A magnetic levitation nozzle calibration system and method

Abstract

This invention relates to the printing field and discloses a magnetic levitation printhead calibration system and method. The magnetic levitation printhead calibration system includes a mover backplate, a mover assembly, a stator coil assembly, a stator backplate, a sensor module, and a controller. The mover backplate is used to mount the printhead. The mover assembly includes a permanent magnet array mounted on the back of the mover backplate. The stator coil assembly includes multiple electromagnetic coils mounted on the front of the stator backplate. A magnetic levitation gap is provided between the permanent magnet array and the electromagnetic coils. The six-degree-of-freedom magnetic levitation of the printhead is achieved through the interaction of the magnetic fields of the electromagnetic coils and the permanent magnet array. The sensor module is used to measure the printhead's pose in real time. The controller is connected to the sensor module and the multiple electromagnetic coils. The controller receives the current printhead pose from the sensor module and controls the current of each electromagnetic coil based on the current printhead pose to adjust the real-time pose of the printhead, enabling online high-precision calibration of the printhead.

Description

Technical Field

[0001] This invention relates to the field of printing, and more particularly to a magnetic levitation printhead calibration system and method. Background Technology

[0002] New display manufacturing, especially next-generation display technologies such as OLED, QLED, and Micro-LED, is increasingly adopting high-precision inkjet printing (IJP) technology as its core fabrication process. This technology precisely deposits functional materials (such as luminescent materials, electrode materials, and optical thin film pastes) onto a substrate in droplet form, enabling the patterned fabrication of pixels, circuits, and functional layers. It offers significant advantages such as high material utilization, a wide range of applicable substrates (including flexible curved surfaces), simplified processes, and ease of large-scale production.

[0003] However, inkjet printing technology faces extremely demanding challenges in the large-scale, high-density, and high-precision manufacturing of display panels. Among these challenges, the precise positioning and attitude control of the printhead are the core factors that determine product yield and performance.

[0004] Extremely high pixelation precision is required: when fabricating RGB pixel arrays, the size of each subpixel is only on the order of micrometers to tens of micrometers. The landing position error of inkjet droplets must be strictly controlled within ±5µm or even smaller. Any tiny deviation may lead to fatal defects such as color mixing, uneven brightness, or short circuits, resulting in high material and production cost losses.

[0005] The substrate surface has a complex topology: the substrate of the display panel is not an ideal plane. The nozzle must be able to dynamically adjust its height (Z-axis) and attitude (pitch, yaw angle) to ensure that the nozzle and the substrate surface maintain the optimal distance (usually several hundred micrometers) throughout the entire printing stroke. Otherwise, it will cause the droplet flight curve to deviate, the impact position to be incorrect, or the droplet shape to become uncontrollable.

[0006] Challenges posed by large-size substrates: G8.5 and later generation production lines need to handle large glass substrates exceeding 2.2m x 2.5m. The printing system must possess extremely high motion smoothness and global consistency; any vibration, thermal deformation, or guide rail error within the span will directly translate into printing defects.

[0007] Capacity and throughput pressures: Display manufacturing is a capital-intensive industry with extremely high requirements for equipment capacity (UPH, Units Per Hour). The printhead calibration process must be fast, automated, and ideally completed online in real time to minimize production downtime. Traditional calibration methods that rely on manual intervention have become a bottleneck for improving production efficiency.

[0008] Therefore, existing technologies still need improvement and development. Summary of the Invention

[0009] The first objective of this invention is to provide a magnetic levitation printhead calibration system, which aims to solve the technical problems of poor printhead calibration accuracy and inconvenience of online calibration in existing inkjet printing equipment.

[0010] To achieve the above objectives, the solution provided by the present invention is as follows:

[0011] A magnetic levitation nozzle calibration system includes a mover backplate, a mover assembly, a stator coil assembly, a stator backplate, a sensor module, and a controller. The mover backplate is used to mount the nozzle. The mover assembly includes a permanent magnet array mounted on the back of the mover backplate. The stator coil assembly includes multiple electromagnetic coils mounted on the front of the stator backplate. A magnetic levitation gap is provided between the permanent magnet array and the electromagnetic coils. The sensor module is mounted on the mover backplate and located beside the nozzle. The sensor module is used to measure the nozzle's pose in real time. The controller is connected to the sensor module and the multiple electromagnetic coils. The controller receives the current nozzle pose fed back by the sensor module and controls the current of each electromagnetic coil based on the current nozzle pose to adjust the real-time pose of the nozzle.

[0012] Preferably, the sensor module includes a displacement sensor, a gap sensor, and an inertial measurement unit. The displacement sensor, the gap sensor, and the inertial measurement unit are all disposed on the back plate of the mover, and are respectively disposed on the periphery of the printhead. The displacement sensor is used to measure the displacement of the printhead in at least one translational direction. The gap sensor is used to measure the distance between the printhead and the substrate to be printed. The inertial measurement unit is used to detect the micro-vibration and angular velocity of the printhead. The controller is connected to the displacement sensor, the gap sensor, and the inertial measurement unit respectively.

[0013] Preferably, the gap sensor includes a capacitive gap sensor body, a first sensor mounting bracket, and a second sensor mounting bracket. The first sensor mounting bracket is mounted on the side of the stator, the capacitive gap sensor body is mounted on the first sensor mounting bracket, and the second sensor mounting bracket is connected to the first sensor mounting bracket and fixes the capacitive gap sensor body on the first sensor mounting bracket.

[0014] Preferably, the mover assembly further includes an upper fixing plate for the mover magnet, a middle fixing plate for the mover magnet, and a lower fixing plate for the mover magnet. The permanent magnet array is mounted on the middle fixing plate for the mover magnet. The upper fixing plate for the mover magnet and the lower fixing plate for the mover magnet are respectively fixed on both sides of the middle fixing plate for the mover magnet. The upper fixing plate for the mover magnet is mounted on the back side of the mover back plate.

[0015] Preferably, the stator coil assembly further includes a stator coil mounting component and a stator coil cover plate. The stator coil mounting component is mounted on the front side of the stator back plate, the electromagnetic coils are mounted on the stator coil mounting component, and the stator coil cover plate covers a plurality of the electromagnetic coils and is connected to the stator coil mounting component.

[0016] The second objective of this invention is to provide a magnetic levitation nozzle calibration method. This method is based on the magnetic levitation nozzle calibration system described above. The method includes: constructing an initial state-space model as a prediction model based on the structure and dynamic characteristics of the magnetic levitation nozzle calibration system, and performing offline pre-training on a Gaussian process regression model to obtain the initial hyperparameters of the Gaussian process regression model; acquiring and fusing displacement, angle, and acceleration information related to the nozzle from the sensor module feedback to obtain the current nozzle pose; predicting the current nozzle pose using the prediction model based on the actual nozzle pose and the electromagnetic coil current vector from the previous moment, and calculating the residual between the current nozzle pose and the predicted value; acquiring the electromagnetic coil current vector at the current moment, and combining the current state-space model with the actual nozzle pose and the electromagnetic coil current vector from the previous moment... The electromagnetic coil current vector at a given moment, along with the residuals of the current nozzle pose and the predicted current nozzle pose, form a new input-output sample pair. This new input-output sample pair is used to update the initial hyperparameters of the Gaussian process regression model. Based on the current nozzle pose and the control variable to be determined, the updated Gaussian process regression model outputs the residual prediction mean and prediction variance. The residual prediction mean is fused with the prediction model to obtain a modified prediction model. Using the modified prediction model as the internal prediction model, with the actual nozzle pose as the initial state and the target pose as the tracking target, a cost function is defined. The constraint boundary is adjusted based on the prediction variance to solve for the optimal control input sequence that satisfies the constraint boundary within a finite future time domain. The instantaneous control variable in the optimal control sequence is converted into a current signal, which drives multiple electromagnetic coils of the stator coil assembly to adjust the real-time pose of the nozzle.

[0017] Preferably, the prediction model is expressed as:

[0018]

[0019] In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, The state matrix, This is the control matrix.

[0020] Preferably, the initial state-space model is constructed as a prediction model based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, and the Gaussian process regression model is pre-trained offline to obtain the initial hyperparameters of the Gaussian process regression model. This includes: constructing an initial state-space model as a prediction model based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system; acquiring historical operating data or simulation data as initial training data, wherein the initial training dataset includes multiple initial input and output sample pairs, each of which includes an electromagnetic coil current vector, a target pose, and an actual nozzle pose; calculating the predicted nozzle pose value for each initial input and output sample pair using the prediction model based on the initial training data, and calculating the residual between the predicted nozzle pose value and the actual nozzle pose; pre-training the Gaussian process regression model offline with the electromagnetic coil current vector and target pose of the initial input and output sample pairs as independent variables and the corresponding residual as dependent variables, and optimizing and determining the initial hyperparameters of the Gaussian process regression model by minimizing the residual prediction error, wherein the initial hyperparameters include kernel function parameters and noise variance.

[0021] Preferably, the modified prediction model is expressed as:

[0022]

[0023] In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, The state matrix, For the control matrix, This represents the mean of the predicted residuals.

[0024] Preferably, the cost function is defined as follows: Then the cost function Represented as:

[0025]

[0026] In the formula, Indicates the target pose. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, This indicates the length of the prediction time domain, and corrects the number of time steps the prediction model advances in each optimization. This indicates the length of the control time domain and corrects the actual number of control steps optimized by the prediction model. , This is a summation index variable used to iterate through discrete time steps.

[0027] In this solution, the six-degree-of-freedom contactless magnetic levitation drive of the printhead is achieved through the interaction of the magnetic fields of the permanent magnet array and the electromagnetic coils, completely avoiding the friction and wear problems of traditional mechanical transmission. Combined with the sensor module to accurately capture the printhead posture in real time, and the controller to dynamically adjust the current of each electromagnetic coil based on the measured posture, it can not only achieve micron-level high-precision positioning and posture calibration to ensure the stability of printing quality, but also support online real-time calibration during the printing process without interrupting the production process, which greatly shortens the calibration time, increases equipment capacity, simplifies the manual intervention process, and reduces maintenance costs, thus meeting the stringent requirements of high precision, high efficiency and high stability in high-end display manufacturing. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the structure of the magnetic levitation nozzle calibration system provided in this embodiment of the invention. Figure 1 ;

[0030] Figure 2 This is a schematic diagram of the structure of the magnetic levitation nozzle calibration system provided in this embodiment of the invention. Figure 2 ;

[0031] Figure 3 This is a side view of the magnetic levitation nozzle calibration system provided in an embodiment of the present invention;

[0032] Figure 4 yes Figure 3 Section along the AA direction;

[0033] Figure 5 This is a flowchart of the magnetic levitation nozzle calibration method provided in the embodiments of the present invention.

[0034] Explanation of icon numbers:

[0035] 10. Mover backplate; 20. Mover assembly; 21. Permanent magnet array; 22. Upper fixing plate of mover magnet; 23. Middle fixing plate of mover magnet; 24. Lower fixing plate of mover magnet; 30. Stator coil assembly; 31. Electromagnetic coil; 32. Stator coil mounting component; 33. Stator coil cover plate; 40. Stator backplate; 50. Sensor module; 51. Displacement sensor; 52. Gap sensor; 521. Capacitive gap sensor body; 522. First sensor mounting bracket; 523. Second sensor mounting bracket; 53. Inertial measurement unit; 60. Magnetic levitation gap; 70. Nozzle; 80. Printing substrate. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0037] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0038] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0039] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0040] like Figures 1 to 4 As shown, this is a magnetic levitation nozzle calibration system according to an embodiment of the present invention.

[0041] Please see Figures 1-4As shown, the magnetic levitation nozzle calibration system of this embodiment includes a mover backplate 10, a mover assembly 20, a stator coil assembly 30, a stator backplate 40, a sensor module 50, and a controller. The mover backplate 10 is used to mount the nozzle 70. The mover assembly 20 includes a permanent magnet array 21, which is mounted on the back of the mover backplate 10. The stator coil assembly 30 includes multiple electromagnetic coils 31, which are mounted on the front of the stator backplate 40. Magnetic levitation is provided between the permanent magnet array 21 and the electromagnetic coils 31. The nozzle 70 is magnetically levitated with six degrees of freedom through the interaction of the magnetic fields of the electromagnetic coil 31 and the permanent magnet array 21. The sensor module 50 is set on the back plate 10 of the mover and located next to the nozzle 70. The sensor module 50 is used to measure the position and orientation of the nozzle 70 in real time. The controller is connected to the sensor module 50 and multiple electromagnetic coils 31 respectively. The controller receives the current position and orientation of the nozzle from the sensor module 50 and controls the current of each electromagnetic coil 31 based on the current position and orientation of the nozzle to adjust the real-time position and orientation of the nozzle 70.

[0042] Understandably, the permanent magnet array 21 is arranged according to a preset Halbach array or other optimized polarity to enhance the unilateral magnetic field and reduce coupling. The electromagnetic coils 31 are distributed in a matrix in space, and the axial direction of each electromagnetic coil 31 is designed so that the magnetic field it generates can effectively interact with the magnetic field of the permanent magnet. By independently and precisely controlling the magnitude and direction of the current in each electromagnetic coil 31, forces and torques of arbitrary direction and magnitude can be synthesized on the mover backplate 10 (carrying nozzle 70), thereby achieving contactless drive and levitation with complete decoupling of six degrees of freedom of translation along the X, Y, and Z axes and rotation around these three axes.

[0043] In this embodiment, the magnetic levitation gap 60 is typically designed to be between 0.1 mm and 1.0 mm. This optimized gap design ensures sufficient control force and stiffness while minimizing magnetic circuit nonlinearity and coupling effects, thus providing a guarantee for achieving precise control with high linearity and high decoupling.

[0044] In this embodiment, the controller adopts an FPGA+DSP heterogeneous architecture. The FPGA is responsible for high-speed synchronous acquisition of multi-sensor data (acquisition delay ≤ 0.1ms) and PWM current drive of electromagnetic coil 31; the DSP is responsible for running an adaptive algorithm that fuses model predictive control (MPC) and Gaussian process (GP) to ensure real-time generation of control commands.

[0045] In other embodiments, the controller can be any processor or combination of processors capable of performing the required high-speed data acquisition and real-time model predictive control algorithm calculations, such as an industrial PC, a multi-core microprocessor, etc.

[0046] In this embodiment, the nozzle 70 is installed on the mover backplate 10. Then, the current nozzle pose is measured by the sensor module 50, and the measured current nozzle pose is fed back to the controller. The controller compares the current nozzle pose with the target pose and determines whether the deviation value between the current nozzle pose and the target pose is within the deviation threshold. If it is, it indicates that the nozzle 70 is installed in place; if not, it indicates that the nozzle 70 is not installed in place. Based on the deviation value between the current nozzle pose and the target pose, a control command is generated, and the current of each electromagnetic coil 31 is controlled according to the control command, thereby adjusting the real-time pose of the nozzle 70 until the current nozzle pose and the target pose are within the deviation threshold.

[0047] In this implementation, the six-degree-of-freedom non-contact magnetic levitation drive of the printhead 70 is achieved through the magnetic field interaction between the permanent magnet array 21 and the electromagnetic coil 31, completely avoiding the friction and wear problems of traditional mechanical transmission. Combined with the sensor module to accurately capture the printhead posture in real time, and the controller to dynamically adjust the current of each electromagnetic coil 31 based on the measured posture, it can not only achieve micron-level high-precision positioning and posture calibration to ensure the stability of printing quality, but also support online real-time calibration during the printing process without interrupting the production process, greatly shortening the calibration time, increasing equipment capacity, simplifying the manual intervention process, reducing maintenance costs, and adapting to the stringent requirements of high precision, high efficiency, and high stability in high-end display manufacturing.

[0048] In this embodiment, the sensor module 50 includes a displacement sensor 51, a gap sensor 52, and an inertial measurement unit 53. All three sensors are mounted on the mover backplate 10 and are located around the nozzle 70. The displacement sensor 51 measures the displacement of the nozzle 70 in at least one translational direction. The gap sensor 52 measures the distance between the nozzle 70 and the substrate 80 to be printed. The inertial measurement unit 53 detects the micro-vibrations and angular velocities of the nozzle 70. A controller is connected to the displacement sensor 51, the gap sensor 52, and the inertial measurement unit 53. This multi-sensor fusion configuration forms an omnidirectional, high-bandwidth sensing network. The displacement sensor 51 provides a global positioning reference, ensuring absolute accuracy in the XY plane. The gap sensor 52 directly monitors key printing parameters (distance between the nozzle 70 and the substrate), overcoming the influence of substrate morphology. The inertial measurement unit 53 captures high-frequency micro-vibrations, providing information for feedforward compensation. The three components work together to provide comprehensive and accurate system status feedback.

[0049] In this embodiment, the displacement sensor 51 is a laser interferometer or a high-resolution grating ruler. A laser interferometer or an ultra-high-resolution grating ruler can provide sub-nanometer resolution, meeting the extreme accuracy requirements for nozzle 70 positioning in display manufacturing.

[0050] In this embodiment, the displacement sensor 51 includes two interferometer mirror groups, one of which is arranged laterally on the stator back plate 40, and the other is arranged longitudinally on the stator back plate 40. This orthogonal arrangement can directly and independently measure the displacement of the nozzle 70 in the X and Y directions, avoiding calculation errors introduced by coordinate transformation, and improving the directness and accuracy of planar positioning measurement.

[0051] In this embodiment, the gap sensor 52 is a capacitive or eddy current non-contact gap sensor 52. The capacitive or eddy current gap sensor 52 has the characteristics of non-contact, high resolution, and fast response, and can accurately measure micron-level changes in working gap without interfering with the printing process.

[0052] In this embodiment, the gap sensor 52 includes a capacitive gap sensor body 521, a first sensor mounting bracket 522, and a second sensor mounting bracket 523. The first sensor mounting bracket 522 is mounted on the side of the stator, the capacitive gap sensor body 521 is mounted on the first sensor mounting bracket 522, and the second sensor mounting bracket 523 is connected to the first sensor mounting bracket 522 and fixes the capacitive gap sensor body 521 on the first sensor mounting bracket 522.

[0053] In this embodiment, the mover assembly 20 further includes an upper moving magnet fixing plate 22, a middle moving magnet fixing plate 23, and a lower moving magnet fixing plate 24. The permanent magnet array 21 is mounted on the middle moving magnet fixing plate 23. The upper moving magnet fixing plate 22 and the lower moving magnet fixing plate 24 are respectively fixed on both sides of the middle moving magnet fixing plate 23. The upper moving magnet fixing plate 22 is mounted on the back of the mover back plate 10. The multi-layer fixing plate structure provides robust mechanical protection and precise positioning for the permanent magnet array 21, preventing damage or displacement during high-speed movement or impact, ensuring the stability and consistency of magnetic force output, and facilitating the assembly and maintenance of the entire mover assembly 20.

[0054] In this embodiment, the stator coil assembly 30 further includes a stator coil mounting component 32 and a stator coil cover plate 33. The stator coil mounting component 32 is mounted on the front side of the stator back plate 40, and the electromagnetic coils 31 are mounted on the stator coil mounting component 32. The stator coil cover plate 33 covers multiple electromagnetic coils 31 and is connected to the stator coil mounting component 32. The stator coil mounting component 32 provides precise positioning and heat dissipation channels for the electromagnetic coils 31, while the stator coil cover plate 33 protects the coils, guides cooling airflow, and prevents foreign objects such as metal debris from entering the magnetic gap, thereby improving the reliability, heat dissipation performance, and cleanliness of the entire actuator, meeting the environmental requirements of high-end manufacturing equipment.

[0055] Please see Figure 5As shown, this embodiment of the invention also provides a nozzle calibration method, including:

[0056] S101. Based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, an initial state-space model is constructed as a prediction model, and the Gaussian process regression model is pre-trained offline to obtain the initial hyperparameters of the Gaussian process regression model.

[0057] S102. Obtain and fuse the displacement, angle and acceleration information related to the nozzle 70 fed back by the sensor module 50 to obtain the current nozzle pose.

[0058] S103. Based on the actual nozzle pose and the electromagnetic coil current vector of the previous moment, the current nozzle pose prediction value is obtained by using the prediction model, and the residual between the current nozzle pose and the current nozzle pose prediction value is calculated.

[0059] S104. Obtain the electromagnetic coil current vector at the current moment, and combine the actual nozzle pose at the previous moment, the electromagnetic coil current vector at the previous moment, and the residual between the current nozzle pose and the predicted current nozzle pose to form a new input and output sample pair, and use the new input and output sample pair to update the initial hyperparameters of the Gaussian process regression model.

[0060] S105. Based on the current nozzle pose and the control variable to be determined, the updated Gaussian process regression model is used to output the residual prediction mean and prediction variance.

[0061] S106. The residual prediction mean is fused with the prediction model to obtain the modified prediction model;

[0062] S107. Using the modified prediction model as the internal prediction model, the actual pose of the nozzle as the initial state, and the target pose as the tracking target, define the cost function, and adjust the constraint boundary based on the prediction variance to solve the optimal control input sequence that satisfies the constraint boundary in the future finite time domain.

[0063] S108. The instantaneous control quantity in the optimal control sequence is converted into a current signal, and the current signal drives multiple electromagnetic coils 31 of the stator coil assembly 30 to adjust the real-time position and orientation of the nozzle 70.

[0064] In this embodiment, an initial state-space prediction model is first constructed and a Gaussian process regression model is pre-trained offline to lay the foundation for precise control. Then, multi-sensor data is fused in real time to obtain the true pose of the printhead 70. By using residual calculation and online model updates to dynamically adapt to the nonlinear characteristics of the system and external disturbances, a high-precision corrected prediction model is obtained by fusing the mean value of residual prediction. Finally, based on the corrected model, the optimal control sequence is solved by combining the cost function and dynamic constraint boundary and driving the electromagnetic coil 31. This not only achieves high-precision calibration of the printhead pose and ensures the stability of printing quality, but also enables real-time calibration during the printing process without interrupting production, significantly shortening calibration time and improving production efficiency. At the same time, it has strong adaptability and robustness, which can effectively cope with complex working conditions such as substrate morphology changes and environmental vibrations, reduce manual intervention and maintenance costs, and meet the stringent requirements of high-end display manufacturing.

[0065] In this embodiment, in step S101, the prediction model is represented as:

[0066]

[0067] In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector is composed of the driving current of each electromagnetic coil 31 in the stator coil assembly 30. The state matrix, This is the control matrix.

[0068] In this embodiment, the actual position of the nozzle includes along... Translation in direction and rotation angles around the X, Y, and Z axes ,like .

[0069] In the formula, T represents the transpose operation, which converts the original row vector into a column vector.

[0070] In this embodiment, .

[0071] In the formula, T represents the transpose operation, which converts the original row vector into a column vector. These represent the electromagnetic coil currents that control the translation of the nozzle along the X, Y, and Z axes, respectively. These represent the electromagnetic coil currents that control the nozzle's rotation around the X, Y, and Z axes, respectively.

[0072] In this embodiment, based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, an initial state-space model is constructed as a prediction model, and an offline pre-training of the Gaussian process regression model is performed to obtain the initial hyperparameters of the Gaussian process regression model. This includes: constructing an initial state-space model as a prediction model based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system; acquiring historical operating data or simulation data as initial training data, the initial training dataset including multiple initial input and output sample pairs, each pair including an electromagnetic coil current vector, target pose, and actual nozzle pose; calculating the predicted nozzle pose value for each initial input and output sample pair using the prediction model based on the initial training data, and calculating the residual between the predicted nozzle pose value and the actual nozzle pose; using the electromagnetic coil current vector and target pose of the initial input and output sample pairs as independent variables and the corresponding residual as dependent variables, offline pre-training of the Gaussian process regression model is performed, and the initial hyperparameters of the Gaussian process regression model are determined by minimizing the residual prediction error. The initial hyperparameters of the Gaussian process regression model include kernel function parameters and noise variance.

[0073] In this embodiment, the initial input and output sample pair is represented as , Indicates the first At any given moment, the current vector of the electromagnetic coil, Indicates the first Target pose at all times. Indicates the first The actual position of the nozzle at all times.

[0074] The residual between the predicted nozzle pose and the actual nozzle pose Represented as:

[0075]

[0076] In the formula, Indicates the first Predicted nozzle position at any given time.

[0077] In this embodiment, the initial training dataset covers the typical operating conditions of the magnetic levitation nozzle calibration system, including historical operating data or simulation data under different substrate sizes (such as G8.5 generation substrate) and different printing speeds (50-200mm / s). Each sample contains electromagnetic coil current vector (6-channel coil current, range 0-5A), target pose (accuracy ±0.1μm), and actual nozzle pose (measured data after sensor fusion). The dataset has no less than 10,000 samples.

[0078] In this embodiment, in step S102, an extended Kalman filter or an unscented Kalman filter algorithm is used to fuse the displacement information, angle information and acceleration information related to the nozzle 70 to obtain the current nozzle pose.

[0079] In this embodiment, in step S104, when updating the Gaussian process regression model using newly added input and output sample pairs, a sparse Gaussian process approximation or a fixed-size sliding time window strategy is adopted. That is, only the most representative sample pairs in the most recent period are retained, or an induced point set is used to approximate the entire dataset, thereby significantly reducing the computational load while ensuring learning ability and meeting the time limit requirements of the real-time control loop.

[0080] In this embodiment, in step S106, the modified prediction model is expressed as:

[0081]

[0082] In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, The state matrix, For the control matrix, This represents the mean of the predicted residuals.

[0083] In this embodiment, in step S107, the cost function Represented as:

[0084]

[0085] In the formula, Indicates the target pose. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, This indicates the length of the prediction time domain, and corrects the number of time steps the prediction model advances in each optimization. This indicates the length of the control time domain and corrects the actual number of control steps optimized by the prediction model. , This is a summation index variable used to iterate through discrete time steps.

[0086] In this embodiment, adjusting the constraint boundary based on the prediction variance means mapping the prediction variance of the Gaussian process output to a constraint safety margin adjustment factor for control or state variables. For example, when the prediction variance is large (high model uncertainty), the current amplitude constraint is tightened or the pose tracking error constraint is relaxed to avoid instability caused by aggressive control when the model is unreliable; conversely, when the prediction variance is small, the current constraint is relaxed or the pose constraint is tightened to pursue better dynamic performance.

[0087] In this embodiment, the constraints include: electromagnetic coil 31 current amplitude constraint (single coil current ≤ 5A), nozzle pose range constraint (X / Y axis translation range ±100μm, Z axis translation range 50-500μm, rotation angle range around the three axes ±0.5°); the safety margin of the constraint boundary is dynamically adjusted based on the prediction variance output by the Gaussian process regression model, when the prediction variance > 1e -8 When the safety margin is increased by 20%, and the prediction variance is less than 1e... -10 At that time, the safety margin is reduced by 10%.

[0088] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A magnetic levitation nozzle calibration system, characterized in that, The system includes a mover backplate, a mover assembly, a stator coil assembly, a stator backplate, a sensor module, and a controller. The mover backplate is used to mount the nozzle. The mover assembly includes a permanent magnet array mounted on the back of the mover backplate. The stator coil assembly includes multiple electromagnetic coils mounted on the front of the stator backplate. A magnetic levitation gap is provided between the permanent magnet array and the electromagnetic coils. The sensor module is mounted on the mover backplate and located beside the nozzle. The sensor module is used to measure the nozzle's position and orientation in real time. The controller is connected to both the sensor module and the multiple electromagnetic coils. The controller receives the current nozzle position and orientation feedback from the sensor module and, based on the current position and orientation, determines the appropriate position and orientation for the nozzle. The current of each electromagnetic coil is controlled by the head pose control to adjust the real-time pose of the printhead. The sensor module includes a displacement sensor, a gap sensor, and an inertial measurement unit. The displacement sensor, the gap sensor, and the inertial measurement unit are all mounted on the back plate of the mover, and are respectively mounted on the periphery of the printhead. The displacement sensor is used to measure the displacement of the printhead in at least one translational direction. The gap sensor is used to measure the distance between the printhead and the substrate to be printed. The inertial measurement unit is used to detect the micro-vibration and angular velocity of the printhead. The controller is connected to the displacement sensor, the gap sensor, and the inertial measurement unit respectively.

2. The magnetic levitation nozzle calibration system as described in claim 1, characterized in that, The gap sensor includes a capacitive gap sensor body, a first sensor mounting bracket, and a second sensor mounting bracket. The first sensor mounting bracket is installed on the side of the stator, and the capacitive gap sensor body is installed on the first sensor mounting bracket. The second sensor mounting bracket is connected to the first sensor mounting bracket and fixes the capacitive gap sensor body on the first sensor mounting bracket.

3. The magnetic levitation nozzle calibration system as described in claim 1, characterized in that, The mover assembly further includes an upper moving magnet fixing plate, a middle moving magnet fixing plate, and a lower moving magnet fixing plate. The permanent magnet array is mounted on the middle moving magnet fixing plate. The upper moving magnet fixing plate and the lower moving magnet fixing plate are respectively fixed on both sides of the middle moving magnet fixing plate. The upper moving magnet fixing plate is mounted on the back of the mover back plate.

4. The magnetic levitation nozzle calibration system as described in claim 1, characterized in that, The stator coil assembly further includes a stator coil mounting component and a stator coil cover plate. The stator coil mounting component is mounted on the front side of the stator back plate, and the electromagnetic coils are mounted on the stator coil mounting component. The stator coil cover plate covers a plurality of the electromagnetic coils and is connected to the stator coil mounting component.

5. A method for calibrating a magnetically levitated nozzle, characterized in that, The nozzle calibration method is implemented based on the magnetic levitation nozzle calibration system as described in any one of claims 1-4, and the nozzle calibration method includes: Based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, an initial state-space model is constructed as a prediction model, and the Gaussian process regression model is pre-trained offline to obtain the initial hyperparameters of the Gaussian process regression model. The displacement, angle, and acceleration information related to the nozzle are obtained from the sensor module and fused to obtain the current nozzle pose. Based on the actual nozzle pose and the electromagnetic coil current vector of the previous moment, the current nozzle pose prediction value is obtained by using the prediction model, and the residual between the current nozzle pose and the current nozzle pose prediction value is calculated. Obtain the electromagnetic coil current vector at the current moment, and combine the actual nozzle pose at the previous moment, the electromagnetic coil current vector at the previous moment, and the residual between the current nozzle pose and the current nozzle pose prediction value to form a new input and output sample pair. Then, use the new input and output sample pair to update the initial hyperparameters of the Gaussian process regression model. Based on the current nozzle pose and the control variable to be determined, the updated Gaussian process regression model is used to output the residual prediction mean and prediction variance. The residual prediction mean is fused with the prediction model to obtain the modified prediction model; Using the modified prediction model as the internal prediction model, the actual pose of the nozzle as the initial state, and the target pose as the tracking target, a cost function is defined, and the constraint boundary is adjusted based on the prediction variance to solve the optimal control input sequence that satisfies the constraint boundary in the future finite time domain. The instantaneous control quantities in the optimal control sequence are converted into current signals, and the current signals drive multiple electromagnetic coils of the stator coil assembly to adjust the real-time pose of the nozzle.

6. The magnetic levitation nozzle calibration method as described in claim 5, characterized in that, The prediction model is expressed as follows: In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, The state matrix, This is the control matrix.

7. The magnetic levitation nozzle calibration method as described in claim 5, characterized in that, Based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, an initial state-space model is constructed as the prediction model, and an offline pre-training of the Gaussian process regression model is performed to obtain the initial hyperparameters of the Gaussian process regression model, including: Based on the structural and dynamic characteristics of the magnetic levitation nozzle calibration system, an initial state-space model is constructed as a prediction model. Historical operating data or simulation data are used as initial training data. The initial training dataset includes multiple initial input and output sample pairs. Each initial input and output sample pair includes the electromagnetic coil current vector, the target pose, and the actual nozzle pose. Based on the initial training data, the prediction model is used to calculate the nozzle pose prediction value for each initial input and output sample pair, and the residual between the nozzle pose prediction value and the actual nozzle pose is calculated. The Gaussian process regression model is pre-trained offline using the electromagnetic coil current vector and target pose of the initial input and output sample pairs as independent variables and the corresponding residuals as dependent variables. The initial hyperparameters of the Gaussian process regression model are determined by minimizing the residual prediction error. The initial hyperparameters include kernel function parameters and noise variance.

8. The magnetic levitation nozzle calibration method as described in claim 5, characterized in that, The modified prediction model is expressed as follows: In the formula, Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, The state matrix, For the control matrix, This represents the mean of the predicted residuals.

9. The magnetic levitation nozzle calibration method as described in claim 5, characterized in that, Define the cost function as Then the cost function Represented as: In the formula, Indicates the target pose. Indicates at discrete time The actual position of the nozzle. Indicates at discrete time The electromagnetic coil current vector, This indicates the length of the prediction time domain, and corrects the number of time steps the prediction model advances in each optimization. This indicates the length of the control time domain and corrects the actual number of control steps optimized by the prediction model. , This is a summation index variable used to iterate through discrete time steps.

Images