Piezoelectric deformable reflector surface shape closed-loop driving and rapid and accurate control method
Through the combination of least squares optimization and the P-I operator model, the hysteresis problem of piezoelectric deformable mirrors is solved, the control accuracy and modulation performance are improved, and the surface shape control with high frequency and high precision is realized.
Patent Information
- Application Number
- CN202411102434.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-22
AI Technical Summary
There are hysteresis problems in piezoelectric deformable mirrors in practical applications, resulting in a decrease in control accuracy and limited modulation performance. It is difficult for the existing technology to effectively solve the impact of hysteresis effect on control speed.
The least squares optimization method is used to characterize the actuator's influence function through the ultra-Gaussian function, and the hysteresis characteristics are described using the P-I operator model, a hysteresis inverse model is established for feedforward compensation, and the control voltage signal is optimized to reduce hysteresis errors.
The mirror surface shape control accuracy and modulation performance of the piezoelectric deformable reflector is improved, the error introduced by hysteresis is suppressed, and the surface shape control with high frequency and high precision is realized.
Smart Images

Figure CN119002039A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of photoelectric detection, and in particular relates to a closed-loop driving and rapid and precise control method for the surface shape of a piezoelectric deformable reflector. Background Art
[0002] As a typical piezoelectric drive device, the working principle of the piezoelectric deformable mirror is mainly based on the piezoelectric effect, that is, by applying an electric field to the piezoelectric material to achieve the change of the mirror surface shape, thereby achieving the adjustment and control of the optical system. The piezoelectric deformable mirror can achieve high-degree-of-freedom and high-precision surface shape changes, providing an important technical means for optical imaging, communication, laser adjustment and other optical fields. However, piezoelectric deformable mirrors often encounter hysteresis problems in practical applications.
[0003] The hysteresis problem of the piezoelectric deformable mirror refers to the hysteresis between the forward voltage-surface shape characteristic curve and the reverse voltage-surface shape characteristic curve. The hysteresis phenomenon is mainly caused by the combined effect of the electrostrictive effect, the ferroelectric effect and the inverse piezoelectric effect. Specifically, the output surface shape of the piezoelectric deformable mirror at the next moment is not only related to the current control voltage and the output surface shape, but also to the historical value of the control voltage and the driving voltage frequency. In addition, the hysteresis loop of the piezoelectric deformable mirror is asymmetric, and there is no symmetry between the forward voltage-surface shape characteristic curve and the reverse voltage-surface shape characteristic curve. Driven by a periodic voltage, the piezoelectric deformable mirror usually cannot return to its initial value.
[0004] Hysteresis nonlinearity has a significant impact on the performance of piezoelectric deformable mirrors. Due to the nonlinear relationship between the input and output of the deformable mirror, there is a certain deviation between the voltage calculated based on the linear superposition assumption and the actual required control voltage, which will lead to a decrease in the open-loop control accuracy of the piezoelectric deformable mirror. Although the influence of the hysteresis effect of the piezoelectric deformable mirror can be eliminated through real-time closed-loop correction, it will also reduce the control speed of the piezoelectric deformable mirror.
[0005] In order to solve the above problems, a hysteresis model is usually established for the piezoelectric deformable mirror to describe its hysteresis characteristics, and then the hysteresis inverse model is used to compensate for the hysteresis of the piezoelectric deformable mirror, thereby optimizing the control method of the piezoelectric deformable mirror, thereby reducing the influence of the hysteresis effect on the piezoelectric deformable mirror and improving the surface control accuracy and modulation performance of the piezoelectric deformable mirror. Summary of the invention
[0006] In order to overcome the shortcomings of the prior art, the present invention provides a closed-loop driving and fast and precise control method for the surface shape of a piezoelectric deformable reflector. Based on the least squares optimization method for solving the influence function matrix of the piezoelectric deformable reflector, a super Gaussian function is used to characterize the influence function of the piezoelectric deformable reflector brake. A (PI) operator model is selected to describe the hysteresis characteristics of piezoelectric ceramics. The operator model has clear expressions and is easy to solve, and can directly solve the error influence, thereby suppressing the error introduced by hysteresis in the control process of the deformable mirror with piezoelectric control braking, and improving the surface shape control accuracy and modulation performance of the piezoelectric deformable reflector.
[0007] The technical solution adopted by the present invention to solve the technical problem is as follows:
[0008] Step 1: Structural parameter identification of piezoelectric deformable mirror based on actuator influence function simulation;
[0009] Step 1-1: Use the improved Gaussian function to characterize the influence function of the actuator, as shown in the following formula:
[0010]
[0011] Where A represents the amplitude of the Gaussian function, B represents the amplitude of the radial basis function, ω represents the coupling coefficient between the actuators, r and d represent the distance from the point on the mirror to the center of the actuator and the distribution spacing of the actuators, α represents the Gaussian index, and γ represents the radial correction width;
[0012] Step 1-2: The stiffness of the connecting column is expressed as:
[0013]
[0014] Where k represents stiffness, E is Young's modulus, r0 and l represent the radius and length of the connecting column respectively;
[0015] The reconstruction capabilities of different structural parameter pairs were compared, and the root mean square error RMSe was used to evaluate the reconstruction accuracy as shown in the following formula:
[0016]
[0017] Where n is the number of samples, Z Mi is the measured value of the aberration surface vector height measured for the ith time, Z Ti is the true value of the aberration surface vector height measured for the i-th time;
[0018] Step 2: Design of high-frequency and high-precision control scheme for piezoelectric deformable mirror based on hysteresis inverse model and multi-degree-of-freedom "surface shape-voltage" transfer model;
[0019] Step 2-1: Use the standard Zernike polynomials to describe the surface shape and wavefront distortion of the deformable mirror, and compose the control voltage signal C in sequence by the channel voltage that drives the actuator to deform:
[0020] C=[c1,c2,...,c i ,...,c N ] (4)
[0021] Where N is the number of actuators in the deformable mirror, i is the actuator number, and C i represents the control voltage applied to the i-th actuator;
[0022] Step 2-2: The deformed mirror shape z(x,y) under the control voltage signal is fitted by the standard Zernike polynomial:
[0023]
[0024] Where Z k (x, y) is the standard Zernike polynomial, and the first M terms are involved in the fitting, S k is the fitting coefficient of the kth standard Zernike polynomial;
[0025] Step 2-3: Define the vector composed of fitting coefficients as the surface characterization vector S:
[0026] S=[s1,s2,...,s i ,...,s N ] (6)
[0027] The influence function of each actuator is fitted using the standard Zernike polynomial and arranged in sequence to obtain the influence function matrix G that reflects the mapping relationship between the deformable mirror shape and the control voltage signal:
[0028]
[0029] The relationship between the surface characterization vector S and the control voltage signal C is expressed as:
[0030] S=GC (8)
[0031] Considering the coupling between actuators, the influence function matrix is obtained using the least squares optimization method:
[0032] G=SC T (CC T ) (9)
[0033] Based on the pseudo-inverse method, the influence function matrix G is decomposed into singular values:
[0034] G=U M×M σ M×N WN×N V T (10)
[0035] Among them, U M×M is a unitary matrix of order M, σ M×N is a semi-positive definite diagonal matrix, W N×N is an N-order unitary matrix, V T is the conjugate transpose;
[0036] The control voltage signal is solved as:
[0037] C=Vσ -1 U T S (11)
[0038] Step 2-4: In the reconstruction of the deformable mirror, the deviation of the surface representation vector between the target surface and the current mirror surface is used as the evaluation function, as shown in the following formula:
[0039] f(x)=e n =S d -S n =S d -GC n (12)
[0040] Among them, S d is the target surface characterization vector, C n is the control voltage signal currently acting on the deformable mirror, S n is the characterization vector of the current deformable mirror surface shape, and the control voltage signal of the deformable mirror in the next iteration is:
[0041] C n+1 =C n +αG T e n (13)
[0042] Step 2-5: Using a non-influence function matrix control algorithm based on a stochastic parallel gradient descent algorithm to control the surface shape of the piezoelectric deformable reflector;
[0043] Randomly add a bidirectional disturbance voltage signal δ to the current control voltage signal C =[δ C1 ,δ C2 ,...,δ CN ], and calculate the change in the evaluation function value after the bidirectional disturbance as the gradient estimate. The change direction of the control voltage signal is the direction of gradient descent. The iterative formula is expressed as:
[0044] C n+1 =C n +γ(E + -E - )δ C (14)
[0045] Where γ is the gain coefficient, and the RMS value of the error between the target surface shape and the current surface shape of the deformable mirror is defined as the evaluation function E.
[0046] Step 2-6: The hysteresis of the piezoelectric actuator is feed-forward compensated and controlled. The PI model is used to describe the hysteresis characteristics of the piezoelectric ceramic, and the parameter identification of its inverse model is completed. Based on this, a "pseudo-linear" system based on the hysteresis inverse model feed-forward compensation is completed.
[0047] The PI model uses a bilateral Play operator and a weighted superposition of weight functions to describe hysteresis nonlinearity, where the Play operator expression is:
[0048] H r =y(t i )=max{v(t i )-r,min{v(t i )+r,y(t i-1 )}} i=1,2,3,...n (15)
[0049] In the formula, v and y represent the input and output of the Play operator respectively, r represents the threshold, n is the number of data sampling points, y0 represents the initial output value of the Play operator, and H r is the Play hysteresis operator with a threshold of r;
[0050] By weighted superposition of multiple Play operators with different thresholds, the PI hysteresis model describing the hysteresis voltage is obtained:
[0051] WH=y(t i ) j ·max{v(t i )-r j ,min{v(t i )+r j ,y j (t i-1 )}}j=1,...N; i=1,...n(16)
[0052] Where N is the number of Play operators, r j is the threshold of the jth Play operator, R = [r1, r2, ..., r N ] is the threshold vector, W=[w1,w2,...,w N ] is the weight vector, H is the matrix composed of Play operators with different thresholds;
[0053] When R is determined artificially, the model parameter W is identified by the least squares method, and the result is expressed as:
[0054] W=(HH T )-1 Hy (17)
[0055] The parameters that need to be identified in the PI inverse model are the threshold vector RR and the weight vector WW, which are obtained from the threshold and weight vector of PI:
[0056]
[0057]
[0058] Preferably, the coupling coefficient ω is between 5% and 12%, and the Gaussian exponent α is between 1.9 and 2.3.
[0059] Preferably, the different pairs of structural parameters include astigmatism, trefoil aberration, coma, spherical aberration, second-order astigmatism, second-order trefoil aberration, and second-order coma.
[0060] Preferably, the initial output value y0 of the Play operator is taken as 0.
[0061] The beneficial effects of the present invention are as follows:
[0062] The present invention introduces a feedforward compensation control model and selects a (PI) operator model with clear expressions and easy solutions to describe the hysteresis characteristics of piezoelectric ceramics, completes the parameter identification of its inverse model, and accordingly establishes a "pseudo-linear" system based on the hysteresis inverse model feedforward compensation to directly solve the error influence, thereby suppressing the error introduced by hysteresis in the control process of the deformable mirror with piezoelectric control braking. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] Figure 1 It is a schematic diagram of the fitting result of the actuator influence function according to an embodiment of the present invention;
[0064] Figure 2 Schematic diagram of simulation of structural parameter changes of a piezoelectric deformable mirror according to an embodiment of the present invention, (a) Gaussian index, (b) coupling coefficient, and (c) structural parameter solution space.
[0065] Figure 3 Schematic diagram of reconstruction error of a typical aberration surface shape of a deformable mirror with different structural parameters according to an embodiment of the present invention;
[0066] Figure 4 A schematic diagram of an influence function matrix data set according to an embodiment of the present invention;
[0067] Figure 5 Schematic diagram of control voltage of actuators of different channels according to an embodiment of the present invention;
[0068] Figure 6 It is a schematic diagram of the reconstruction result of the astigmatism surface shape by the SPGD algorithm according to an embodiment of the present invention;
[0069] Figure 7 Schematic diagram of input and output of the hysteresis simulation system according to an embodiment of the present invention, (a) hysteresis loop of the actuator, (b) input and output of the simulation system;
[0070] Figure 8 Schematic diagram of describing hysteresis by the PI model according to an embodiment of the present invention, (a) original output and PI simulated output, (b) original hysteresis loop and PI simulated hysteresis;
[0071] Fig. 9 It is a schematic diagram of the piezoelectric hysteresis compensation principle based on the PI hysteresis inverse model according to an embodiment of the present invention;
[0072] Fig.10 Schematic diagram of the hysteresis compensation effect of the piezoelectric actuator according to an embodiment of the present invention, (a) output displacement compensation, (b) the size of the hysteresis loop before and after compensation. DETAILED DESCRIPTION
[0073] The present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0074] The present invention aims to solve the problem of realizing structural optimization design and precise control of a piezoelectric deformable reflector controlled by a multi-channel actuator.
[0075] To solve the above technical problems, the present invention proposes a method for solving the influence function matrix of a piezoelectric deformable reflector based on least squares optimization, and uses a super-Gaussian function to characterize the influence function of a piezoelectric deformable reflector actuator. The present invention specifically includes the following contents.
[0076] (1) Structural parameter identification of piezoelectric deformable mirror based on actuator influence function simulation
[0077] Due to the large stiffness difference between the actuator and the mirror, when a single actuator acts, there will be a phenomenon that the mirror is deformed in the opposite direction of the actuator in a region outside the actuator. As a result, the traditional Gaussian function cannot fit this part of the mirror deformation well. To solve the above problem, the improved Gaussian function is used to characterize the influence function of the actuator, as shown in the following formula:
[0078]
[0079] Among them, A represents the magnitude of the Gaussian function amplitude, B represents the magnitude of the radial basis function amplitude, ω represents the coupling coefficient between actuators, r and d represent the distance from the point on the mirror to the center of the actuator and the distribution spacing of the actuator, respectively, α represents the Gaussian index, and γ represents the radial correction width. The expression of the actuator influence function directly affects the surface reconstruction accuracy of the deformable mirror. The reasonable range of the coupling coefficient ω between actuators should be between 5% and 12%. If the coupling coefficient is too large, it will affect the actuator and reduce the ability of the deformable mirror to reconstruct high-order surfaces. If the coupling coefficient is too small, it will cause the local fluctuation of the mirror surface to be too large and unable to form a continuous wavefront. The Gaussian index of the influence function also has an important influence on the control stability and surface reconstruction ability of the deformable mirror. The reasonable Gaussian index should be between 1.9 and 2.3.
[0080] After the deformable mirror leaves the factory, the mirror diameter, the number of actuators, and the arrangement cannot be changed. Therefore, the structural parameters that can be optimized are mainly the thickness of the mirror and the stiffness of the connecting column. The stiffness of the connecting column can be expressed as:
[0081]
[0082] Where k represents stiffness, E is Young's modulus, r0 and l represent the radius and length of the connecting column, respectively. Since the exponent of r0 is 2, the change of r0 has a more significant effect on stiffness than l. By traversing the thickness of the mirror and the radius of the connecting column, the changes in the exponent of the actuator influence function and the coupling coefficient are observed. Specifically, the reconstruction capabilities of different structural parameters for several typical aberrations such as astigmatism, trefoil aberration, coma, spherical aberration, second-order astigmatism, second-order trefoil aberration, and second-order coma are compared, and the root mean square error RMSe shown in the following formula is used to evaluate the reconstruction accuracy.
[0083]
[0084] Where n is the number of samples, Z Mi is the measured value of the aberration surface vector height measured for the ith time, Z Ti is the true value of the aberration surface vector height measured for the i-th time.
[0085] (2) Design of high-frequency and high-precision control scheme for piezoelectric deformable mirror based on hysteresis inverse model and multi-degree-of-freedom "surface shape-voltage" transfer model
[0086] The standard Zernike polynomials are used to describe the surface shape and wavefront distortion of the deformable mirror. Usually, a deformable mirror contains multiple actuators. Under the action of voltage, each actuator deforms to different degrees, thereby driving the mirror surface to deform. The channel voltages that drive the actuator to deform are sequentially composed into a control voltage signal C.
[0087] C=[c1,c2,...,c i ,...,cN ] (4)
[0088] Where N is the number of actuators in the deformable mirror, i is the actuator number, and C i represents the control voltage applied to the actuator No. i. The deformed mirror shape z(x,y) under the control voltage signal can be fitted by the standard Zernike polynomial.
[0089]
[0090] Where Z k (x, y) is the standard Zernike polynomial, and the first M terms are involved in the fitting, S k is the fitting coefficient of the kth standard Zernike polynomial. The vector composed of the fitting coefficients is defined as the surface characterization vector S.
[0091] S=[s1,s2,...,s i ,...,s N ] (6)
[0092] By fitting the influence function of each actuator using the standard Zernike polynomial and arranging them in sequence, we can obtain the influence function matrix G that reflects the mapping relationship between the deformable mirror shape and the control voltage signal:
[0093]
[0094] The relationship between the surface characterization vector S and the control voltage signal C is expressed as:
[0095] S=GC (8)
[0096] Considering the coupling between actuators, the influence function matrix is obtained using the least squares optimization method:
[0097] G=SC T (CC T ) (9)
[0098] Among them, C represents a random control voltage signal sample set that obeys Gaussian distribution, and S represents the corresponding surface shape characterization vector sample set.
[0099] Since the influence function matrix generally has unequal rows and columns, and the number of rows is usually less than the number of columns, that is, the number of standard Zernike terms used for fitting is less than the number of actuators, the solution of equation (1.7) is essentially the solution of an underdetermined linear equation system. The influence function matrix G is subjected to singular value decomposition based on the pseudo-inverse method:
[0100] G=U M×M σ M×N W N×N V T (10)
[0101] Among them, U M×M is a unitary matrix of order M, σ M×N is a semi-positive definite diagonal matrix, W N×N is an N-order unitary matrix, V T The control voltage signal can be solved as:
[0102] C=Vσ -1 U T S (11)
[0103] The principle of the gradient descent method is that if the evaluation function f(x) is differentiable at α, then f(x) decreases fastest in the opposite direction of the gradient of α. In the reconstruction of the deformable mirror, the deviation of the surface representation vector between the target surface and the current mirror surface is used as the evaluation function, as shown in the following formula:
[0104] f(x)=e n =S d -S n =S d -GC n (12)
[0105] Among them, S d is the target surface characterization vector, C n is the control voltage signal currently acting on the deformable mirror, S n is the characterization vector of the current deformable mirror surface shape, and the control voltage signal of the deformable mirror in the next iteration is:
[0106] C n+1 =C n +αG T e n (13)
[0107] The piezoelectric deformable mirror surface shape control is performed using a matrix control algorithm based on a random parallel gradient descent algorithm. Specifically, a bidirectional disturbance voltage signal δ is randomly added to the current control voltage signal. C =[δ C1 ,δ C2 ,...,δ CN ] (or addition or subtraction), and calculate the change in the evaluation function value after the bidirectional disturbance as the gradient estimate. The change direction of the control voltage signal is the direction of gradient descent. The iterative formula can be expressed as:
[0108] C n+1 =C n +γ(E + -E - )δ C (14)
[0109] Where γ is the gain coefficient, and the RMS value of the error between the target surface shape and the current surface shape of the deformable mirror is defined as the evaluation function E.
[0110] In order to further improve the control accuracy of the deformable mirror, feedforward compensation control is performed on the hysteresis of the piezoelectric actuator during the control process. The Prandtl-Ishilinskii (PI) operator model with clear expression and easy solution is used to describe the hysteresis characteristics of piezoelectric ceramics, and the parameters of its inverse model are identified. Based on this, a "pseudo-linear" system based on feedforward compensation of the hysteresis inverse model is completed.
[0111] The PI model uses a bilateral Play operator and a weighted superposition of weight functions to describe hysteresis nonlinearity, where the Play operator expression is:
[0112] H r =y(t i )=max{v(t i )-r,min{v(t i )+r,y(t i-1 )}} i=1,2,3,...n (15)
[0113] In the formula, v and y represent the input and output of the Play operator respectively, r represents the threshold, n is the number of data sampling points, y0 represents the initial output value of the Play operator, which is generally taken as 0, and H r is the Play hysteresis operator with a threshold of r. By weighted superposition of multiple Play operators with different thresholds, the PI hysteresis model describing the hysteresis voltage can be obtained:
[0114] WH=y(t i ) j ·max{v(t i )-r j ,min{v(t i )+r j ,y j (t i-1 )}}j=1,...N; i=1,...n(16)
[0115] Where N is the number of Play operators, r j is the threshold of the jth Play operator, R = [r1, r2, ..., r N ] is the threshold vector, which is usually selected by arithmetic or geometric progression within an interval, W = [w1,w2,...,w N ] is the weight vector, and H is the matrix composed of different threshold Play operators. When R is determined artificially, the model parameter W can be identified by the least squares method, and the result can be expressed as:
[0116] W=(HHT ) -1 Hy (17)
[0117] The parameters that need to be identified in the PI inverse model are the threshold vector RR and the weight vector WW, which can be obtained from the threshold and weight vector of PI:
[0118]
[0119] Example:
[0120] (1) Structural parameter identification of piezoelectric deformable mirror based on actuator influence function simulation
[0121] The method for identifying structural parameters of a piezoelectric deformable mirror based on actuator influence function simulation includes the following steps:
[0122] In the first step, the influence function of the actuator is characterized by using the improved Gaussian function shown in formula (1).
[0123] In the second step, equation (2) is used to express the stiffness of the connecting column of the piezoelectric deformable mirror.
[0124] The third step is to establish a simulation model of the piezoelectric deformable mirror in the finite element software, traverse the thickness of the mirror and the radius of the connecting column, observe the changes in the Gaussian index and coupling coefficient of the actuator influence function fitting results, and complete the selection and identification of the deformable mirror structural parameters. Specifically, compare the reconstruction capabilities of different structural parameters for several typical aberrations such as astigmatism, trefoil aberration, coma, spherical aberration, second-order astigmatism, second-order trefoil aberration, and second-order coma, and use RMSe to evaluate the reconstruction accuracy.
[0125] A specific embodiment of the present invention is described in detail below. A method for identifying the structural parameters of a deformable mirror based on actuator influence function simulation is implemented in the following manner:
[0126] In the first step, the improved Gaussian function shown in formula (1) is used to characterize the influence function of the actuator. The piezoelectric deformable mirror used in the present invention has a mirror diameter of 100 mm, the number of actuators is 55, the actuation amount reaches 100 μm, and the actuators are arranged in a hexagonal pattern. Figure 1 The figure shows the result of fitting the influence function of the piezoelectric deformable mirror using the improved Gaussian function. It can be seen that the improved Gaussian function can well fit the surface deformation of the piezoelectric deformable mirror.
[0127] In the second step, equation (2) is used to express the stiffness of the connecting column of the piezoelectric deformable mirror.
[0128] The third step is to establish a simulation model of the piezoelectric deformable mirror in the finite element software, and traverse the thickness of the mirror and the radius of the connecting column, observe the changes in the Gaussian index and coupling coefficient of the actuator influence function fitting results, and complete the selection and identification of the deformable mirror structure parameters. The results are shown in the attached figure. Figure 2 As shown. The horizontal axis represents the mirror thickness, the vertical axis represents the connection column radius, and different gray values represent the size of the Gaussian index and the coupling coefficient. Each point in the figure corresponds to a set of structural parameters. The simulation results show that as the mirror thickness increases, the Gaussian index decreases and the coupling coefficient increases. On the contrary, as the connection column radius increases, the Gaussian index increases and the coupling coefficient decreases. According to the reasonable range of the Gaussian index and the coupling coefficient, the intersection of the two is taken ( Figure 2 (c) The shaded area) can be used to obtain the reasonable range of mirror thickness and transition column radius. Three points A (1, 1.25), B (1.2, 1.25), and C (1.5, 1.5) are selected from the reasonable interval, and a point D (0.6, 1.5) is selected outside the reasonable interval. The reconstruction capabilities of the four groups of structural parameters for several typical aberrations such as astigmatism, trefoil aberration, coma, spherical aberration, second-order astigmatism, second-order trefoil aberration, and second-order coma are compared, and RMSe is used to evaluate the reconstruction accuracy.
[0129] The reconstruction error is shown in the attached Figure 3 As shown in the figure, it can be seen that the external structural parameter D outside the reasonable range has the largest reconstruction error when reconstructing the low-order aberration surface shape, while the errors of the four groups are similar when reconstructing the high-order aberration surface shape. The simulation results show that the three groups of structural parameters A, B, and C within the reasonable range have similar surface reconstruction capabilities. For the purpose of large actuator stroke and weight reduction, point A is selected as the optimized deformable mirror structural parameter.
[0130] (2) A high-frequency and high-precision control scheme for a deformable mirror based on a hysteresis inverse model and a multi-degree-of-freedom “surface shape-voltage” transfer model
[0131] In the first step, the standard Zernike polynomials shown in formula (5) are used to describe the surface shape and wavefront distortion of the deformable mirror. The channel voltages that drive the actuator to deform are sequentially composed into the control voltage signal C shown in formula (4), and the vector composed of the fitting coefficients is defined as the surface shape characterization vector S shown in formula (6).
[0132] In the second step, the influence function of each actuator is fitted using the standard Zernike polynomial and arranged in sequence, and the influence function matrix G reflecting the mapping relationship between the deformable mirror shape and the control voltage signal is represented as shown in formula (7). The influence function matrix G is obtained using the least squares optimization method shown in formula (9).
[0133] In the third step, the influence function matrix G is subjected to singular value decomposition based on the pseudo-inverse method shown in equation (10), and the control voltage signal C is solved using equation (11).
[0134] In the fourth step, the non-influence function matrix control algorithm based on the random parallel gradient descent algorithm is used to control the surface shape of the piezoelectric deformable reflector. The iterative formula is shown in formula (14).
[0135] In the fifth step, in order to further improve the control accuracy of the deformable mirror, the hysteresis of the piezoelectric actuator during the control process is feedforward compensated. The bilateral Play operator PI model shown in Equation (15) and Equation (16) is used to describe the hysteresis characteristics of the piezoelectric ceramic, and the parameter identification of its inverse model is completed. Based on this, the "pseudo-linear" system based on the hysteresis inverse model feedforward compensation is completed.
[0136] Step 6. Given the value of R, identify the PI model parameter W by the least squares method and express the result in the form shown in formula (1.17).
[0137] A specific embodiment of the present invention is described in detail below. A high-frequency and high-precision control scheme for a deformable mirror based on a hysteresis inverse model and a multi-degree-of-freedom "surface shape-voltage" transfer model is designed in the following manner:
[0138] In the first step, the first 28 terms of the standard Zernike polynomials shown in formula (5) are used to describe the surface shape and wavefront distortion of the deformable mirror with an aperture of 100 mm. The channel voltage that drives the actuator to deform is sequentially composed into the control voltage signal C shown in formula (4), and the vector composed of the fitting coefficients is defined as the surface shape characterization vector S shown in formula (6).
[0139] In the second step, the influence function of each actuator is fitted using the standard Zernike polynomial and arranged in sequence, and the influence function matrix G reflecting the mapping relationship between the deformable mirror surface shape and the control voltage signal is represented as shown in formula (7). The influence function matrix G of the deformable mirror surface with a diameter of 100 mm is obtained using the least squares optimization method shown in formula (9), and the influence function matrix G is plotted in the form of a two-dimensional matrix diagram to obtain the result, as shown in the attached figure. Figure 4 shown.
[0140] In the third step, the influence function matrix G is subjected to singular value decomposition based on the pseudo-inverse method shown in formula (10), and the control voltage signal C is solved using formula (11). The solution is shown in the attached figure. Figure 5 shown.
[0141] The fourth step is to use the non-influence function matrix control algorithm based on the random parallel gradient descent algorithm to control the surface shape of the piezoelectric deformable reflector. The iterative formula is shown in equation (14). The surface shape reconstruction structure is shown in the attached figure. Figure 6 As shown in Figure 3, after 600 iterations, the reconstructed surface shape distribution is close to the target surface shape.
[0142] In the fifth step, in order to further improve the control accuracy of the deformable mirror, the hysteresis of the piezoelectric actuator during the control process is feedforward compensated. The bilateral Play operator PI model shown in Equation (15) and Equation (16) is used to describe the hysteresis characteristics of the piezoelectric ceramic, and the parameter identification of its inverse model is completed. Based on this, the "pseudo-linear" system based on the hysteresis inverse model feedforward compensation is completed.
[0143] Specifically, assuming an ideal sinusoidal voltage signal v(t) = sin(t), the input is as shown in the following figure. Figure 7 The linear system with hysteresis (assuming k = 1) is shown in Figure 1. The output is the displacement of the piezoelectric actuator. It can be seen that due to the hysteresis effect, the output has a phase delay and a reduced amplitude. A PI model is established to describe the hysteresis behavior, as shown in the attached figure. Figure 8 As shown, the original hysteresis curve and the curve described by the PI model are highly overlapped, indicating that the selected threshold vector and the corresponding threshold vector can well simulate the hysteresis of the piezoelectric ceramic.
[0144] In the sixth step, given the value of R, the PI model parameter W is identified by the least squares method, and the result is expressed as shown in formula (17).
[0145] Specifically, for the attached Figure 8 The inverse model of the PI model is used for parameter identification, and the principle diagram of using the hysteresis inverse model to compensate for piezoelectric hysteresis is shown in the attached figure. Fig. 9 As shown, the signal v(t) exhibits hysteresis after passing through the linear system and the piezoelectric actuator, and then undergoes hysteresis compensation through the hysteresis inverse model of the piezoelectric actuator. Overall, the output y(t) remains linear with the input v(t).
[0146] Compensation effect as attached Fig.10 As shown, it can be seen that the output signal after piezoelectric hysteresis compensation is basically matched with the output signal of the system without piezoelectric hysteresis, and the hysteresis loop changes from a solid line before compensation to a dotted line after compensation, almost reaching linearity, indicating that the compensation principle based on the hysteresis inverse model can solve the problem of piezoelectric hysteresis nonlinearity.
Claims
1. A closed-loop driving and rapid and precise control method for the surface shape of a piezoelectric deformable reflector, characterized in that: The steps include: Step 1: Structural parameter identification of piezoelectric deformable mirror based on actuator influence function simulation; Step 1-1: Use the improved Gaussian function to characterize the influence function of the actuator, as shown in the following formula: Where A represents the amplitude of the Gaussian function, B represents the amplitude of the radial basis function, ω represents the coupling coefficient between the actuators, r and d represent the distance from the point on the mirror to the center of the actuator and the distribution spacing of the actuators, α represents the Gaussian index, and γ represents the radial correction width; Step 1-2: The stiffness of the connecting column is expressed as: Where k represents stiffness, E is Young's modulus, r0 and l represent the radius and length of the connecting column respectively; The reconstruction capabilities of different structural parameter pairs were compared, and the root mean square error RMSe was used to evaluate the reconstruction accuracy as shown in the following formula: Where n is the number of samples, Z Mi is the measured value of the aberration surface vector height measured for the ith time, Z Ti is the true value of the aberration surface vector height measured for the i-th time; Step 2: Design of high-frequency and high-precision control scheme for piezoelectric deformable mirror based on hysteresis inverse model and multi-degree-of-freedom "surface shape-voltage" transfer model; Step 2-1: Use the standard Zernike polynomials to describe the surface shape and wavefront distortion of the deformable mirror, and compose the control voltage signal C in sequence by the channel voltage that drives the actuator to deform: C=[c1,c2,...,c i ,...,c N ] (4) Where N is the number of actuators in the deformable mirror, i is the actuator number, and C i represents the control voltage applied to the i-th actuator; Step 2-2: The deformed mirror shape z(x,y) under the control voltage signal is fitted by the standard Zernike polynomial: Where Z k (x, y) is the standard Zernike polynomial, and the first M terms are involved in the fitting, S k is the fitting coefficient of the kth standard Zernike polynomial; Step 2-3: Define the vector composed of fitting coefficients as the surface characterization vector S: S=[s1,s2,...,s i ,...,s N ] (6) The influence function of each actuator is fitted using the standard Zernike polynomial and arranged in sequence to obtain the influence function matrix G that reflects the mapping relationship between the deformable mirror shape and the control voltage signal: The relationship between the surface characterization vector S and the control voltage signal C is expressed as: S=GC (8) Considering the coupling between actuators, the influence function matrix is obtained using the least squares optimization method: G=SC T (CC T ) (9) Based on the pseudo-inverse method, the influence function matrix G is decomposed into singular values: G=U M×M s M×N W N×N V T (10) Among them, U M×M is a unitary matrix of order M, σ M×N is a semi-positive definite diagonal matrix, W N×N is an N-order unitary matrix, V T is the conjugate transpose; The control voltage signal is solved as: C=Vσ -1 U T S (11) Step 2-4: In the reconstruction of the deformable mirror, the deviation of the surface representation vector between the target surface and the current mirror surface is used as the evaluation function, as shown in the following formula: f(x)=e n =S d -S n =S d -GC n (12) Among them, S d is the target surface characterization vector, C n is the control voltage signal currently acting on the deformable mirror, S n is the characterization vector of the current deformable mirror surface shape, and the control voltage signal of the deformable mirror in the next iteration is: C n+1 =C n +αG T e n (13) Step 2-5: Using a non-influence function matrix control algorithm based on a stochastic parallel gradient descent algorithm to control the surface shape of the piezoelectric deformable reflector; Randomly add a bidirectional disturbance voltage signal δ to the current control voltage signal C =[δ C1 ,δ C2 ,...,δ CN ], and calculate the change in the evaluation function value after the bidirectional disturbance as the gradient estimate. The change direction of the control voltage signal is the direction of gradient descent. The iterative formula is expressed as: C n+1 =C n +γ(E + -E - )d C (14) Where γ is the gain coefficient, and the RMS value of the error between the target surface shape and the current surface shape of the deformable mirror is defined as the evaluation function E; Step 2-6: The hysteresis of the piezoelectric actuator is feed-forward compensated and controlled. The PI model is used to describe the hysteresis characteristics of the piezoelectric ceramic, and the parameter identification of its inverse model is completed. Based on this, a "pseudo-linear" system based on the hysteresis inverse model feed-forward compensation is completed. The PI model uses a bilateral Play operator and a weighted superposition of weight functions to describe hysteresis nonlinearity, where the Play operator expression is: H r =y(t i )=max{v(t i )-r,min{v(t i )+r,y(t i-1 )}} i=1,2,3,...n (15) In the formula, v and y represent the input and output of the Play operator respectively, r represents the threshold, n is the number of data sampling points, y0 represents the initial output value of the Play operator, and H r is the Play hysteresis operator with a threshold of r; By weighted superposition of multiple Play operators with different thresholds, the PI hysteresis model describing the hysteresis voltage is obtained: WH=y(t i )w j ·max{v(t i )-r j ,min{v(t i )+r j ,y j (t i-1 )}}j=1,...N;i=1,...n(16) Where N is the number of Play operators, r j is the threshold of the jth Play operator, R = [r1, r2, ..., r N ] is the threshold vector, W=[w1,w2,...,w N ] is the weight vector, H is the matrix composed of Play operators with different thresholds; When R is determined artificially, the model parameter W is identified by the least squares method, and the result is expressed as: W=(HH T ) -1 Hy (17) The parameters that need to be identified in the PI inverse model are the threshold vector RR and the weight vector WW, which are obtained from the threshold and weight vector of PI:
2. The closed-loop driving and rapid and precise control method of the piezoelectric deformable reflector surface shape according to claim 1 is characterized in that: The coupling coefficient ω is between 5% and 12%, and the Gaussian exponent α is between 1.9 and 2.
3.
3. The closed-loop driving and rapid and precise control method of the piezoelectric deformable reflector surface shape according to claim 1 is characterized in that: The different pairs of structural parameters include astigmatism, trefoil aberration, coma, spherical aberration, second-order astigmatism, second-order trefoil aberration, and second-order coma.
4. The closed-loop driving and rapid and precise control method of the piezoelectric deformable reflector surface shape according to claim 1 is characterized in that: The initial output value y0 of the Play operator is taken as 0.
Citation Information
Cited By
Piezoelectric deformable mirror flattening control method
CN120065499A
