Magnetic suspension micro-motion working platform stabilizing method and system based on stress sensing

Abstract

The invention discloses a magnetic suspension micro-motion working platform stabilizing method and system based on stress sensing. The method is applied to magnetic suspension working platforms. A fiber bragg grating stress sensor is arranged on the magnetic suspension working platform; and in each time slot tau, sensing the stress of the magnetic suspension micro-motion working platform, measuring the influence of the comprehensive action of the operation factors and the environmental factors on the magnetic suspension micro-motion platform, and determining the compensation current of the stator winding of the magnetic suspension micro-motion working platform in combination with a Markov decision process according to the stress state of the magnetic suspension micro-motion platform. Therefore, the driving force of the magnetic suspension micro-motion platform is adjusted, the influence of operation and environmental factors is counteracted by continuously iteratively adjusting the driving current of the stator winding, and the position stability of the magnetic suspension micro-motion platform is maintained.

Description

Technical Field

[0001] The present invention belongs to the field of precision machining, and more specifically, relates to a method and system for stabilizing a magnetic suspension micro-motion working platform based on stress sensing. Background Art

[0002] With the rapid development of science and technology, the magnetic suspension micro-motion worktable has become a key equipment in the fields of precision manufacturing, micro-nano processing and high-precision measurement. Its unique non-contact control method and high sensitivity make it stand out in application scenarios that pursue high stability and high precision. However, the performance of the magnetic suspension micro-motion worktable is deeply affected by the accuracy and reliability of its sensor system and position and posture measurement technology.

[0003] The magnetic suspension micro-motion workbench uses non-contact magnetic suspension drive technology for the motion platform and drive mechanism, eliminating the influence of friction and wear on motion accuracy, so it is easy to achieve large-scale high-precision micro-motion. The magnetic suspension micro-motion workbench is generally composed of electromagnetic suspension mechanism, drive circuit, precision measurement and motion control, and can achieve micron / nano-level micro-motion. However, when the magnetic suspension motion workbench is mechanically processed, it is inevitably affected by the processing operation and uncontrollable environmental factors, and position fluctuations occur. Improving the stability of the magnetic suspension platform at a specific position is an important function of motion control. Summary of the invention

[0004] In response to the above defects or improvement needs of the prior art, the present invention provides a stress-sensing-based magnetic levitation micro-motion work platform stabilization method and system, the purpose of which is to measure the combined influence of operating factors and environmental factors on the magnetic levitation micro-motion work platform when it is suspended through a stress sensor, thereby performing current compensation, adjusting the driving force of the magnetic levitation micro-motion work platform, and offsetting the fluctuations caused by operating factors and environmental factors, thereby maintaining the position of the magnetic levitation micro-motion work platform unchanged, improving the stability of the magnetic levitation micro-motion work platform, and thereby solving the technical problem that the stability of the existing magnetic levitation micro-motion work platform is affected by operating factors and environmental factors.

[0005] To achieve the above-mentioned purpose, according to one aspect of the present invention, a method for stabilizing a magnetic suspension micro-motion work platform based on stress sensing is provided, which is applied to a magnetic suspension work platform; a fiber Bragg grating stress sensor is arranged on the magnetic suspension work platform;

[0006] In each time slot τ, the following steps are performed:

[0007] (1) Read the stress value vector of the fiber grating array using the spectrum detector is the reading of the i-th fiber Bragg grating stress sensor;

[0008] (2) Obtain the output values ​​of each posture sensor of the magnetic levitation platform as the current posture vector of the magnetic levitation platform Pose value is the output of the jth position or attitude sensor; the position sensor of the magnetic levitation platform is used to determine the position of a specific sampling point of the magnetic levitation platform; the attitude sensor of the magnetic levitation platform is used to determine the deflection state of the magnetic levitation platform;

[0009] (3) According to the stress value vector S(τ) obtained in step (1) and the posture vector P(τ) obtained in step (2), a model-free reinforcement learning algorithm based on Markov process is used to determine the stator winding current compensation value of the magnetic suspension micro-motion platform in the next time slot τ+1. is the current compensation value of the kth stator winding;

[0010] (4) For the kth stator winding, the current value at the current moment is The sum of the current compensation value at the next moment As the current value of the stator winding at the next moment Right now

[0011] Preferably, in the stress-sensing-based magnetic suspension micro-motion work platform stabilization method, the magnetic suspension work platform has a fiber Bragg grating stress sensor array evenly arranged on it, and the direction of the fiber Bragg grating is parallel to the diagonal direction of the magnetic suspension platform.

[0012] Preferably, in the stress-sensing-based magnetic suspension micro-motion work platform stabilization method, the magnetic suspension platform posture vector includes at least one posture value which is the output of a position sensor.

[0013] Preferably, the stress-sensing-based magnetic levitation micro-motion working platform stabilization method uses 3 to 5 position sensors and 3 to 5 attitude sensors to determine the current state of the magnetic levitation platform.

[0014] Preferably, in the stress-sensing-based magnetic levitation micro-motion work platform stabilization method, in step (3), the state of the Markov process-based model-free reinforcement learning algorithm is defined as a binary pair of a stress value vector S(τ) and a posture vector P(τ).

[0015] Preferably, in the stress-sensing-based magnetic levitation micro-motion working platform stabilization method, in step (3), the action of the Markov process-based model-free reinforcement learning algorithm is defined as the magnetic levitation micro-motion platform stator winding current compensation value ΔI(τ).

[0016] Preferably, in the stress-sensing-based magnetic suspension micro-motion working platform stabilization method, the reward function r(τ) of the Markov process-based model-free reinforcement learning algorithm in step (3) is defined as the negative value of the sum of the difference between the output value of the posture sensor and the theoretical posture value of the magnetic suspension platform at a preset position, which is expressed as:

[0017]

[0018] in is the theoretical position and posture value corresponding to the jth position or posture sensor.

[0019] Preferably, in the stress-sensing-based magnetic levitation micro-motion work platform stabilization method, the optimization goal of the Markov process-based model-free reinforcement learning algorithm in step (3) is to maximize the reward function value.

[0020] Preferably, in the stress-sensing-based magnetic suspension micro-motion working platform stabilization method, the Markov process-based model-free reinforcement learning algorithm in step (3) adopts a DQN network to maximize decision utility.

[0021] According to another aspect of the present invention, there is provided a magnetic suspension micro-motion work platform stabilization system based on stress sensing, which is an electronic device or a non-transitory computer-readable storage medium;

[0022] The electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing provided by the present invention when executing the program;

[0023] The non-transitory computer-readable storage medium stores a computer program thereon, and is characterized in that when the computer program is executed by a processor, the steps of the stress-sensing-based magnetic suspension micro-motion work platform stabilization method provided by the present invention are implemented.

[0024] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0025] The present invention provides a method and system for stabilizing a magnetic levitation micro-motion working platform based on stress sensing. The method and system sense the stress of the magnetic levitation micro-motion working platform, measure the influence of the combined effects of operating factors and environmental factors on the magnetic levitation micro-motion working platform, determine the compensation current of the stator winding of the magnetic levitation micro-motion working platform according to the stress state of the magnetic levitation micro-motion platform and combine the Markov decision process, thereby adjusting the driving force of the magnetic levitation micro-motion platform. By continuously and iteratively adjusting the driving current of the stator winding, the influence of the operating and environmental factors is offset, and the position stability of the magnetic levitation micro-motion platform is maintained. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a structural schematic diagram of a magnetic suspension micro-motion work platform to which the magnetic suspension micro-motion work platform stabilization method based on stress sensing provided by the present invention is applied;

[0027] Figure 2 It is a schematic diagram of assembling a stress sensor of a magnetic suspension micro-motion working platform provided by an embodiment of the present invention.

[0028] In all the drawings, the same reference numerals are used to represent the same elements or structures, wherein: 1 is a motion platform, 2 is a permanent magnet array, 3 is a stator winding, 4 is a fiber grating array, 5 is a fiber grating, and 6 is a temperature-compensated fiber grating. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0030] The present invention provides a method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing, such as Figure 1 The device is applied to a magnetic suspension work platform, on which a fiber Bragg grating stress sensor is arranged. Preferably, a fiber Bragg grating stress sensor array is evenly arranged, and the direction of the fiber Bragg grating is parallel to the diagonal direction of the magnetic suspension platform to obtain the best measurement sensitivity;

[0031] Fiber Bragg grating sensors are used to measure stress with extremely high sensitivity and are not affected by magnetic fields or electric fields. They have almost no interference with the magnetic suspension platform that is arranged with permanent magnets and driven by electromagnetic coils, and will not cause driving errors. Fiber Bragg grating array sensors are preferably used, which can simultaneously monitor the stress changes of fiber Bragg grating sensors by spectral detection, while reducing the measurement time difference due to the hard gas polling of the sensor.

[0032] In each time slot τ, the following steps are performed:

[0033] (1) Read the stress value vector of the fiber grating array using the spectrum detector is the reading of the i-th fiber Bragg grating stress sensor;

[0034] (2) Obtain the output values ​​of each posture sensor of the magnetic levitation platform as the current posture vector of the magnetic levitation platform Pose value is the output of the jth position or attitude sensor; the position sensor of the magnetic levitation platform is used to determine the position of a specific sampling point of the magnetic levitation platform; the attitude sensor of the magnetic levitation platform is used to determine the deflection state of the magnetic levitation platform. The magnetic levitation platform posture vector includes at least one posture value as the output of the position sensor. If the magnetic levitation platform is regarded as a rigid platform, at least 3 position sensors can be used to determine the current state of the magnetic levitation platform. However, since the magnetic levitation platform is used for high-precision displacement, the use of more position and attitude sensors can better solve the problem of sensor and measurement errors. However, too many position and attitude sensors will introduce more complex control. The preferred solution uses 3 to 5 position sensors and 3 to 5 attitude sensors to determine the current state of the magnetic levitation platform.

[0035] (3) According to the stress value vector S(τ) obtained in step (1) and the posture vector P(τ) obtained in step (2), a model-free reinforcement learning algorithm based on Markov process is used to determine the stator winding current compensation value of the magnetic suspension micro-motion platform in the next time slot τ+1. is the current compensation value of the kth stator winding.

[0036] Markov decision process is called "Markov Decision Process" in English, or MDP for short. It is a mathematical framework used to describe decision-making problems in reinforcement learning, in which the agent takes actions in the environment to maximize a certain cumulative reward. MDP contains the following elements: state set (S), action set (A), state transition probability (P), reward function (R), and discount factor (γ). In the Markov decision process, the future state depends only on the current state and the current action, and has nothing to do with the previous historical state and action. The motion process of the magnetic levitation micro-motion working platform has the Markov property, that is, the position of the magnetic levitation micro-motion platform at the next moment is the same as the position of the magnetic levitation micro-motion platform at the current moment, the current state of the stress, and the current action, that is, the current compensation value.

[0037] Factors that have a complex impact on the posture state of the magnetic levitation micro-motion platform include ambient temperature, pressure, etc., and their impact on the magnetic levitation micro-motion platform is comprehensively reflected in the stress of the magnetic levitation micro-motion platform. The stress sensor array can comprehensively reflect the interference to the magnetic levitation micro-motion platform. The stress value vector is incorporated into the state consideration of the Markov decision process, and the strategy is continuously updated and adjusted according to the reward function to solve the nonlinear problem of stator winding current control, and the position of the micro-motion platform is adaptively stabilized at the preset position.

[0038] The model-free reinforcement learning algorithm based on Markov process adopted in the present invention is as follows:

[0039] The state is defined as a pair of stress value vector S(τ) and posture vector P(τ);

[0040] Action is defined as the stator winding current compensation value ΔI(τ) of the magnetic suspension micro-motion platform;

[0041] The reward function r(τ) is defined as the negative value of the sum of the differences between the output value of the posture sensor and the theoretical posture value of the magnetic levitation platform at a stable preset position, denoted as:

[0042]

[0043] in is the theoretical position and posture value corresponding to the jth position or posture sensor.

[0044] The optimization goal is to maximize the reward function value, and the DQN network is preferably used to maximize the decision utility.

[0045] (4) For the kth stator winding, the current value at the current moment is The sum of the current compensation value at the next moment As the current value of the stator winding at the next moment Right now

[0046] The following are examples:

[0047] The magnetic suspension micro-motion working platform stabilization method based on stress sensing provided in this embodiment uses a magnetic suspension micro-motion working platform structure such as Figure 1 As shown, four groups of permanent magnet arrays are respectively embedded in the middle of the four sides of the bottom surface of the motion platform, and four groups of stator windings are arranged correspondingly; the bottom surface of the motion platform is glued and fixed with 9 fiber grating arrays, as shown in FIG. Figure 2 As shown, the fiber Bragg grating is arranged along the diagonal and in parallel with the diagonal, and the temperature compensation fiber Bragg grating connected in series with the fiber Bragg grating array is suspended, connected to the detection light source through spatial coupling and coupled to the spectrum analyzer; the motion platform position sensor and attitude sensor both use Moire fringe sensors, which are arranged at the four corners and three axes of the platform respectively, with a total of 7 attitude sensors, and the attitude values ​​are observed by imaging equipment.

[0048] The motion platform is levitated and moves under the interaction of the electromagnetic force between the permanent magnet array and the stator winding. By controlling the magnitude and direction of the voltage or current passed into the four sets of stator windings, the micro-motion of the motion platform in six degrees of freedom, namely x, y, z, Hx, Hy, and Hz, can be achieved.

[0049] like Figure 2 The device is applied to a magnetic suspension work platform, on which a fiber Bragg grating stress sensor array is evenly arranged, and the direction of the fiber Bragg grating is parallel to the diagonal direction of the magnetic suspension platform to obtain the best measurement sensitivity;

[0050] Fiber Bragg grating sensors are used to measure stress with extremely high sensitivity and are not affected by magnetic fields or electric fields. They have almost no interference with the magnetic suspension platform that is arranged with permanent magnets and driven by electromagnetic coils, and will not cause driving errors. Using fiber Bragg grating array sensors, the stress changes of fiber Bragg grating sensors can be monitored simultaneously by spectral detection, while reducing the measurement time difference due to the hard gas polling of the sensor.

[0051] The moiré fringe sensor, combined with high-precision imaging, achieves nanometer-level detection sensitivity. It is also compatible with strong magnetic environments, is not affected by magnetic fields, and does not interfere with the drive of the magnetic levitation micro-motion platform.

[0052] In each time slot τ, the following steps are performed:

[0053] (1) Read the stress value vector of the fiber grating array using the spectrum detector is the reading of the i-th fiber Bragg grating stress sensor; n=9;

[0054] (2) Obtain the output values ​​of each posture sensor of the magnetic levitation platform as the current posture vector of the magnetic levitation platform Pose value is the output of the jth position or attitude sensor; the position sensor of the magnetic levitation platform is used to determine the position of a specific sampling point of the magnetic levitation platform; the attitude sensor of the magnetic levitation platform is used to determine the deflection state of the magnetic levitation platform. This embodiment uses 4 position sensors and 3 attitude sensors to determine the current state of the magnetic levitation platform, that is, m=7.

[0055] (3) According to the stress value vector S(τ) obtained in step (1) and the posture vector P(τ) obtained in step (2), a model-free reinforcement learning algorithm based on Markov process is used to determine the stator winding current compensation value of the magnetic suspension micro-motion platform in the next time slot τ+1. is the current compensation value of the kth stator winding, and in this embodiment, l=4.

[0056] The model-free reinforcement learning algorithm based on Markov process adopted in this embodiment is as follows:

[0057] The state is defined as a pair of stress value vector S(τ) and posture vector P(τ), that is, the state is s(τ) = (S(τ), P(τ));

[0058] The action is defined as the stator winding current compensation value ΔI(τ) of the magnetic suspension micro-motion platform, denoted as a(τ) = ΔI(τ+1);

[0059] The reward function r(τ) is defined as the negative value of the sum of the differences between the output value of the posture sensor and the theoretical posture value of the magnetic levitation platform at a stable preset position, denoted as:

[0060]

[0061] in is the theoretical position and posture value corresponding to the jth position or posture sensor.

[0062] The optimization goal is to maximize the reward function value. This embodiment uses the DQN network to maximize the decision utility, as follows:

[0063] Its strategy π(s(τ), a(τ)) is a function of the probability of choosing action a(τ) when the state s(τ) is given. The goal is to maximize the accumulated reward value in the preset time period before time τ, so as to avoid the problem of delayed strategy update due to drastic changes in the production environment due to too long accumulation time.

[0064] Define the action value function Q under the strategy π(s(τ), a(τ)) π (s(τ), a(τ)) is as follows:

[0065]

[0066] Among them, γ∈(0,1] represents the discount factor, and E[·] is the mathematical expectation.

[0067] The optimal action value Q * (s(τ), a(τ)) satisfies the following equation:

[0068] Q * (s(τ),a(τ))=max Q π (s(τ),a(τ))

[0069] In the learning algorithm, the present invention uses Q π (s(τ), a(τ)) Q table to estimate the optimal action value function Q * (s(τ), a(τ)), the update rule of the Q table is as follows:

[0070] Q(s(τ),a(τ))

[0071] =(1-α)Q(s(τ),a(τ))

[0072] +α(r(τ+1)+γmax(s(τ+1),a(τ+1)))

[0073] Among them, α∈(0,1) is the learning rate.

[0074] In time slot τ, the control device obtains the current state s(τ) and uses action a(τ) to determine the current compensation value for the next time slot τ+1 Then we get the reward r(τ+1) from the environment, and then the state space is passed to the next state s(τ+1), and the Q value is updated using (s(τ), a(τ)).

[0075] (4) For the kth stator winding, the current value at the current moment is and the current compensation value at the next moment The sum is taken as the current value of the stator winding at the next moment Right now

[0076] Then enter the next time slot τ+1.

[0077] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing, characterized in that: Applied to a magnetic suspension work platform; the magnetic suspension work platform is provided with a fiber Bragg grating stress sensor; At each time slot τ, the following steps are performed: (1) Read the stress value vector of the fiber grating array using the spectrum detector is the reading of the i-th fiber Bragg grating stress sensor; (2) Obtain the output values ​​of each posture sensor of the magnetic levitation platform as the current posture vector of the magnetic levitation platform Pose value is the output of the jth position or attitude sensor; the position sensor of the magnetic levitation platform is used to determine the position of a specific sampling point of the magnetic levitation platform; the attitude sensor of the magnetic levitation platform is used to determine the deflection state of the magnetic levitation platform; (3) According to the stress value vector S(τ) obtained in step (1) and the posture vector P(τ) obtained in step (2), a model-free reinforcement learning algorithm based on Markov process is used to determine the stator winding current compensation value of the magnetic suspension micro-motion platform in the next time slot τ+1. is the current compensation value of the kth stator winding; (4) For the kth stator winding, the current value at the current moment is The sum of the current compensation value at the next moment As the current value of the stator winding at the next moment Right now 2. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing as claimed in claim 1, characterized in that: The magnetic suspension working platform has an array of fiber grating stress sensors evenly arranged on it, and the direction of the fiber grating is parallel to the diagonal direction of the magnetic suspension platform.

3. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing as claimed in claim 1, characterized in that: The magnetic levitation platform posture vector includes at least one posture value which is the output of the position sensor.

4. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing as claimed in claim 1, characterized in that: 3 to 5 position sensors and 3 to 5 attitude sensors are used to determine the current state of the magnetic levitation platform.

5. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing according to claim 1, characterized in that: The state of the Markov process-based model-free reinforcement learning algorithm in step (3) is defined as a binary pair of a stress value vector S(τ) and a posture vector P(τ).

6. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing according to claim 1, characterized in that: The action of the Markov process-based model-free reinforcement learning algorithm in step (3) is defined as the stator winding current compensation value ΔI(τ) of the magnetic levitation micro-motion platform.

7. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing according to claim 1, characterized in that: The reward function r(τ) of the Markov process-based model-free reinforcement learning algorithm in step (3) is defined as the negative value of the sum of the difference between the output value of the posture sensor and the theoretical posture value of the magnetic levitation platform at a stable preset position, denoted as: in is the theoretical position and posture value corresponding to the jth position or posture sensor.

8. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing according to any one of claims 5 to 7, characterized in that: The optimization goal of the Markov process-based model-free reinforcement learning algorithm in step (3) is to maximize the reward function value.

9. The method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing as claimed in claim 8, characterized in that: The Markov process-based model-free reinforcement learning algorithm described in step (3) uses the DQN network to maximize the decision utility.

10. A magnetic suspension micro-motion work platform stabilization system based on stress sensing, characterized in that: An electronic device or a non-transitory computer-readable storage medium; The electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for stabilizing a magnetic suspension micro-motion working platform based on stress sensing according to any one of claims 1 to 9 when executing the program; The non-transitory computer-readable storage medium stores a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the stress-sensing-based magnetic suspension micro-motion work platform stabilization method as described in any one of claims 1 to 9.

Images