A double-target learning type thrust fluctuation suppression method for a permanent magnet synchronous linear motor

Abstract

The application belongs to the technical field of numerical control, and discloses a double-target learning type thrust fluctuation suppression method for a permanent magnet synchronous linear motor, which comprises the following steps: determining an initial range of iterative learning gain according to a system model and a convergence condition; taking the average of multiple peak values of position tracking error and the standard deviation of speed fluctuation as double targets, and optimizing to obtain an optimal gain matrix; constructing a speed and acceleration feedforward controller to reduce the initial error; collecting the tracking error of the current iteration period, inputting the learning law containing a forgetting factor and a zero-phase filter after data alignment, calculating and storing the compensation signal of the next period; and finally implementing online compensation in the form of current feedforward. The application improves the iterative convergence efficiency through feedforward pre-compensation, suppresses the noise influence through a zero-phase filter, and simultaneously suppresses the position tracking error and speed fluctuation through double-target optimization, thereby significantly improving the dynamic accuracy and running stability of the linear motor feeding system.

Description

Technical Field

[0001] This invention belongs to the field of numerical control technology, and more specifically, relates to a dual-objective learning-type thrust fluctuation suppression method for permanent magnet synchronous linear motors. Background Technology

[0002] Permanent magnet linear motors (PMLSMs) are widely used in high-end manufacturing scenarios with stringent requirements for motion accuracy and stability, such as precision machine tool feed systems, semiconductor packaging equipment, and integrated circuit exposure mechanisms, due to their lack of intermediate transmission links, compact structure, rapid response, high positioning accuracy, and high thrust density.

[0003] However, during actual operation, permanent magnet linear motor feed systems are inevitably affected by multiple factors, including cogging effect, end effect, uneven permanent magnet arrangement, current commutation distortion, and mechanical friction, resulting in thrust fluctuations. These fluctuations reduce tracking accuracy, cause speed fluctuations, and affect the system's control performance, severely limiting their application in ultra-high precision machining. Therefore, suppressing thrust fluctuations is a core challenge in optimizing such systems.

[0004] To address the aforementioned issues, extensive research has been conducted both domestically and internationally, resulting in two main categories of solutions: motor structure optimization and control strategy improvement. Among these, control strategy improvement requires no hardware modifications, offering both flexibility and cost advantages, and is the mainstream approach for suppressing thrust fluctuations in mid-to-high-end feed systems. Iterative Learning Control (ILC) methods are adapted to repetitive motion scenarios and are a well-known and widely used intelligent control strategy in this field.

[0005] The core principle of this technology is as follows: relying on the repetitive operation characteristic of the system's cyclic trajectory reproduction, it collects the output error and control input data from the previous iteration cycle, corrects the current control command through a preset learning law, and reduces the tracking error through multiple iterations, making the output conform to the desired trajectory. It has low dependence on mathematical models and can suppress periodic disturbances without precise modeling.

[0006] Existing thrust fluctuation suppression technology based on ILC, exemplified by the published paper "Thrust Fluctuation Suppression Method for Permanent Magnet Linear Synchronous Motors Based on Iterative Learning Pre-compensation Strategy" (VIP Journal, Document No.: 7201682697), is the closest existing technology to the concept of this invention. Its specific implementation involves: constructing a two-degree-of-freedom position controller containing an acceleration feedforward controller and a disturbance observer based on the position-current dual closed-loop structure of the permanent magnet linear synchronous motor. Acceleration feedforward compensates for acceleration and deceleration dynamic errors, and the disturbance observer initially suppresses thrust fluctuations. A pre-defined ILC controller is introduced, and the position tracking error is reduced through an iterative process. Experiments show that this scheme can reduce the root mean square value of the position tracking error by more than 10% under different speed no-load and light-load conditions, demonstrating certain engineering applicability.

[0007] This type of technology has significant advantages: First, it has a remarkable effect on suppressing periodic fluctuations such as tooth cogging force and end force, which can improve positioning accuracy and operational stability; second, it only optimizes the control algorithm without reconstructing the hardware, making the modification difficult and cost controllable; third, it is robust to parameter perturbations, without the need to accurately obtain the motor's electromagnetic parameters and disturbance model, and has a wide range of applicability.

[0008] However, this type of technology has inherent shortcomings and cannot meet the requirements of ultra-high precision: First, the suppression effect depends on iterative convergence, and the derivation of iterative convergence conditions depends on a precise understanding of the entire system model, which makes it difficult to determine the initial iteration parameters; Second, iterative learning is based on system error learning, and the initial error value of traditional PID feedback control is large, affecting the convergence speed; Third, improvement schemes such as iterative learning and pre-compensation require the introduction of additional compensators and fitting algorithms, increasing computational complexity, increasing controller hardware requirements and system costs; Fourth, the error of the previous iteration cycle is used as the input of the iterative learning controller, which inevitably introduces noise caused by data sampling and system disturbances, affecting the controller performance; Fifth, existing learning laws only consider the suppression of position tracking errors, but speed fluctuations in the actual processing process are also an important factor affecting processing accuracy.

[0009] In summary, existing thrust fluctuation suppression technologies for linear motor feed systems based on ILC still have room for improvement in terms of optimizing iterative learning laws and practical engineering applications. There is an urgent need for a dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors to meet the stringent requirements of high precision and high stability for feed systems in high-end manufacturing. Summary of the Invention

[0010] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors. Its purpose is to improve iterative convergence efficiency, suppress measurement noise, and simultaneously optimize position tracking error and speed fluctuation. This solves the problems of slow iterative convergence, noise issues caused by data sampling and system disturbances, and the technical issues of existing learning laws that only consider position tracking error suppression.

[0011] To achieve the above objectives, according to one aspect of the present invention, a dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors is provided, comprising the following steps: S1: For the target permanent magnet linear synchronous motor feed system, establish its state space equation and determine the initial range of the iterative learning gain matrix according to the convergence condition of iterative learning control. S2: Based on the dynamic equations and state-space equations of the feeding system, establish its mathematical simulation model, and use the iterative learning gain matrix as the optimization parameter. Set the initial value based on the range of the iterative learning gain matrix determined in S1, and construct a dual-objective fitness function with the quantitative indicators of position tracking error and velocity fluctuation. Perform multiple rounds of iterative optimization on the simulation model to obtain the optimal iterative learning gain matrix. S3: Position of the feed system speed In the current three-loop control structure, a velocity feedforward controller and an acceleration feedforward controller are constructed. The velocity feedforward and acceleration feedforward quantities are calculated in real time using the position command signal generated by system interpolation, and then superimposed on the controller output to compensate for the initial tracking error and accelerate the iterative convergence rate. S4: Execute the motion control command for the current iteration cycle based on the feedforward controller described in S3, collect the tracking error data fed back by the servo system, and preprocess the collected data; S5: The tracking error sequence preprocessed in S4 is used as the input of the iterative learning controller. Based on the preset learning law and the optimal iterative learning gain matrix obtained in S2, the compensation control signal sequence for the next iteration cycle is calculated and stored. The learning law introduces a forgetting factor to lock the historical optimal control signal and introduces a zero-phase filter to suppress signal noise and meet the strong timing requirements of iterative learning. S6: In the next iteration cycle, read the compensation control signal sequence stored in S5 and superimpose it onto the input of the servo driver in the form of position feedforward to realize online compensation for thrust fluctuation. S7: Repeat S4 to S6 until the tracking error converges to the preset accuracy or the maximum number of iterations is reached.

[0012] Preferably, the convergence condition in step S1 is as follows:

[0013] in, It is the identity matrix. For iterative learning of the gain matrix, the matrix , For the bounded, well-defined matrix related to the learning gain matrix, Sampling time, denoted as the spectral radius of the matrix.

[0014] Preferably, the quantitative index of the position tracking error in step S2 is the peak-to-average value of the tracking error sequence. The calculation formula is as follows:

[0015] in, The total number of peak values. It is a discrete sampled data sequence. Indicates the sequence number One element, This indicates that this element is the peak value and is the [number]th [element]. One peak.

[0016] Preferably, the quantification index of the velocity fluctuation in step S2 is the standard deviation of the high-frequency component sequence obtained after bandpass filtering the tracking error sequence. The calculation formula is as follows:

[0017] in, This represents the total number of sample points after filtering. Indicates the sampling time. for High-frequency tracking error at all times Its average value.

[0018] Preferably, the bandpass filter is a second-order digital filter. Domain transfer function The calculation formula is as follows:

[0019] in, This represents the unit delay operator, which corresponds to a signal delay of 1 sampling period in the time domain; This indicates a delay of 2 sampling periods; This is the gain coefficient, used to adjust the overall output amplitude of the system; The coefficients of the denominator polynomial are directly related to the damping characteristics of the system poles. The coefficients of the denominator polynomial are equal to the squares of the system pole magnitudes.

[0020] Preferably, the velocity feedforward in step S3 and acceleration feedforward The calculation formulas are as follows:

[0021]

[0022] in, For total load mass, This is the equivalent inertia of the current loop. This is the torque coefficient. and For adjacent control cycles, For the present The speed of instructions at any given moment. For system control cycle; current Instruction speed at any moment The calculation formula is as follows:

[0023] in, They are respectively -1 time and The position command signal is a multi-level continuous signal interpolated by the CNC system at all times.

[0024] Preferably, the data preprocessing in step S4 includes a data alignment operation to correct the data length inconsistency caused by the response delay of the CNC system.

[0025] Preferably, the expression for the learning law in step S5 is:

[0026] in, and For adjacent iteration periods, For the first Control signal for each iteration cycle For the 1st~ The optimal control signal generated in the next iteration, where β is the forgetting factor. This represents the function of a zero-phase filter, used to suppress signal noise without incurring filtering delay. The differential learning gain matrix is... For proportional learning gain matrix, For the first iteration Tracking error at any given moment.

[0027] Preferably, in step S6, the compensation control signal sequence needs to be converted according to the resolution quantization calculation unit inside the driver before feedforward compensation to ensure the accuracy of feedforward control.

[0028] According to another aspect of the present invention, a permanent magnet linear synchronous motor feed system is provided, comprising: Permanent magnet linear synchronous motor; The position detection unit is used to detect the position of the mover in real time and output a position feedback signal; A servo driver, electrically connected to the permanent magnet linear synchronous motor, is used to drive the motor to run according to control commands; The controller is communicatively connected to the host computer, the position detection unit, and the servo driver, and is used to execute the above-mentioned method for pre-compensation and dual-objective learning thrust fluctuation suppression of permanent magnet synchronous linear motor.

[0029] Overall, compared with the prior art, the dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors provided by this invention has the following beneficial effects: 1. Achieving Dual-Objective Optimization: This invention employs a dual-objective optimization algorithm to suppress both position tracking error and velocity fluctuation. It innovatively uses the peak-to-average value of the position tracking error and the standard deviation of the high-frequency components of the velocity fluctuation as the fitness function of the optimization algorithm to optimize the iterative learning gain. This method can simultaneously suppress position error and velocity fluctuation, comprehensively improving the dynamic performance and machining quality of the feeding system.

[0030] 2. Improved convergence efficiency: By establishing explicit iterative convergence conditions and introducing velocity and acceleration feedforward controllers, this invention significantly reduces the initial tracking error of the system and improves iterative convergence efficiency.

[0031] 3. Enhanced robustness: This invention uses a zero-phase filter to filter the output of the learning law, which effectively suppresses noise introduced by data sampling and system disturbances, avoids the phase delay problem caused by traditional filters, and improves the quality of the compensation signal and the stability of the system.

[0032] 4. High engineering practicality: This invention requires no hardware modification and takes into account engineering implementation details such as data alignment and unit conversion, making it easy to integrate and apply on existing CNC systems. Attached Figure Description

[0033] Figure 1 This is a schematic diagram illustrating the principle of the iterative learning law parameter optimization algorithm in an embodiment of the present invention; Figure 2 This is a block diagram of the control method provided in an embodiment of the present invention; Figure 3This is a sequence diagram of the tracking error in the current iteration cycle collected in this embodiment of the invention; Figure 4 This is a sequence diagram of the control signal for the next iteration cycle calculated in an embodiment of the present invention; Figure 5 This is a comparison chart showing the application effects of the embodiments of the present invention on the Kollmorgen linear motor test bench. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0035] The embodiments of this invention are mainly operated on the Kollmorgen linear motor (IC22050A1) test bench. The main parameters of the Kollmorgen permanent magnet synchronous linear motor are shown in Table 1.

[0036]

[0037] This invention provides a dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors, comprising the following steps: S1: For the target permanent magnet linear synchronous motor feed system, establish its state space equation and determine the initial range of the iterative learning gain matrix according to the convergence condition of iterative learning control. Furthermore, the convergence condition is as follows:

[0038] in, It is the identity matrix. For iterative learning of the gain matrix, the matrix , For the bounded, well-defined matrix related to the learning gain matrix, Sampling time, denoted as the spectral radius of the matrix.

[0039] Specifically, for this linear motor feed system, its position and velocity are defined as state variables. Control input , for Based on the actual current at any given time, and combining the electromagnetic thrust equation and the dynamic equation, establish its state-space equation. Substitute it into equation (1) to obtain the norm of the iterative learning gain matrix. ≤0.012.

[0040] S2: Based on the dynamic equations and state-space equations of the feeding system, establish its mathematical simulation model, and use the iterative learning gain matrix as the optimization parameter. Set the initial value based on the range of the iterative learning gain matrix determined in S1, and construct a dual-objective fitness function with the quantitative indicators of position tracking error and velocity fluctuation. Perform multiple rounds of iterative optimization on the simulation model to obtain the optimal iterative learning gain matrix and accelerate the iteration rate. Specifically, for the linear motor feed system, a mathematical simulation model of the system is established based on the dynamic equations. Taking the range of the iterative gain matrix in step (1) as the initial range, the optimal iterative gain matrix is ​​obtained using the optimization algorithm in this invention. = For the optimization algorithm principle, please refer to [link / reference]. Figure 1 .

[0041] Furthermore, the quantification index of the position tracking error in step S2 is the peak-to-average value of the tracking error sequence. The calculation formula is as follows:

[0042] in, The total number of peak values. It is a discrete sampled data sequence. Indicates the sequence number One element, This indicates that this element is the peak value and is the [number]th [element]. One peak.

[0043] The quantification index of the velocity fluctuation mentioned in step S2 is the standard deviation of the high-frequency component sequence obtained after bandpass filtering the tracking error sequence. The calculation formula is as follows:

[0044] in, This represents the total number of sample points after filtering. for High-frequency tracking error at all times Its average value.

[0045] The bandpass filter is a second-order digital filter. Domain transfer function The calculation formula is as follows:

[0046] in, This represents the unit delay operator, which corresponds to a signal delay of 1 sampling period in the time domain; This indicates a delay of 2 sampling periods; This is the gain coefficient, used to adjust the overall output amplitude of the system; The coefficients of the denominator polynomial are directly related to the damping characteristics of the system poles. The coefficients of the denominator polynomial are equal to the squares of the system pole magnitudes.

[0047] S3: Position of the feed system speed In the current three-loop control structure, a velocity feedforward controller and an acceleration feedforward controller are constructed. The velocity feedforward and acceleration feedforward quantities are calculated in real time using the position command signal generated by system interpolation, and then superimposed on the controller output to compensate for the initial error. Specifically, the constructed speed and acceleration feedforward controller requires the current feed rate of the system and related parameters for its output feedforward calculation. After the machine tool starts, the relevant parameters required for the feedforward calculation are first obtained, as shown in Table 2.

[0048] Table 2

[0049] Furthermore, the velocity feedforward in step S3 and acceleration feedforward The calculation formulas are as follows:

[0050]

[0051] in, , , , Obtain from the parameter table. For total load mass, This is the equivalent inertia of the current loop. This is the torque coefficient. and For adjacent control cycles, For the present The speed of instructions at any given moment. For system control cycle; current Instruction speed at any moment The calculation formula is as follows:

[0052] in, They are respectively -1 time and The position command signals are continuously interpolated by the CNC system at various levels.

[0053] In this embodiment, assuming the obtained position command signal sequence is [0, 0.1, 0.2, 0.4, 0.7], its velocity sequence is calculated to be [0.1, 0.1, 0.2, 0.3], and its acceleration feedforward sequence is [0, 0.000697, 0, 0.000697]. For the control framework including the feedforward controller, please refer to [link / reference needed]. Figure 2 .

[0054] S4: Execute the motion control command for the current iteration cycle based on the feedforward controller described in S3, collect the tracking error data fed back by the servo system, and preprocess the collected data; S5: The tracking error sequence preprocessed in S4 is used as the input of the iterative learning controller. Based on the preset learning law and the optimal iterative learning gain matrix obtained in S2, the compensation control signal sequence for the next iteration cycle is calculated and stored. The learning law introduces a forgetting factor to lock the historical optimal control signal and introduces a zero-phase filter to suppress signal noise and meet the strong timing requirements of iterative learning. Furthermore, the expression for the learning law in step S5 is:

[0055] in, and For adjacent iteration periods, For the first Control signal for each iteration cycle For the 1st~ The optimal control signal generated in the next iteration, where β is the forgetting factor. This represents the function of a zero-phase filter, used to suppress signal noise without incurring filtering delay. The differential learning gain matrix is... For proportional learning gain matrix, For the first iteration Tracking error at any given moment.

[0056] Specifically, in this embodiment, the machine tool is started to execute a specific instruction, and the tracking error signal sequence within the current iteration cycle of the feed system is collected (see [link to relevant documentation]). Figure 3 Using this sequence as input to the iterative learning controller, the control signal sequence for the next iteration is output after passing through the learning law. Please refer to [link / reference]. Figure 4 .

[0057] S6: In the next iteration cycle, read the compensation control signal sequence stored in S5 and superimpose it onto the input of the servo driver in the form of position feedforward to realize online compensation for thrust fluctuation. Before performing feedforward compensation, the compensation control signal sequence needs to be converted according to the resolution quantization unit inside the driver to ensure the accuracy of feedforward control.

[0058] Specifically, in this embodiment, the feedforward compensation switch on the CNC system side is turned on, and the attached data is read... Figure 4 Compensation is performed in the form of feedforward after the control signal sequence.

[0059] S7: Repeat steps S4 to S6 until the tracking error converges to the preset accuracy or the maximum number of iterations is reached. This process effectively reduces the position tracking error and velocity fluctuations throughout the entire motion cycle. For a detailed explanation of the technical effects of this embodiment, please refer to [link to relevant documentation]. Figure 5 The actual effect was verified on the Kollmorgen linear motor workbench. Taking a uniform speed signal of 0.083 m / s as an example, the maximum tracking error of the uniform speed segment was reduced from 3.5 μm to less than 0.5 μm, and the average speed fluctuation of the uniform speed segment was reduced from 20 mm / min to less than 12 mm / min. Based on the suppression effect on position tracking error and speed fluctuation, the suppression effect reached 80% and 40% respectively, which is significant.

[0060] Those skilled in the art will readily understand that the above description is merely 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 within the scope of protection of the present invention.

Claims

1. A dual-objective learning-based thrust fluctuation suppression method for permanent magnet synchronous linear motors, characterized in that: Includes the following steps: S1: For the target permanent magnet linear synchronous motor feed system, establish its state space equation and determine the initial range of the iterative learning gain matrix according to the convergence condition of iterative learning control. S2: Based on the dynamic equations and state-space equations of the feeding system, establish its mathematical simulation model, and use the iterative learning gain matrix as the optimization parameter. Set the initial value based on the range of the iterative learning gain matrix determined in S1, and construct a dual-objective fitness function with the quantitative indicators of position tracking error and velocity fluctuation. Perform multiple rounds of iterative optimization on the simulation model to obtain the optimal iterative learning gain matrix. S3: In the position-velocity-current three-loop control structure of the feed system, a velocity feedforward controller and an acceleration feedforward controller are constructed. The velocity feedforward and acceleration feedforward quantities are calculated in real time using the position command signal generated by system interpolation, and then superimposed on the controller output. S4: Execute the motion control command for the current iteration cycle based on the feedforward controller described in S3, collect the tracking error data fed back by the servo system, and preprocess the collected data; S5: The tracking error sequence preprocessed in S4 is used as the input of the iterative learning controller. Based on the preset learning law and the optimal iterative learning gain matrix obtained in S2, the compensation control signal sequence for the next iteration cycle is calculated and stored. The learning law introduces a forgetting factor to lock the historical optimal control signal and introduces a zero-phase filter to suppress signal noise and meet the strong timing requirements of iterative learning. S6: In the next iteration cycle, read the compensation control signal sequence stored in S5 and superimpose it onto the input of the servo driver in the form of position feedforward to realize online compensation for thrust fluctuation. S7: Repeat S4 to S6 until the tracking error converges to the preset accuracy or the maximum number of iterations is reached.

2. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: The convergence condition described in step S1 is as follows: in, It is the identity matrix. For iterative learning of the gain matrix, the matrix , For the bounded, well-defined matrix related to the learning gain matrix, Sampling time, denoted as the spectral radius of the matrix.

3. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: The quantitative index of the position tracking error in step S2 is the peak-to-average value of the tracking error sequence. The calculation formula is as follows: in, The total number of peak values. It is a discrete sampled data sequence. Indicates the sequence number One element, This indicates that this element is the peak value and is the [number]th [element]. One peak.

4. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: The quantification index of the velocity fluctuation mentioned in step S2 is the standard deviation of the high-frequency component sequence obtained after bandpass filtering the tracking error sequence. The calculation formula is as follows: in, This represents the total number of sample points after filtering. Indicates the sampling time. for High-frequency tracking error at all times Its average value.

5. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 4, is characterized in that: The bandpass filter is a second-order digital filter. Domain transfer function The calculation formula is as follows: in, This represents the unit delay operator, which corresponds to a signal delay of 1 sampling period in the time domain; This indicates a delay of 2 sampling periods; This is the gain coefficient, used to adjust the overall output amplitude of the system; The coefficients of the denominator polynomial are directly related to the damping characteristics of the system poles. The coefficients of the denominator polynomial are equal to the squares of the system pole magnitudes.

6. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: Velocity feedforward in step S3 and acceleration feedforward The calculation formulas are as follows: in, For total load mass, This is the equivalent inertia of the current loop. This is the torque coefficient. and For adjacent control cycles, For the present The speed of instructions at any given moment. For system control cycle; current Instruction speed at any moment The calculation formula is as follows: in, They are respectively -1 time and The position command signal is a multi-level continuous signal interpolated by the CNC system at all times.

7. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: Step S4, data preprocessing, includes data alignment operations to correct data length inconsistencies caused by CNC system response delays.

8. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: The expression for the learning law mentioned in step S5 is: in, and For adjacent iteration periods, For the first Control signal for each iteration cycle For the 1st~ The optimal control signal generated in the next iteration, where β is the forgetting factor. This represents the function of a zero-phase filter, used to suppress signal noise without incurring filtering delay. The differential learning gain matrix is... For proportional learning gain matrix, For the first iteration Tracking error at any given moment.

9. The method for suppressing thrust fluctuations in a permanent magnet synchronous linear motor with dual objectives of learning, as described in claim 1, is characterized in that: In step S6, the compensation control signal sequence needs to be converted according to the resolution quantization unit inside the driver before feedforward compensation to ensure the accuracy of feedforward control.

10. A permanent magnet linear synchronous motor feed system, characterized in that, include: Permanent magnet linear synchronous motor; A position detection unit is used to detect the position of the mover in real time and output a position feedback signal; a servo driver is electrically connected to the permanent magnet linear synchronous motor and is used to drive the motor to run according to control commands; a controller is communicatively connected to the host computer, the position detection unit and the servo driver, and is used to execute the dual-objective learning thrust fluctuation suppression method for permanent magnet synchronous linear motors as described in any one of claims 1 to 9.

Images