An iterative learning robust control method for magnetic suspension worktable based on error tracking

By using an iterative learning robust control method of error tracking on the magnetic levitation workbench, combining adaptive robust control and iterative learning control terms, the stability and tracking error problems under the influence of initial position deviation are solved, and high accuracy and stability are achieved Multi-degree of freedom magnetic levitation table control.

CN115236987BInactive Publication Date: 2024-12-06WUHAN UNIV

Patent Information

Application Number
CN202210913606.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-27
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

When the existing multi-degree-of-freedom magnetic levitation table control method exists, the system stability is affected, the tracking error is large, and iterative learning control depends on the condition that there is no initial position deviation, and the accumulation of initial error cannot be effectively suppressed. .

Method used

Using the iterative learning robust control method of magnetic levitation workbench based on error tracking, designing nonlinear system models through dynamic decoupling, combining the expected state trajectory and attenuation function, and using the Lyapunov method to analyze stability and error convergence, construct a parallel structure controller to overcome the influence of initial position error.

Benefits of technology

It effectively eliminates the impact of the system's initial position error on iterative learning control, improves the tracking accuracy and stability of the control system, and ensures the high robustness and parameter adaptability of the magnetic levitation workbench under initial error conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115236987B_ABST
    Figure CN115236987B_ABST
Patent Text Reader

Abstract

An iterative learning robust control method for a magnetic levitation workbench based on error tracking utilizes the dynamic decoupling of a magnetic levitation planar motor to transform multi-freedom control into independent single-degree-of-freedom, constructs a nonlinear system model, combines the desired state trajectory, and utilizes the attenuation function to design the desired error trajectory. The sliding surface is constructed based on the adaptive robust control term to ensure the stability of the magnetic levitation workbench and eliminate the uncertain terms. The adaptive robust control term is combined with the iterative learning control term through a parallel structure to obtain an iterative learning robust controller for the magnetic levitation workbench based on error tracking. The stability and error convergence are analyzed, and the control algorithm is applied to an actual magnetic levitation workbench to verify the effectiveness of the control algorithm. The method can overcome the limitations of the existing control method for multi-degree-of-freedom magnetic levitation workbench, while taking into account the tracking accuracy and suppressing the influence of the initial position error of the magnetic levitation workbench on the iterative learning control method.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of magnetic suspension workbench control, in particular to an iterative learning robust control method for a magnetic suspension workbench based on error tracking. Background Art

[0002] As a new type of driving element, the magnetic levitation workbench is one of the core components of precision machining and manufacturing equipment, and its motion performance is the key to ensuring the processing performance of the equipment. Compared with traditional mechanical superposition or air-floating drive platforms, the magnetic levitation workbench can drive a single rigid body to achieve multi-degree-of-freedom motion through electromagnetic force, and has the advantages of frictionless, vacuum operation, and heat source separation, which is more in line with the current process requirements in the field of precision machining and manufacturing.

[0003] Adaptive robust control has good robustness and system parameter adaptability, but it only suppresses the state-related uncertainties and disturbances in the system through robust feedback, resulting in a large steady-state tracking error. The iterative learning control method can effectively compensate for model uncertainties and state follow-up disturbances without relying on an accurate system model. Combining adaptive robust control with iterative learning control can effectively improve the trajectory tracking performance of the magnetic levitation worktable.

[0004] However, the control method based on iterative learning relies on the condition that the system does not have an initial position deviation. The initial position deviation in the system will continue to accumulate with the increase of iteration cycles, affecting the stability of the entire system. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an iterative learning robust control method for a magnetic levitation worktable based on error tracking, which can overcome the limitations of the existing multi-degree-of-freedom magnetic levitation worktable control method, take into account the tracking accuracy while suppressing the influence of the system's initial position error on the iterative learning control method.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] An iterative learning robust control method for a magnetic suspension workbench based on error tracking comprises the following steps:

[0008] Step 1: Use the dynamic decoupling of the magnetic levitation planar motor to convert the multi-freedom control into independent single degrees of freedom for design, and build a nonlinear system model for each degree of freedom;

[0009] Step 2: Based on the nonlinear system model constructed in step 1 and in combination with the expected state trajectory, the expected error trajectory is designed using the attenuation function, the sliding surface is constructed based on the error tracking deviation and the error tracking deviation derivative, the adaptive robust control term is designed to ensure the stability of the magnetic levitation workbench, and the iterative learning control term is designed to eliminate the model uncertainty term of the magnetic levitation workbench. The adaptive robust control term is combined with the iterative learning control term through a parallel structure, thereby obtaining an iterative learning robust controller of the magnetic levitation workbench based on error tracking;

[0010] Step 3, using the Lyapunov direct method to analyze the stability and error convergence of the robust iterative learning control method of the magnetic suspension workbench based on error tracking obtained in step 2;

[0011] Step 4: Apply the control algorithm to the actual magnetic levitation workbench, perform trajectory tracking in the presence of initial position error and interference, and verify the effectiveness of the iterative learning robust controller of the magnetic levitation workbench based on error tracking.

[0012] Furthermore, in step 1, based on the dynamic decoupling strategy, each degree of freedom can be controlled as an independent single-input single-output system for closed-loop control. Therefore, without losing generality, taking the x-axis translation as an example (x-axis rotation, y-axis translation, y-axis rotation, z-axis translation, z-axis rotation) at the kth iteration, the nonlinear system of the x-axis translation of the magnetic suspension worktable is It is expressed as:

[0013]

[0014] Among them, x k Indicates the displacement of the worktable along the x-axis in the kth iteration, u k (t) is the control output of the kth iteration, θ is the parameter of the magnetic suspension workbench, f(x k , t) is the combination of uncertainty related to the state of the magnetic suspension workbench and state follow-up disturbance, Δ d represents a non-repetitive external disturbance;

[0015] Define the trajectory tracking error e in the kth iteration k , e k The calculation formula is:

[0016] e k =x d -x k (2)

[0017] Among them, x d represents the reference trajectory point;

[0018] According to formula (3), formula (4) can be rewritten as follows:

[0019]

[0020] Similarly, a nonlinear system of rotation of the magnetic levitation worktable in the x-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of rotation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the z-axis direction, and a nonlinear system of rotation of the magnetic levitation worktable in the z-axis direction are constructed.

[0021] Furthermore, the tracking error e k It is necessary to follow the expected error trajectory to converge within the entire trajectory period [0, T], so as to make corresponding transitions at the beginning so that the actual state trajectory x of the magnetic suspension workbench k Can be in [T r , T] follows the desired state trajectory, T r is the correction time, correction time T r Satisfy the condition 0<T r <T, correction time T r It needs to be predetermined according to actual needs and the range of control input. Achieving good state tracking requires a correction process for the initial state error. The expected error trajectory The design is as follows:

[0022]

[0023] Among them, t is a certain moment and ζ(t) is the attenuation function.

[0024] Furthermore, the decay function ζ(t) needs to ensure the desired error trajectory It is continuously differentiable in the time domain.

[0025] Furthermore, the attenuation function ζ(t) is constructed as:

[0026]

[0027] Furthermore, the tracking deviation ε of the error is defined as k (t):

[0028]

[0029] in, is the trajectory tracking error in the kth iteration, is the expected error trajectory;

[0030] Then the tracking deviation derivative of the error is

[0031]

[0032] According to formula (6) and formula (7), the sliding surface s based on the error tracking deviation can be constructed k (t):

[0033]

[0034] Where c represents the sliding surface parameter, c>0.

[0035] Furthermore, if the control output can make s k (t) tends to zero, then the tracking error can also coincide with the expected error trajectory;

[0036] Time derivative of sliding surface It can be expressed as:

[0037]

[0038] The second derivative of the tracking error Substituting formula (2) into formula (9), we can get u(t) and The relationship between them is as follows:

[0039]

[0040] in, represents the model compensation at the kth iteration;

[0041] According to formula (10), the calculation formula of the control output of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is:

[0042]

[0043] in, represents the control output of the iterative learning control term, is the output of the adaptive robust control term;

[0044] According to formula (10), the output of the adaptive robust control term of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is The calculation formula is:

[0045]

[0046] Among them, w k (t) represents the improved switch term, which helps to reduce the chattering of the control magnetic suspension worktable, w k The calculation formula of (t) is as follows:

[0047] w k (t) = ηk ·sgn(s k )+l s s k (t) (13)

[0048] Among them, sgn(·) is the switching function, l s is the feedback compensation coefficient. For simplicity, sgn(s k (t)) is abbreviated as sgn(s k ), magnetic levitation workbench estimated parameters Generated by the adaptive term in the adaptive robust control term, the parameter is estimated based on the parameter boundary using a parameter adaptive algorithm based on discontinuous mapping. It is derived by the following formula:

[0049]

[0050] Among them, γ is a fixed parameter estimation adaptation rate. This mapping correction method can ensure Always keep within the allowable range of parameters and do not affect its convergence. In addition, when choosing the initial value of the estimated parameter When

[0051] Furthermore, the step 3 specifically includes the following steps:

[0052] Step 301: Use Lyapunov direct method to construct the energy function E of the kth iteration k (t):

[0053]

[0054] in, is the learning error of iterative learning, which can be Calculate and obtain;

[0055] Bias in parameter estimates Able to pass Calculate and obtain;

[0056] Step 302: derive the E between two adjacent iterations. k The difference ΔE of (t) k , as shown below:

[0057]

[0058] According to V k Definition, the difference between two consecutive iterations ΔV k Expressed as:

[0059]

[0060] There is no initial tracking error bias, Can be rewritten in integral form Based on this expression, by substituting formula (12) and formula (13) into formula (17), we can obtain:

[0061]

[0062] Similarly, The derivative in the iteration domain can be expanded into the following form:

[0063]

[0064] Among them, ξ is the learning rate;

[0065] For any constant a and For example, the semi-saturated function It can satisfy the following inequality:

[0066]

[0067] Therefore, according to the properties of the saturation function and the learning law of the iterative learning term, the integrand in formula (18) is rewritten as follows:

[0068]

[0069] Substituting it for the corresponding integral term in formula (19) yields:

[0070]

[0071] according to The definition of , between the kth and k-1th iterations, the difference It can be expressed as:

[0072]

[0073] According to the iterative update law (11) and combined with formulas (16)-(23), the time between the kth and k-1th iterations is k The difference can be simplified to:

[0074]

[0075] If the desired switching gain η d Satisfy the condition η d ≥Δ max , then the above inequality is guaranteed to hold;

[0076] According to the definition of Lyapunov functional, it can be verified that E0(t)≥0, Ek (t) ≥ 0 and E0(t) is bounded, which means that when k → ∞, s k It can converge to zero asymptotically under the L2 norm definition, because is a bounded value, and it is possible to determine s k (t) have continuous consistency;

[0077] Based on Barbalat's lemma, we can conclude that Tracking error satisfies When t∈[T r , T], Therefore, the iterative learning robust controller of the magnetic levitation worktable based on error tracking can ensure the monotonic convergence of the magnetic levitation worktable.

[0078] Furthermore, the output of the iterative learning control term It can effectively suppress the uncertainty related to the unmodeled state f(x k , t) influence, therefore, in the iterative learning update law, a partial saturation learning strategy is introduced to ensure the boundedness of the iterative learning result. Therefore, the output of the ILC control item Updated by the following learning rule:

[0079]

[0080] in, is the control output of the previous iteration, where ξ represents the learning gain, is a saturation function, and its specific expression is:

[0081]

[0082] Among them, ε d It is the upper boundary value of the uncertainty item of the magnetic levitation workbench.

[0083] Furthermore, if the effectiveness of the iterative learning robust controller of the magnetic levitation workbench based on error tracking is not strong, adjust the learning rate or increase the number of iterations.

[0084] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0085] The present invention provides an iterative learning robust control method for a magnetic levitation workbench based on error tracking. By designing the expected error trajectory and using error tracking instead of traditional position tracking, the influence of the system's initial position error on iterative learning control is eliminated, and the initial error correction capability and the stability of the control system are improved.

[0086] The control method of the present invention can solve the tracking control problem of a multi-degree-of-freedom magnetically suspended planar motor under the condition that the system has an initial iterative position error, and the algorithm has strong robustness and parameter adaptability, and has high steady-state tracking accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 Schematic diagram of a multi-degree-of-freedom magnetically levitated planar motor.

[0088] Figure 2 It is a flow chart of the steps of the control method of the present invention.

[0089] Figure 3 This is a structural diagram of an iterative learning robust control method for a magnetic suspension workbench based on error tracking according to the present invention.

[0090] Figure 4 This is a diagram of the position tracking error convergence process during the final folding process of the control method obtained by actually applying the present invention to the magnetic levitation system. DETAILED DESCRIPTION

[0091] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0092] An iterative learning robust control method for a magnetic suspension worktable based on error tracking, such as Figure 1 and Figure 2 As shown, the following steps are included:

[0093] Step 1: Use the dynamic decoupling of the magnetic levitation planar motor to convert the multi-freedom control into independent single degrees of freedom for design, and build a nonlinear system model for each degree of freedom;

[0094] Step 2: Based on the nonlinear system model constructed in step 1 and in combination with the expected state trajectory, the expected error trajectory is designed using the attenuation function, the sliding surface is constructed based on the error tracking deviation and the error tracking deviation derivative, the adaptive robust control term is designed to ensure the stability of the magnetic levitation workbench, and the iterative learning control term is designed to eliminate the model uncertainty term of the magnetic levitation workbench. The adaptive robust control term is combined with the iterative learning control term through a parallel structure, thereby obtaining an iterative learning robust controller of the magnetic levitation workbench based on error tracking;

[0095] Step 3, using the Lyapunov direct method to analyze the stability and error convergence of the robust iterative learning control method of the magnetic suspension workbench based on error tracking obtained in step 2;

[0096] Step 4: Apply the control algorithm to the actual magnetic levitation workbench, perform trajectory tracking when the magnetic levitation workbench has initial position error and interference, and verify the effectiveness of the control algorithm.

[0097] The present invention eliminates the influence of the initial position error of the magnetic suspension workbench on iterative learning control by designing the expected error trajectory and using error tracking instead of traditional position tracking, thereby improving the correction capability for the initial error and the stability of the control system.

[0098] The control method of the present invention can solve the tracking control problem of a multi-degree-of-freedom magnetic levitation planar motor under the condition of initial iterative position error of a magnetic levitation workbench, and the algorithm has strong robustness and parameter adaptability, and high steady-state tracking accuracy.

[0099] In the present invention, in step 1, based on the dynamic decoupling strategy, each degree of freedom can be controlled in a closed loop as an independent single-input single-output system. Therefore, without losing generality, taking the translation in the x-axis direction as an example (the same applies to the rotation in the x-axis direction, the translation in the y-axis direction, the rotation in the y-axis direction, the translation in the z-axis direction, and the rotation in the z-axis direction), at the kth iteration, the nonlinear system of the x-axis translation of the magnetic suspension worktable is expressed as

[0100]

[0101] Among them, x k Indicates the displacement of the worktable along the x-axis in the kth iteration, u k (t) is the control output of the kth iteration, θ is the parameter of the magnetic suspension workbench, f(x k , t) is the combination of uncertainty related to the state of the magnetic suspension workbench and state follow-up disturbance, Δ d represents a non-repetitive external disturbance;

[0102] Define the trajectory tracking error e in the kth iteration k , e k The calculation formula is:

[0103] e k =x d -x k (2)

[0104] Among them, x d represents the reference trajectory point;

[0105] According to formula (5), formula (6) can be rewritten as follows:

[0106]

[0107] Similarly, a nonlinear system of rotation of the magnetic levitation worktable in the x-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of rotation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the z-axis direction, and a nonlinear system of rotation of the magnetic levitation worktable in the z-axis direction are constructed.

[0108] For state tracking, the goal is to make the output of the magnetic levitation table track the desired state trajectory, that is, the tracking error converges to zero, while for error tracking, the goal is to make the tracking error itself converge according to the desired error convergence trajectory, so that the tracking error converges to the desired error trajectory.

[0109] Different from designing a controller based on the expected state trajectory, the expected error trajectory itself is independent of the expected state trajectory, and the same expected error trajectory can be applied to tasks with different expected state trajectories.

[0110] Therefore, in the reference error tracking strategy, the tracking error e k It is necessary to follow the expected error trajectory to converge within the entire trajectory period [0, T], so as to make corresponding transitions at the beginning so that the actual state trajectory x of the magnetic suspension workbench k Able to r , T] to follow the desired state trajectory.

[0111] For the non-zero initial state error that may exist in the magnetic suspension workbench, since the actual control output is limited, a correction process for the initial state error is required to achieve good state tracking. Correction time T r Satisfy the condition 0<T r <T, its size needs to be predetermined according to actual needs and the range of control input. The construction of the expected error trajectory is to correct the initial error of the magnetic suspension workbench. The design is as follows:

[0112]

[0113] Among them, ζ(t) is the decay function;

[0114] In practice, ζ(t) needs to ensure the expected error trajectory It is continuously differentiable in the time domain, and the construction method of the attenuation function ζ(t) is not unique. It only needs to meet the conditions of asymptotic convergence and high-order differentiability.

[0115] The decay function ζ(t) is constructed as:

[0116]

[0117] The present invention only takes formula (5) as an example.

[0118] The tracking deviation ε is defined as the error k (t):

[0119]

[0120] Then the tracking deviation derivative of the error is

[0121]

[0122] According to formula (6) and formula (7), the sliding surface s based on the error tracking deviation can be constructed k (t):

[0123]

[0124] Where c represents the sliding surface parameter, c>0.

[0125] If the control output can make s k (t) tends to zero, then the tracking error can also coincide with the expected error trajectory.

[0126] Time derivative of sliding surface It can be expressed as:

[0127]

[0128] The second derivative of the tracking error Substituting formula (2) into formula (9), we can get u(t) and The relationship between them is as follows:

[0129]

[0130] in, represents the model compensation at the kth iteration;

[0131] like Figure 3 As shown in the control structure block diagram, according to formula (10), the calculation formula of the control output of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is:

[0132]

[0133] in, represents the control output of the iterative learning control term, is the output of the adaptive robust control term.

[0134] According to formula (10), the output of the adaptive robust control term of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is The calculation formula is:

[0135]

[0136] Among them, w k(t) represents the improved switch term, which helps to reduce the chattering of the control magnetic suspension worktable, w k The calculation formula of (t) is as follows:

[0137] ω k (t) = η k ·sgn(s k )+l s s k (t) (13)

[0138] Among them, sgn(·) is the switching function, l s is the feedback compensation coefficient. For simplicity, sgn(s k (t)) is abbreviated as sgn(s k ). Estimated parameters of magnetic levitation workbench Generated by the adaptive term in the adaptive robust control term. According to the parameter boundary, the parameter adaptive algorithm based on discontinuous mapping is used to estimate the parameter It is derived by the following formula:

[0139]

[0140] Among them, γ is a fixed parameter estimation adaptation rate. This mapping correction method can ensure Always keep within the allowable range of parameters and do not affect its convergence. In addition, when choosing the initial value of the estimated parameter When

[0141] In the present invention, in step 3, the stability of the obtained iterative learning robust controller based on error tracking and the convergence of the error are analyzed using the Lyapunov direct method, which specifically includes the following steps:

[0142] Step 301: Use Lyapunov direct method to construct the energy function E of the kth iteration k (t):

[0143]

[0144] in, is the learning error of iterative learning, which can be Calculate and obtain;

[0145] Bias in parameter estimates Able to pass Calculate and obtain;

[0146] Step 302: derive the E between two adjacent iterations. k The difference ΔE of (t) k, as shown below:

[0147]

[0148] According to V k Definition, the difference between two consecutive iterations ΔV k Expressed as:

[0149]

[0150] There is no initial tracking error bias, Can be rewritten in integral form Based on this expression, by substituting formula (12) and formula (13) into formula (17), we can obtain:

[0151]

[0152] Similarly, The derivative in the iteration domain can be expanded into the following form:

[0153]

[0154] Among them, ξ is the learning rate;

[0155] For any constant a and For example, the semi-saturated function It can satisfy the following inequality:

[0156]

[0157] Therefore, according to the properties of the saturation function and the learning law of the iterative learning term, the integrand in formula (18) is rewritten as follows:

[0158]

[0159] Substituting it for the corresponding integral term in formula (19) yields:

[0160]

[0161] according to The definition of , between the kth and k-1th iterations, the difference It can be expressed as:

[0162]

[0163] According to the iterative update law (11) and combined with formulas (16)-(23), the time between the kth and k-1th iterations is k The difference can be simplified to:

[0164]

[0165] If the desired switching gain η d Satisfy the condition η d ≥Δ max , then the above inequality is guaranteed to hold.

[0166] According to the definition of Lyapunov functional, it can be verified that E0(t)≥0, E k (t) ≥ 0 and E0(t) is bounded, which means that when k → ∞, s k It can converge to zero asymptotically under the L2 norm definition, because is a bounded value, and it is possible to determine s k (t)Has continuous consistency.

[0167] Based on Barbalat's lemma, we can conclude that Tracking error satisfies When t∈[T r , T], Therefore, the iterative learning robust controller of the magnetic levitation workbench based on error tracking proposed in the present invention can ensure the monotonic convergence of the magnetic levitation workbench.

[0168] In the present invention, Figure 2 As shown, the output of the iterative learning control term It can effectively suppress the uncertainty related to the unmodeled state f(x k , t). Therefore, in the iterative learning update law, a partial saturation learning strategy is introduced to ensure the boundedness of the iterative learning result. Therefore, the output of the ILC control item Updated by the following learning rule:

[0169]

[0170] in, is the control output of the previous iteration, where ξ represents the learning gain, is a saturation function, and its specific expression is:

[0171]

[0172] Among them, ε d It is the upper boundary value of the uncertainty item of the magnetic levitation workbench.

[0173] In one embodiment of the present invention, the process of applying the proposed control method to the magnetic suspension workbench is as follows: first, the structural parameters of the controller are determined according to the magnetic suspension workbench, c = 10, g = 60, and the parameter boundary of the magnetic suspension workbench is set to θ min =2.34kg,θ max =4kg, ε d=0.158,Δ max =1.25, learning rate ξ = 0.25, correction time is set to T r =0.02s.

[0174] In order to test the tracking performance of the controller, a typical sinusoidal trajectory tracking experiment along the horizontal x-axis is carried out. d For: x d = sin(πt)(mm), where the tracking angular velocity of the trajectory is 3.14 rad / s and the speed is 1.57 cos(πt) mm / s. d is the expected trajectory point. Set the initial position of each iteration to x k (0) = (0.09 + 0.02rand (k)) mm, so that in each iteration, the magnetic suspension worktable has an initial error of about 0.1 mm. The root mean square error of the evaluation index e rms The maximum absolute value of the sum error e max Evaluate the performance of the magnetic levitation workbench.

[0175] The actual control effect of the control method is as follows Figure 4 As shown in the figure, according to the changes in the tracking error and the actual tracking trajectory, it can be seen that after multiple iterations of learning, the actual motion trajectory of the magnetic levitation workbench is very close to the reference trajectory, and the tracking root mean square error is less than 1.4um.

[0176] When there is an initial position error in the magnetic levitation worktable, after a finite number of iterations, the proposed control method can effectively eliminate the influence of the initial position error and can correct the initial error according to the pre-designed expected error trajectory, thereby achieving better steady-state tracking accuracy.

[0177] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A robust control method for iterative learning of a magnetic suspension worktable based on error tracking, characterized in that: The following steps are involved: Step 1: Use the dynamic decoupling of the magnetic levitation planar motor to convert the multi-freedom control into independent single degrees of freedom for design, and build a nonlinear system model for each degree of freedom; Step 2: Based on the nonlinear system model constructed in step 1 and in combination with the expected state trajectory, the expected error trajectory is designed using the attenuation function, the sliding surface is constructed based on the error tracking deviation and the error tracking deviation derivative, the adaptive robust control term is designed to ensure the stability of the magnetic levitation workbench, and the iterative learning control term is designed to eliminate the model uncertainty term of the magnetic levitation workbench. The adaptive robust control term is combined with the iterative learning control term through a parallel structure, thereby obtaining an iterative learning robust controller of the magnetic levitation workbench based on error tracking; Step 3, using the Lyapunov direct method to analyze the stability and error convergence of the robust iterative learning control method of the magnetic suspension workbench based on error tracking obtained in step 2; Tracking Error k It is necessary to follow the expected error trajectory to converge within the entire trajectory period [0, T], so as to make corresponding transitions at the beginning so that the actual state trajectory x of the magnetic suspension workbench k Able to r , T] follows the desired state trajectory, T r is the correction time, correction time T r Satisfy the condition 0<T r <T, correction time T r It needs to be predetermined according to actual needs and the range of control input. Achieving good state tracking requires a correction process for the initial state error. The expected error trajectory The design is as follows: Among them, t is a certain moment, ζ(t) is the decay function; The tracking deviation ε of the error is defined k (t): Among them, e k (t) is the tracking error in the kth iteration; Then the tracking deviation derivative of the error is According to formula (6) and formula (7), the sliding surface s based on the error tracking deviation can be constructed k (t): Where c represents the sliding surface parameter, c>0; Step 4: Apply the control algorithm to the actual magnetic levitation workbench, perform trajectory tracking in the presence of initial position error and interference, and verify the effectiveness of the iterative learning robust controller of the magnetic levitation workbench based on error tracking.

2. The method for iterative learning robust control of a magnetic suspension workbench based on error tracking according to claim 1 is characterized in that: In step 1, based on the dynamic decoupling strategy, each degree of freedom can be controlled as an independent single-input single-output system for closed-loop control. Therefore, without losing generality, taking the x-axis translation as an example, the x-axis rotation, y-axis translation, y-axis rotation, z-axis translation, and z-axis rotation are similar. At the kth iteration, the nonlinear system of the x-axis translation of the magnetic suspension worktable is It is expressed as: Among them, x k Indicates the displacement of the worktable along the x-axis in the kth iteration, u k (t) is the control output of the kth iteration, θ is the parameter of the magnetic suspension workbench, f(x k , t) is the combination of uncertainty related to the state of the magnetic suspension workbench and state follow-up disturbance, Δ d represents a non-repetitive external disturbance; Define the trajectory tracking error e in the kth iteration k , e k The calculation formula is: e k =x d -x k (2) Among them, x d represents the reference trajectory point; According to formula (1), formula (2) can be rewritten as follows: Similarly, a nonlinear system of rotation of the magnetic levitation worktable in the x-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of rotation of the magnetic levitation worktable in the y-axis direction, a nonlinear system of translation of the magnetic levitation worktable in the z-axis direction, and a nonlinear system of rotation of the magnetic levitation worktable in the z-axis direction are constructed.

3. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 1 is characterized in that: The decay function ζ(t) needs to ensure the expected error trajectory It is continuously differentiable in the time domain.

4. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 3 is characterized in that: The decay function ζ(t) is constructed as:

5. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 1 is characterized in that: If the control output can make s k (t) tends to zero, then the tracking error can also coincide with the expected error trajectory; Time derivative of sliding surface It can be expressed as: The second derivative of the tracking error Substituting formula (2) into formula (9), we can get u(t) and The relationship between them is as follows: in, represents the model compensation at the kth iteration; According to formula (10), the calculation formula of the control output of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is: in, represents the control output of the iterative learning control term, is the output of the adaptive robust control term; According to formula (10), the output of the adaptive robust control term of the iterative learning robust controller of the magnetic suspension workbench based on error tracking at the kth iteration is The calculation formula is: Among them, w k (t) represents the improved switch term, which helps to reduce the chattering of the control magnetic suspension worktable, w k The calculation formula of (t) is as follows: w k (t)=η k ·sgn(s k )+l s s k (t) (13) Among them, sgn(·) is the switching function, l s is the feedback compensation coefficient. For simplicity, sgn(s k (t)) is abbreviated as sgn(s k ), magnetic levitation workbench estimated parameters Generated by the adaptive term in the adaptive robust control term, the parameter is estimated based on the parameter boundary using a parameter adaptive algorithm based on discontinuous mapping. It is derived by the following formula: Among them, γ is a fixed parameter estimation adaptation rate. This mapping correction method can ensure Always keep within the allowable range of parameters and do not affect its convergence. In addition, when choosing the initial value of the estimated parameter When 6. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 1 is characterized in that: The step 3 specifically comprises the following steps: Step 301: Use Lyapunov direct method to construct the energy function E of the kth iteration k (t): in, is the learning error of iterative learning, which can be Calculate and obtain; Bias in parameter estimates Able to pass Calculate and obtain; Step 302: derive the E between two adjacent iterations. k The difference ΔE of (t) k , as shown below: According to V k Definition, the difference between two consecutive iterations ΔV k It is expressed as: There is no initial tracking error bias, Can be rewritten in integral form Based on this expression, by substituting formula (12) and formula (13) into formula (17), we can obtain: Similarly, The derivative in the iteration domain can be expanded into the following form: Among them, ξ is the learning rate; For any constant a and For example, the semi-saturated function It can satisfy the following inequality: Therefore, according to the properties of the saturation function and the learning law of the iterative learning term, the integrand in formula (18) is rewritten as follows: Substituting it for the corresponding integral term in formula (19) yields: according to The definition of , between the kth and k-1th iterations, the difference It can be expressed as: According to the iterative update law (11) and combined with formulas (16)-(23), the time between the kth and k-1th iterations is k The difference can be simplified to: If the desired switching gain η d Satisfy the condition η d ≥Δ max , then the above inequality is guaranteed to hold; According to the definition of Lyapunov functional, it can be verified that E0(t)≥0, E k (t) ≥ 0 and E0(t) is bounded, which means that when k → ∞, s k It can converge to zero asymptotically under the L2 norm definition, because is a bounded value, and it is possible to determine s k (t) have continuous consistency; Based on Barbalat's lemma, we can conclude that Tracking error satisfies When t∈[T r , T], Therefore, the iterative learning robust controller of the magnetic levitation worktable based on error tracking can ensure the monotonic convergence of the magnetic levitation worktable.

7. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 5 is characterized in that: Output of iterative learning control term It can effectively suppress the uncertainty related to the unmodeled state f(x k , t) influence, therefore, in the iterative learning update law, a partial saturation learning strategy is introduced to ensure the boundedness of the iterative learning result. Therefore, the output of the ILC control item Updated by the following learning rule: in, is the control output of the previous iteration, where ξ represents the learning gain, is a saturation function, and its specific expression is: Among them, ε d It is the upper boundary value of the uncertainty item of the magnetic levitation workbench.

8. The method for iterative learning robust control of a magnetic suspension worktable based on error tracking according to claim 6 is characterized in that: If the iterative learning robust controller of the magnetic levitation workbench based on error tracking is not effective, adjust the learning rate or increase the number of iterations.

Citation Information

Patent Citations

  • Self-adaptive learning sliding mode control method for multi-degree-of-freedom magnetic suspension planar motor

    CN112859618A

  • Robust heuristic iterative learning control method for metal bar temperature distribution system

    WO2022088857A1

Cited By

  • Multi-degree-of-freedom motion control system and method for high-precision magnetic suspension motor

    CN120222852A

  • Multi-degree-of-freedom motion control system and method for magnetic levitation motor

    CN120222852B