured mean value filtered by the measured parameter mean value filtering module for the m-th time, β m is the regularization parameter for the m-th identification, and I is the identity matrix;

[0022] A4.2. Curve fitting is performed on the solution value obtained in step A4.1, and the U-curve method is used to determine the point with the maximum curvature on the fitting curve, and the corresponding regularization parameter β m is the optimized regularization parameter β m ;

[0023] Fifth step, using the optimized regularization parameter β determined in the fourth step m , the recursive least squares method is used to correct the m+1th measurement mean value of the servo parameter to be identified;

[0024] Sixth step, determine whether the formula (11) is satisfied. If not, let m = m + 1 and then return to the second step, and repeat the second to sixth steps until the formula (11) is satisfied,

[0025]

[0026] wherein, is the control accuracy.

[0027] Advantages of the present invention compared with the prior art:

[0028] (1) By dynamically selecting the optimal regularization parameter during the identification process and using the dynamically selected optimal regularization parameter to correct the least squares method, the present invention solves the ill-conditioned problem that appears in the least squares method identification process and improves the identification accuracy of the servo system parameters;

[0029] (2) The present invention uses the U-curve method for the process of selecting the optimal regularization parameter, takes the regularization parameter as the parameter to be solved, and takes the regularization parameter corresponding to the point with the largest change in curvature as the optimal value, thus solving the problems of the traditional method depending on the curve fitting accuracy, large calculation amount, and easy curve divergence, and further improving the identification accuracy of the servo system parameters;

[0030] (3) The present invention introduces the optimal regularization parameter into the recursive least squares method, and by dynamically selecting the optimal regularization parameter, continuously corrects the mth calculation result to make it gradually approach the accurate value, thus effectively improving the accuracy of the servo system parameters to be identified;

[0031] (4) Before setting the control parameters of the servo system, the present invention obtains the parameter model to be identified based on the three-loop system of the permanent magnet synchronous motor, and uses a high-precision identification method to obtain parameters such as the armature inductance, armature resistance, rotor and reducer moment of inertia, system equivalent stiffness and damping of the actual system, thereby improving the accuracy of the servo system control parameters. Description of the Drawings

[0032] Figure 1 is the flow chart of the present invention;

[0033] Figure 2 These are the identification results of OLS, RLS, and ARLS in the embodiments of the present invention. Detailed implementation manners

[0034] The present invention provides a high-precision servo parameter identification system for a permanent magnet synchronous motor, including a measurement parameter mean filtering module, a regularization parameter β range determination module, a regularization parameter β solution module, and an identification parameter solution module.

[0035] The measurement parameter mean filtering module performs filtering processing on the m-th measurement value Y of the servo parameter to be identified m * to obtain the measurement mean value.

[0036] In the present invention, the measurement parameter mean filtering module performs mean filtering on the measurement values in the m-th identification process to reduce the interference of measurement errors and make the measurement values tend to be stable. Mean filtering is a well-known technology in the art.

[0037] Furthermore, the servo parameters to be identified include the sum of the moments of inertia I Mg (the moment of inertia I of the motor rotor M and the moment of inertia I of the speed reducer gear sum), the armature inductance L of the motor a and the armature resistance R a .

[0038] The regularization parameter β range determination module corrects the least squares method by introducing the regularization parameter β, and based on the principle of mean square error invariance, obtains the value range of the regularization parameter β m in the m-th identification process.

[0039] Furthermore, the regularization parameter β range determination module corrects the least squares method (Ordinary least square, OLS) using the ridge regression theory (Ordinary ridgeregression, ORR), and the correction equation is shown in formula (1):

[0040]

[0041] Wherein, is the estimated value of the servo parameter to be identified in the m-th identification, G m * is the transfer function matrix of the servo parameter to be identified in the m-th time, is the measurement mean value filtered by the measurement parameter mean filtering module in the m-th time, β m is the regularization parameter in the m-th identification, and I is the identity matrix.

[0042] Based on the principle of mean square error invariance, solve formula (1) to obtain the regularization parameter β during the m - time identification process m The value range is as shown in formula (2):

[0043]

[0044] where is the eigenvalue of G m * i = 1, 2, …, n, representing the number of servo parameters to be identified, σ is the standard deviation of the measurement noise is an intermediate variable in the solution process, Φ is the eigenvector corresponding to the eigenvalue of G

[0045] The transfer function matrix G of the servo parameters to be identified * is a mathematical model of the servo parameters to be identified determined according to the current loop model of the servo system. The matrix with the servo parameters to be identified as elements is a well - known technology in the art

[0046] The regularization parameter β solving module determines the optimal regularization parameter β m The regularization parameter β solving module performs random sampling on the regularization parameter β determined by the regularization parameter β range determination module, and solves for β m The solution formula is as follows: m where

[0047]

[0048] where is for calculating the mean square error formula

[0049] The regularization parameter β solving module selects the point with the maximum curvature as the optimal regularization parameter β according to the solution results of formulas (3) and (4) using the U - curve method m .

[0050] The curvature of the U - curve method can be expressed as formula (5): (The U - curve method is used to calculate the value of β at the point with the maximum curvature of the curve) m

[0051]

[0052] where

[0053]

[0054] The identification parameter solving module, according to the optimal regularization parameter β obtained by the regularization parameter β solving module m ​, the recursive least squares (RLS) method is used to solve the parameters to be identified. By continuously iterating until the accuracy requirement is met, high-precision parameters to be identified can be effectively obtained.

[0055] Furthermore, based on the solved optimal β m value, the recursive least squares method is used to correct the m-th calculation result, making the parameters to be identified gradually approach the exact value. The solution formula of the corrected recursive algorithm is as follows:

[0056]

[0057] Among them,

[0058]

[0059] Among them, is the estimated value of the parameter to be identified at the m-th time; is the prediction of the (m + 1)-th measurement value based on the m-th measurement value, is the mean value of the m-th measurement, is the prediction error, also known as the innovation, and I is the identity matrix.

[0060] As the recursion progresses, the influence of the initial value is gradually reduced. When the parameter estimation of the recursive least squares algorithm meets the accuracy requirement, the following formula can be selected to end the loop.

[0061]

[0062] Among them, is the control accuracy, and the general engineering value range is less than or equal to 10 -6 .

[0063] The identification method of the adaptive recursive least squares (ARLS) algorithm based on the ridge regression theory in the present invention, that is, each time the recursive least squares algorithm is performed, by dynamically selecting the optimal value of the regularization parameter β m , the m-th calculation result is corrected to gradually approach the exact value.

[0064] The present invention also provides a method for identifying servo parameters of a permanent magnet synchronous motor. As Figure 1 shown, before the servo system control parameters are tuned, a parameter model to be identified is obtained based on the three-loop system of the permanent magnet synchronous motor, and high-precision identification methods are used to obtain parameters such as the armature inductance, armature resistance, rotor and reducer moment of inertia, system equivalent stiffness and damping of the motor in the actual system, thereby improving the accuracy of the servo system control parameters.

[0065] Specifically, it includes the following steps:

[0066] In the first step, determine the servo parameters to be identified of the servo system. According to the current loop model of the servo system, determine the transfer function matrix G of the servo parameters to be identified * .

[0067] Furthermore, the servo parameters to be identified of the servo system in this step include the sum of moments of inertia I Mg (the sum of the moment of inertia of the motor rotor I M and the moment of inertia of the reducer I gear ), the armature inductance L of the motor a , and the armature resistance R a .

[0068] In the second step, excite the current loop of the servo system to obtain the measured value Y of the servo parameters to be identified for the m-th time m * , and perform filtering processing on the measured value Y m * to obtain the measured mean value of the servo parameters to be identified for the m-th time

[0069] In this step, the filtering processing is a well-known technology in the art.

[0070] In the third step, correct the least squares formula through the regularization parameter β m and obtain the value range of the regularization parameter β m in the m-th identification process based on the invariant characteristic of the mean square error.

[0071] Furthermore, this step includes

[0072] A3.1. Use the regularization parameter β m to correct the least squares formula of formula (1),

[0073]

[0074] wherein is the estimated value of the servo parameters to be identified for the m-th identification, G m * is the transfer function matrix of the servo parameters to be identified for the m-th time, is the measured mean value filtered by the measured parameter mean value filtering module for the m-th time, β m is the regularization parameter for the m-th identification, and I is the identity matrix;

[0075] A3.2. Based on the principle of invariant mean square error, solve formula (1) to obtain the value range of the regularization parameter β m in the m-th identification process of formula (2),

[0076]

[0077] Among them, is the eigenvalue of G m * , i = 1, 2, …, n, representing the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is an intermediate variable in the solution process, and Φ is the eigenvector corresponding to the eigenvalue of.

[0078] Step 4: Determine the optimal regularization parameter β m ,

[0079] A4.1. Randomly sample within the range of the regularization parameter β determined in Step 3, and use formulas (3) and (4) to solve for β m , m .

[0080]

[0081] Among them, is the eigenvalue of G m * , i = 1, 2, …, n, representing the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is an intermediate variable in the solution process, Φ is the eigenvector corresponding to the eigenvalue of, is the estimated value of the servo parameter to be identified in the m-th identification, and G m * is the transfer function matrix of the servo parameter to be identified in the m-th time, is the measurement mean filtered by the measurement parameter mean filtering module in the m-th time, β m is the regularization parameter in the m-th identification, and I is the identity matrix.

[0082] A4.2. Perform curve fitting on the solution values in Step A4.1, and use the U-curve method to determine the point with the maximum curvature on the fitting curve. The regularization parameter β m corresponding to this point is the optimal regularization parameter β m .

[0083] Furthermore, in this step, the U-curve method determines the value of the optimal regularization parameter β m corresponding to the point with the maximum curvature of the curve according to formula (5),

[0084]

[0085] Among them,

[0086]

[0087] In the fifth step, the optimized regularization parameter β determined in the fourth step is used m , and the recursive least squares method is adopted to correct the mean value of the (m + 1)-th measurement of the servo parameters to be identified.

[0088] Furthermore, in this step, the correction is performed using the formula group (8),

[0089]

[0090] wherein,

[0091]

[0092] wherein, is the estimated value of the parameter to be identified at the m-th time; is the prediction of the (m + 1)-th measurement value based on the m-th measurement value, is the mean value of the m-th measurement, is the prediction error, also known as the innovation, and I is the identity matrix.

[0093] In the sixth step, it is judged whether the formula (11) is satisfied. If not, let m = m + 1 and then return to the second step, repeating the second to the sixth steps until the formula (11) is satisfied,

[0094]

[0095] wherein, is the control accuracy.

[0096] Furthermore, the control accuracy in this step generally has a value range of less than or equal to 10 -6 , and those skilled in the art can select according to the actual situation.

[0097] The present invention will be described in detail below with reference to specific examples and drawings.

[0098] A high-precision servo system parameter identification system for a permanent magnet synchronous motor provided by this example includes a measurement parameter mean value filtering module, a regularization parameter β range determination module, a regularization parameter β solution module, and an identification parameter solution module.

[0099] The parameter mean value filtering module reduces the interference of measurement errors by adopting m identification processes, making the measurement parameters tend to be stable.

[0100] The regularization parameter β range determination module corrects the least squares method by introducing the regularization parameter β to obtain the value range of the regularization parameter β in the m identification processes m .

[0101] The regularization parameter β solution module processes β in the m identification processesm Perform sampling and use the U-curve method to obtain the optimal β m value.

[0102] Based on the obtained optimal β m value, use the recursive least squares method to solve the parameters to be identified. Continuously iterate until the accuracy requirement is met, so as to effectively obtain the high-precision parameters to be identified.

[0103] Furthermore, this embodiment also provides a method for identifying the parameters of a high-precision servo system of a permanent magnet synchronous motor, including the following steps:

[0104] The first step is to determine the servo system parameters to be identified, including the sum of the moments of inertia I Mg (the sum of the moment of inertia of the motor rotor I M and the moment of inertia of the reducer I gear ), the armature inductance L a of the motor, and the armature resistance R a .

[0105] In practical applications, factors such as temperature have a greater impact on the servo system. Therefore, before adjusting the control parameters, it is necessary to identify some parameters of the servo system. The main parameters that usually affect the identification of the servo system are the armature inductance and armature resistance of the servo motor, the equivalent moment of inertia of the controlled system, etc. The nominal parameters of the servo system of a conventional motor are shown in Table 1.

[0106] Table 1 Nominal parameters of the servo system

[0107]

[0108] The second step is to determine the servo system identification model.

[0109] According to the current loop model of the servo system, determine the sum of the moments of inertia I Mg (the sum of the moment of inertia of the motor rotor I M and the moment of inertia of the reducer I gear ), the armature inductance L a of the motor, and the armature resistance R a parameter model to be identified. The parameter model to be identified is a transfer function matrix, including the sum of the moments of inertia I Mg , the armature inductance L a of the motor, and the armature resistance R a mathematical model.

[0110] The third step is to measure the parameter mean filtering.

[0111] In this example, an M-sequence excitation signal with an amplitude of 0.1 A and a frequency of 500 Hz is used to excite the current loop, so as to obtain the measured values of the parameters to be identified. At the same time, the mean filtering process is performed on the measured values of the identified parameters during the identification process, so that the accuracy of the measured values tends to be stable.

[0112] Step 4, regularization parameter β m Range determination.

[0113] By introducing the regularization parameter β m The least squares formula is corrected, and the range of the regularization parameter β in the m identification processes is obtained based on the invariant characteristic of the mean square error. m Value range.

[0114] Step 5, optimal regularization parameter β m Determination.

[0115] Sample β in the m identification processes m and use the U-curve method to obtain the β value at the point with the maximum curvature, that is, the optimal β value required for this point. m Value. m Value.

[0116] Step 6, solve the parameters to be identified.

[0117] Based on the obtained optimal β m value, solve the parameters to be identified, and select the optimal β value in each iteration process until the parameters to be identified to be solved meet the accuracy requirements, as shown in Table 2. m Value until the parameters to be identified to be solved meet the accuracy requirements, as shown in Table 2.

[0118] Compare the identification convergence situations of the method of the present invention (ARLS) with the least squares method (OLS) and the recursive least squares method (RLS), as Figure 2 shown. From Figure 2 it can be seen that the present invention can quickly achieve convergence and is closer to the true value, so it can effectively improve the accuracy of the parameters to be identified.

[0119] Table 2 Final identification results of ARLS

[0120]

[0121] The parts not detailed in the present invention are well-known technologies to those skilled in the art.

Claims

1. A servo system parameter identification system, characterized in that: It includes a measurement parameter mean filtering module, a regularization parameter range determination module, a regularization parameter solution module and an identification parameter solution module; The measurement parameter mean value filtering module performs filtering processing on the mth measurement value of the servo parameter to be identified to obtain a measurement mean value; The regularization parameter range determination module modifies the least square method by introducing a regularization parameter, and obtains the value range of the regularization parameter in the m-times identification process based on the mean square error invariance principle; The regularization parameter solving module randomly samples the regularization parameter within its value range, solves the formulas (3) and (4) for the regularization parameter, and performs curve fitting based on the solution results to determine the regularization parameter corresponding to the maximum point of the curve curvature as the optimized regularization parameter. in, G m * The characteristic root of , i = 1, 2, ..., n, represents the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is the intermediate variable of the solution process, Φ is the characteristic root The characteristic vector of is the estimated value of the servo parameter to be identified for the mth time, G m * is the transfer function matrix of the mth servo parameter to be identified, is the mth measurement mean after filtering by the measurement parameter mean filtering module, β m is the regularization parameter of the mth identification, I is the unit matrix; The identification parameter solving module uses the recursive least square method to solve the servo parameters to be identified according to the optimized regularization parameters obtained by the regularization parameter solving module, and obtains the servo parameters to be identified that meet the accuracy requirements through iteration.

2. A servo system parameter identification system according to claim 1, characterized in that: The regularization parameter range determination module adopts ridge regression theory to correct the least squares method. The correction equation is shown in formula (1): in, is the estimated value of the servo parameter to be identified for the mth time, G m * is the transfer function matrix of the mth servo parameter to be identified, is the mth measurement mean after filtering by the measurement parameter mean filtering module, β m is the regularization parameter of the mth identification, and I is the identity matrix.

3. A servo system parameter identification system according to claim 2, characterized in that: The regularization parameter range determination module solves formula (1) based on the mean square error invariance principle to obtain the regularization parameter β in the m-th identification process. m The value range is shown in formula (2): in, G m * The characteristic root of , i = 1, 2, ..., n, represents the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is the intermediate variable of the solution process, Φ is the characteristic root The feature vector of .

4. A servo system parameter identification system according to claim 3, characterized in that: The regularization parameter solving module adopts the U-curve method, and the curve curvature is calculated by formula (5): in, 5. A servo system parameter identification system according to claim 3, characterized in that: The servo parameters to be identified include the sum of the moments of inertia I Mg , Motor armature inductance L a and armature resistance R a .

6. A servo system parameter identification system according to claim 3, characterized in that: The identification parameter solving module uses the formula group (8) based on the solved optimal regularization parameter to correct the m-th identification result by using the recursive least square method. in, P m =((G m * ) T G m * ) -1 (9) P m+1 =((G m+1 * ) T G m+1 * ) -1 (10) in, is the estimated value of the parameter to be identified for the mth time; It is the prediction of the m+1th measurement value based on the mth measurement value. is the mean value of the mth measurement, is the prediction error, also known as the new information, and I is the unit matrix.

7. A servo system parameter identification system according to claim 6, characterized in that: The identification parameter solving module uses formula (11) to determine whether the iteration is completed. in, To control the accuracy, the engineering value range is generally less than or equal to 10 -6 .

8. A servo system parameter identification method, characterized in that: The following steps are involved: The first step is to determine the servo parameters to be identified of the servo system. According to the servo system current loop model, the transfer function matrix G of the servo parameters to be identified is determined. * ; The second step is to excite the servo system current loop and obtain the mth measurement value Y of the servo parameter to be identified. m * , for the measured value Y m * Perform filtering to obtain the measured mean value of the mth servo parameter to be identified The third step is to use the regularization parameter β m The least squares formula is modified, and the regularization parameter β in the m-th identification process is obtained based on the mean square error invariance characteristic. m The value range of The fourth step is to determine the optimal regularization parameter β m , A4.

1. Regularization parameter β determined in step 3 m Random sampling is performed in the range, and formulas (3) and (4) are used to calculate β m To solve, in, G m * The characteristic root of , i = 1, 2, ..., n, represents the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is the intermediate variable of the solution process, Φ is the characteristic root The characteristic vector of is the estimated value of the servo parameter to be identified for the mth time, G m * is the transfer function matrix of the mth servo parameter to be identified, is the mth measurement mean after filtering by the measurement parameter mean filtering module, β m is the regularization parameter of the mth identification, I is the unit matrix; A4.

2. Perform curve fitting on the solution value of step A4.1 and use the U-curve method to determine the point with the maximum curvature on the fitting curve. The regularization parameter β corresponding to this point is m To optimize the regularization parameter β m ; Step 5: Use the optimized regularization parameter β determined in step 4 m , the recursive least square method is used to correct the m+1th measurement mean of the servo parameter to be identified; Step 6: Determine whether formula (11) is satisfied. If not, set m=m+1 and return to step 2. Repeat steps 2 to 6 until formula (11) is satisfied. in, To control accuracy.

9. A servo system parameter identification method according to claim 8, characterized in that: The third step includes the following steps, A3.

1. Using the regularization parameter β m The least squares formula of formula (1) is modified. in, is the estimated value of the servo parameter to be identified for the mth time, G m * is the transfer function matrix of the mth servo parameter to be identified, is the mth measurement mean after filtering by the measurement parameter mean filtering module, β m is the regularization parameter of the mth identification, I is the unit matrix; A3.

2. Based on the mean square error invariance principle, solve formula (1) and obtain the regularization parameter β in the m-th identification process of formula (2): m Value range, in, G m * The characteristic root of , i = 1, 2, ..., n, represents the number of servo parameters to be identified, σ is the standard deviation of the measurement noise, is the intermediate variable of the solution process, Φ is the characteristic root The feature vector of .

10. A servo system parameter identification method according to claim 9, characterized in that: The step A4.2 U-curve method determines the optimal regularization parameter β corresponding to the maximum point of the curve curvature according to formula (5) m The value of in, In the fifth step, the formula group (8) is used for correction: in, P m =((G m * ) T G m * ) -1 (9) P m+1 =((G m+1 * ) T G m+1 * ) -1 (10) in, is the estimated value of the parameter to be identified for the mth time; It is the prediction of the m+1th measurement value based on the mth measurement value. is the mean value of the mth measurement, is the prediction error, also known as the innovation, and I is the unit matrix; or In the first step, the servo parameters to be identified of the servo system include the sum of the moments of inertia I Mg , Motor armature inductance L a and armature resistance R a .

Images