High-speed wire bonding equipment motion platform displacement error dynamic compensation method and device

Abstract

The application relates to the technical field of bonding equipment displacement compensation, in particular to a high-speed wire bonding equipment motion platform displacement error dynamic compensation method and device, which comprises the following steps: collecting the state parameters of a motion platform in real time to obtain multi-dimensional real-time monitoring data; based on the multi-dimensional real-time monitoring data, a plurality of preset identification algorithms are executed in parallel to independently estimate platform real-time characteristic parameters; the platform real-time characteristic parameters are input into a preset multi-error source coupling prediction algorithm to calculate the predicted displacement deviation of the motion platform in a future predetermined time window; according to the requirement of a current bonding process step on positioning accuracy, a corresponding position accuracy tolerance interval is determined; a multi-objective optimization problem is established; the multi-objective optimization problem is solved in a rolling time domain to generate a cooperative compensation instruction in real time, and the cooperative compensation instruction is executed. Through the application, various time-varying coupling errors under high-speed motion conditions can be accurately predicted and cooperatively dynamically compensated.

Description

Technical Field

[0001] This invention relates to the technical field of displacement compensation for bonding equipment, and in particular to a method and apparatus for dynamic compensation of displacement error of the motion platform of high-speed wire bonding equipment. Background Technology

[0002] As a core piece of equipment in the semiconductor back-end packaging process, the displacement and positioning accuracy of the motion platform of the wire bonding machine directly determines the yield and reliability of wire bonding. With the continuous improvement of chip integration and the continuous shrinking of package size, the market has become more stringent in its requirements for the motion performance of high-speed wire bonding equipment. For example, the X / Y axis acceleration of the motion platform needs to reach 15G, the Z axis acceleration needs to exceed 100G, and the positioning accuracy needs to be stable within ±1μm to meet the needs of high-density, high-precision microelectronic packaging.

[0003] To ensure the positioning accuracy of the motion platform, existing dynamic compensation methods for motion platform displacement errors typically use sensors to collect state parameters such as displacement and velocity of the motion platform, and then perform compensation actions after estimating the error through specific identification algorithms. However, the displacement error of the motion platform in high-speed wire bonding equipment is caused by the superposition of multiple error sources, such as dynamic coupling, mechanical vibration, and inertial impact. Moreover, the influence characteristics of each error source change dynamically with the motion state and process stage. Existing compensation methods mostly use a single identification algorithm to estimate the platform characteristic parameters, which makes it difficult to capture the coupling effect of multiple error sources. At the same time, since compensation is performed based on the error data at the current moment, it is impossible to effectively predict the displacement deviation within a predetermined time window in the future, resulting in limited compensation accuracy. Summary of the Invention

[0004] This invention provides a method and apparatus for dynamic compensation of displacement error of the motion platform of a high-speed wire bonding device, which can effectively solve the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A dynamic compensation method for displacement error of the motion platform in high-speed wire bonding equipment includes:

[0007] The motion platform's status parameters are collected in real time to obtain multi-dimensional real-time monitoring data;

[0008] Based on the multi-dimensional real-time monitoring data, the platform's real-time characteristic parameters are independently estimated by multiple preset identification algorithms executed in parallel.

[0009] The real-time characteristic parameters of the platform are input into a preset multi-error-source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined time window in the future.

[0010] Based on the positioning accuracy requirements of the current bonding process steps, determine the corresponding position accuracy tolerance range;

[0011] A multi-objective optimization problem is established with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objectives.

[0012] A model predictive control optimizer is used to solve the multi-objective optimization problem in the rolling time domain under the constraint of satisfying the physical limits of the motion actuator, generate cooperative compensation instructions in real time, and execute them.

[0013] Furthermore, the multi-dimensional real-time monitoring data includes position feedback, vibration acceleration, temperature distribution, and process forces.

[0014] Furthermore, the real-time characteristic parameters of the platform include servo system dynamic parameters, mechanical structure vibration mode parameters, thermally induced deformation parameters, and load disturbance parameters.

[0015] Furthermore, the method of independently estimating the platform's real-time characteristic parameters through multiple preset identification algorithms executed in parallel includes:

[0016] For the dynamic parameters of the servo system, the recursive least squares algorithm is used to estimate the equivalent inertia, viscous damping coefficient and Coulomb friction coefficient of the servo loop in real time based on the deviation between the current command and position feedback of the servo motor.

[0017] For the vibration modal parameters of the mechanical structure, the vibration acceleration of the motion platform is analyzed in real time to extract the first and second dominant vibration frequencies, damping ratio and mode participation factor under the current motion state;

[0018] For thermal deformation parameters, based on the temperature distribution on the linear motor stator, guide rail and frame, and combined with the pre-calibrated transfer matrix between the thermal expansion coefficient and structural deformation, the platform positioning reference drift caused by thermal deformation is calculated in real time.

[0019] For the load disturbance parameters, the temporal correlation between the process force of the welding head and the motion trajectory command is analyzed, and the load disturbance components caused by bonding pressure and ultrasonic vibration reaction force are separated and quantified in real time.

[0020] Furthermore, the method for constructing the multi-error-source coupled prediction algorithm includes:

[0021] A servo tracking error sub-model is established, with motion commands and dynamic parameters of the servo system as inputs, and the output being the predicted tracking error caused by servo response lag and disturbances.

[0022] A vibration-induced error sub-model is established, whose inputs are the acceleration / jump command of the motion platform and the vibration mode parameters of the mechanical structure, and whose output is the predicted value of displacement oscillation caused by structural vibration.

[0023] A thermal deformation error sub-model is established, with the temperature distribution and thermal deformation parameters as inputs and the predicted values ​​of thermal expansion / contraction of the platform structure caused by the temperature gradient as outputs.

[0024] A load disturbance error sub-model is established, with the process force and load disturbance parameters as inputs and the predicted value of additional platform deformation caused by the dynamic force of the bonding process as output.

[0025] The predicted tracking error, displacement oscillation, thermal expansion / contraction, and additional deformation are vector-superimposed and fused using a preset time-varying weighting matrix to output the predicted displacement deviation.

[0026] Furthermore, the position accuracy tolerance range includes:

[0027] Different precision levels are preset for different bonding process steps;

[0028] Based on the currently executing step, the corresponding precision level is invoked and converted into a tolerance range centered on the target position with upper and lower limits;

[0029] Specifically, for the first solder joint of a spherical bond, the width of the tolerance interval is smaller than that for the second solder joint of a wedge bond.

[0030] Furthermore, the collaborative compensation command includes at least one of the following: a trajectory feedforward correction amount for the motion controller, a vibration suppression amount for the active damping device, and a parameter adjustment amount for the force-position control loop.

[0031] Furthermore, the position feedback is acquired using a dual-source synchronous acquisition and real-time calibration method using a grating ruler displacement sensor and a laser interferometer. The laser interferometer is symmetrically arranged on the outer ends of the crossbeams on both sides of the motion platform gantry mechanism, and the measuring optical axis is collinear with the measuring axis of the grating ruler.

[0032] Furthermore, the vibration acceleration is acquired using a piezoelectric triaxial accelerometer, which is connected to the platform via a flexible connection assembly. The flexible connection assembly includes a silicone damping pad and a metal fixing plate, which are used to isolate the transmission of non-target vibrations.

[0033] On the other hand, the present invention also provides a dynamic compensation device for displacement error of the motion platform of a high-speed wire bonding device, comprising:

[0034] The real-time acquisition module is used to collect the status parameters of the motion platform in real time and obtain multi-dimensional real-time monitoring data.

[0035] The parallel identification module is used to independently estimate the platform's real-time characteristic parameters based on the multi-dimensional real-time monitoring data through multiple preset identification algorithms executed in parallel.

[0036] The multi-error-source prediction module is used to input the real-time characteristic parameters of the platform into a pre-stored multi-error-source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined future time window.

[0037] The tolerance range determination module is used to determine the corresponding position accuracy tolerance range based on the positioning accuracy requirements of the current bonding process step.

[0038] The optimization problem construction module is used to establish a multi-objective optimization problem with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objectives.

[0039] The control optimization module is used to solve the multi-objective optimization problem in the rolling time domain under the constraint of satisfying the physical limits of the motion actuator, generate cooperative compensation instructions in real time, and control the actuator to execute the instructions.

[0040] The technical solution of this invention achieves the following technical effects: First, multiple preset identification algorithms executed in parallel independently estimate the platform's real-time characteristic parameters, and these parameters are input into a preset multi-error-source coupled prediction algorithm. This combined mechanism overcomes the limitation of a single model being unable to characterize complex coupled dynamics, enabling the system to simultaneously decouple and quantify the error contributions from different physical sources, and thereby accurately predict future displacement deviations in the time domain. Second, by combining the predicted displacement deviation with the position accuracy tolerance range determined according to process dynamics, and using this as the optimization objective, a model predictive control optimizer is employed for rolling time-domain solution, achieving a fundamental shift in control logic. The system transforms from a passive response to past or current errors to an active optimization of future states, thus eliminating the compensation lag problem caused by system inertia and control delay. Overall, the system not only improves dynamic accuracy through predictive feedforward, but also, due to the flexible constraints and global optimization capabilities introduced in multi-objective optimization by the model predictive control framework, the system can automatically achieve the optimal balance between energy consumption compensation, mechanical shock, and tracking performance while strictly meeting process accuracy requirements. This ensures the stability of the equipment under long-term high-speed operation and achieves synergistic optimization of accuracy maintenance and long-term equipment service.

[0041] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0042] 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 recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart illustrating the dynamic compensation method for displacement error of the high-speed wire bonding equipment motion platform in an embodiment of the present invention.

[0044] Figure 2 This is a structural block diagram of the dynamic compensation device for displacement error of the high-speed wire bonding equipment motion platform in an embodiment of the present invention. Detailed Implementation

[0045] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0047] like Figure 1 As shown, the dynamic compensation method for displacement error of the high-speed wire bonding equipment motion platform of the present invention specifically includes the following steps:

[0048] Step S100: Collect the status parameters of the motion platform in real time to obtain multi-dimensional real-time monitoring data;

[0049] Step S200: Based on multi-dimensional real-time monitoring data, the platform's real-time characteristic parameters are independently estimated by multiple preset identification algorithms executed in parallel.

[0050] Step S300: Input the real-time characteristic parameters of the platform into the preset multi-error source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined time window in the future.

[0051] Step S400: Determine the corresponding position accuracy tolerance range based on the positioning accuracy requirements of the current bonding process step;

[0052] Step S500: Establish a multi-objective optimization problem with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objective.

[0053] Step S600: Using a model predictive control optimizer, under the constraint of satisfying the physical limits of the motion actuator, the multi-objective optimization problem is solved in the rolling time domain, and collaborative compensation instructions are generated and executed in real time.

[0054] In this embodiment, firstly, multiple preset identification algorithms executed in parallel independently estimate the platform's real-time characteristic parameters, and these parameters are input into a preset multi-error-source coupled prediction algorithm. This combined mechanism overcomes the limitation of a single model being unable to characterize complex coupled dynamics, enabling the system to simultaneously decouple and quantify the error contributions from different physical sources, and thereby accurately predict future displacement deviations in the time domain. Secondly, the predicted displacement deviation is combined with the position accuracy tolerance range determined based on process dynamics, and this is used as the optimization objective. A model predictive control optimizer is employed for rolling time-domain solution, achieving a fundamental shift in control logic. The system transforms from a passive response to past or current errors to an active optimization of future states, thereby eliminating the compensation lag problem caused by system inertia and control delay. Overall, the system not only improves dynamic accuracy through predictive feedforward, but also, due to the flexible constraints and global optimization capabilities introduced in the model predictive control framework for multi-objective optimization, the system can automatically achieve the optimal balance between energy consumption compensation, mechanical shock, and tracking performance while strictly meeting process accuracy requirements. This ensures the stability of the equipment under long-term high-speed operation and achieves synergistic optimization of accuracy maintenance and long-term equipment service.

[0055] In a specific implementation, as one example, the multi-dimensional real-time monitoring data includes position feedback, vibration acceleration, temperature distribution, and process forces. By combining the core structural features of the high-speed wire bonding equipment's motion platform—a gantry-type dual-axis structure, direct drive of X / Y axis voice coil motors, Z-axis piezoelectric ceramic drive, and dual control of welding head force and position—as well as the high-speed, high-acceleration, micron-level positioning, and multi-error-source coupling characteristics of the equipment's motion platform, and relying on the equipment's structure for precise selection and structured deployment of sensing elements, a synchronous acquisition architecture at the hardware level is designed. The specific implementation is as follows:

[0056] For position feedback acquisition, grating ruler displacement sensors and laser interferometers are selected as position feedback acquisition sensing elements. Grating ruler displacement sensors are respectively deployed at the connection ends of the voice coil motor movers and slide rails on the X / Y axes of the motion platform, and at the connection ends of the piezoelectric ceramic drive end and welding head mounting base on the Z axis. The main scale of the grating ruler is connected to the fixed structure of the platform, and the reading head is connected to the moving parts of the platform. Laser interferometers are symmetrically deployed at the outer ends of the crossbeams on both sides of the gantry mechanism of the motion platform. The measuring optical axis of the laser interferometer is collinear with the measuring axis of the grating ruler. The laser reflector is fixed at the end of the corresponding moving part of the platform. The measurement range of the laser interferometer covers the entire motion stroke of the platform's X / Y / Z axes. The signal output ends of all grating ruler displacement sensors and laser interferometers are connected to the signal conditioning module. The laser interferometer performs real-time calibration on the data acquired by the grating ruler displacement sensors. The calibration cycle is synchronized with the servo cycle of the equipment motion controller. The signal conditioning module performs zero drift compensation and amplification processing on the calibrated position feedback signal.

[0057] For vibration acceleration acquisition, a piezoelectric triaxial accelerometer is selected as the sensing element. The sensor is deployed on the welding head mounting base of the motion platform, the X / Y axis guide rail slider, the connection end between the voice coil motor base and the frame, and the outer wall of the Z-axis piezoelectric ceramic drive end. A flexible connection component is added between the sensor and the platform. The flexible connection component consists of a silicone damping pad and a metal fixing plate. The metal fixing plate is fixed to the connection surfaces of the sensor and the platform, and the silicone damping pad is sandwiched between the metal fixing plate and the platform connection surface. The sensor and the flexible connection component are integrated into a single package. The sensor's output signal is transmitted to the signal conditioning module via a shielded cable. The signal conditioning module performs filtering and analog-to-digital conversion on the acquired signal. The filtering process eliminates the influence of high-frequency electromagnetic interference on the acquired signal.

[0058] For temperature distribution acquisition, NTC thermistor sensors were selected as the temperature sensing elements. The sensing elements were arranged in an array to form a temperature acquisition array. The sensing nodes were respectively deployed on the stator windings of the X / Y axis linear motors of the motion platform, the dynamic and static contact surfaces of the X / Y axis guide rail pairs, the temperature-sensitive areas of the crossbeams and longitudinal beams of the gantry mechanism, and the outer wall of the Z-axis piezoelectric ceramic drive end. All sensing nodes were fixed to the platform structure surface with high-temperature thermally conductive adhesive, which ensured the heat conduction efficiency between the sensing nodes and the platform structure. The connecting cables of the sensing nodes were made of heat-resistant insulated cables, which were laid along the cable grooves on the platform structure surface and filled with flame-retardant insulating adhesive. The signals from all sensing nodes were connected to a multi-channel temperature acquisition module to achieve centralized acquisition of temperature signals from multiple nodes.

[0059] For the acquisition of process forces, a miniature piezoelectric sensor is selected as the sensing element. The sensor is positioned between the connecting flange of the welding head on the motion platform and the Z-axis piezoelectric ceramic drive end. Both the upper and lower surfaces of the sensor are in contact with the connecting flange and the welding head base through ceramic gaskets, which ensure uniform force transmission and insulation performance. The signal output of the piezoelectric sensor, the signal acquisition end of the bonding force position control circuit, and the signal output of the grating ruler displacement sensor are all connected to a hardware trigger module. The hardware trigger module realizes the timing binding of the process force acquisition signal and the position feedback acquisition signal. The sensor acquisition signal covers the bonding pressure and ultrasonic vibration reaction force of the welding head during the bonding process. The acquired signal is transmitted to the signal conditioning module for amplification and analog-to-digital conversion processing via a shielded cable.

[0060] Furthermore, a synchronous acquisition and control module based on an FPGA is constructed. Signals from the grating ruler displacement sensor, laser interferometer, piezoelectric triaxial accelerometer, NTC thermistor sensor, and miniature piezoelectric sensor, after processing, are all connected to this synchronous acquisition and control module. The module provides a unified hardware trigger clock and sampling frequency for all sensing elements, with the sampling frequency matching the servo cycle of the device's motion controller. The module performs preliminary preprocessing on the digital signals acquired from each channel, including zero-drift compensation, outlier removal, and signal normalization. The preprocessed data is stored in an external buffer chip connected to the synchronous acquisition and control module. This buffer chip establishes a bidirectional data interaction channel with the subsequent parallel identification algorithm module via a high-speed bus, enabling real-time retrieval of the acquired data.

[0061] Furthermore, all signal cables for the sensing elements are double-shielded, with the outer shielding layer grounded and the inner shielding layer connected to the signal ground. The synchronous acquisition control module, signal conditioning module, and multi-channel temperature acquisition module are all encapsulated in metal shielding boxes, which are grounded. The power supply for the acquisition system is an isolated switching power supply to prevent voltage fluctuations in the main circuit of the equipment from interfering with the acquisition system. This comprehensive anti-interference design of the acquisition system is achieved from signal transmission, module encapsulation, and power supply.

[0062] In this embodiment, the FPGA-based hardware synchronous acquisition architecture provides a unified trigger clock and sampling frequency for all sensing elements. The sampling frequency matches the servo cycle of the device motion controller, ensuring consistent timing of multi-dimensional data acquisition for position feedback, vibration acceleration, temperature distribution, and process forces. This adapts to the platform's high-speed, high-acceleration operating characteristics and avoids parameter estimation deviations caused by timing differences. The position feedback employs a dual-source acquisition and real-time calibration design using a grating ruler displacement sensor and a laser interferometer. This effectively corrects the inherent system errors of the grating ruler displacement sensor, improving the accuracy of the motion platform's position feedback data and meeting the micron-level positioning requirements of the high-speed wire bonding equipment's motion platform. The vibration acceleration sensing element is connected to the platform via a flexible connection component, effectively isolating the sensor from the platform. Non-target vibration transmission is achieved by deploying sensing nodes in the vibration-sensitive areas of the platform. The collected data accurately reflects the actual vibration state of the motion platform under high-speed and high-acceleration operation. Temperature parameters are obtained by using an array of sensing nodes to form a temperature acquisition array. The collected data accurately reflects the temperature distribution and gradient characteristics of key structures such as the stator winding of the linear motor, the guide rail pair, the gantry mechanism, and the piezoelectric ceramic drive end. Combined with the efficient heat conduction of the high-temperature thermally conductive adhesive, the collected data can reflect the spatial distribution law of thermal deformation of the platform in real time. The process force and position feedback signal are time-bound through a hardware trigger module, eliminating the signal delay caused by software timing matching. This achieves precise spatiotemporal matching between the process force data and the corresponding position motion parameters, and the collected data accurately reflects the changes in process force during the bonding process.

[0063] In some embodiments of the present invention, based on the multi-dimensional real-time monitoring data provided in step S100, the dynamic parameters of the servo system, the vibration modal parameters of the mechanical structure, the thermal deformation parameters, and the load disturbance parameters are independently estimated by multiple parallel-executed identification algorithm modules.

[0064] Specifically, the identification of dynamic parameters of the servo system is implemented as follows: A second-order simplified dynamic model of the servo system is established. The model input is the current command of the servo motor, and the model output is the position response of the motor. The model parameters include equivalent inertia, viscous damping coefficient, and Coulomb friction coefficient. The real-time current command sequence obtained in step S100 and the highly synchronized grating ruler position feedback sequence are input into a recursive least squares parameter estimator. In each servo control cycle, the estimator recursively updates the estimated values ​​of the above model parameters using the latest input-output data pairs. The recursive least squares algorithm adopts a form with a forgetting factor, which is dynamically adjusted according to the current acceleration command of the motion platform. A smaller forgetting factor is used in the high-speed segment or under high acceleration command to quickly track parameter changes, and a larger forgetting factor is used in the low-speed stable segment to maintain estimation stability. In this way, the system obtains the servo loop dynamic parameters that evolve in real time with the motion state online.

[0065] The identification of vibration modal parameters of the mechanical structure is implemented as follows: The synchronous vibration acceleration signals collected from multiple key parts in step S100 are input into the real-time vibration modal analysis module. This module first performs windowing processing on the acceleration data of each channel, and the window function length is associated with an integer multiple of the current main motion cycle of the platform. Subsequently, joint time-frequency analysis is performed on the windowed multi-channel signal set, specifically using an online variant of the random subspace identification algorithm based on covariance driving. This algorithm directly extracts the state space matrix of the system from the time domain data, and then calculates the frequency, damping ratio, and mode participation factor of the first and second dominant vibration modes in real time through eigenvalue decomposition. This process is completed independently within each analysis window, thereby realizing quasi-real-time updating of vibration modal parameters, and its update rate is independent of and usually lower than the servo cycle.

[0066] The identification of thermally induced deformation parameters is implemented as follows: The temperature distribution data obtained by the temperature sensor array in step S100, after spatial interpolation, is combined with the pre-generated thermal structure deformation transfer matrix. This transfer matrix is ​​obtained in advance through a combination of finite element analysis and experimental calibration to establish a quantitative relationship between the unit temperature change at each temperature measuring point and the thermally induced displacement of the end of the motion platform in the X, Y, and Z directions. During real-time operation, the thermally induced deformation parameters, i.e., the platform positioning reference drift caused by the current temperature field, are calculated as follows: The difference between the temperature value of each temperature measuring point at the current moment and the set reference temperature such as the cold state temperature of the equipment is used as the input vector, and the thermal structure deformation transfer matrix is ​​multiplied on the left to directly output the three-dimensional real-time thermal drift vector. This calculation process is executed immediately after the temperature data is updated, and its output is the thermally induced deformation parameters.

[0067] The identification of load disturbance parameters is implemented as follows: The load disturbance parameters aim to quantify the dynamic disturbances introduced by the bonding process itself. The process force signal collected in step S100 and the corresponding platform motion trajectory command, which is strictly timestamped, are acquired simultaneously. The process force signal includes static pressure and dynamic ultrasonic components. First, the periodic dynamic force component related to the ultrasonic transducer driving frequency is separated from the process force signal through an adaptive filter. Then, the correlation between the remaining non-periodic force component and motion trajectory commands such as acceleration and jerk commands in the time domain is analyzed. By calculating the cross-correlation function within a sliding time window and identifying its peak value and phase, the non-periodic force is decomposed into a component coupled with the motion inertial force and a pure process force component directly related to the bonding contact process. The final output load disturbance parameters are a structured data set including the dynamic ultrasonic force amplitude, the inertial coupling coefficient, and the pure process force amplitude.

[0068] In this embodiment, recursive least squares with an adaptive forgetting factor is used for servo parameters, which can accurately match the electromechanical time constant of the servo loop and achieve rapid tracking of parameters such as inertia and damping. Online random subspace identification is used for vibration modes, which can directly extract modal parameters from the time domain response, avoiding the limitation of traditional frequency domain methods that require steady-state data and adapting to transient processes. The thermally induced deformation parameters are calculated by direct mapping of the transfer matrix, which can transform complex unsteady heat conduction calculations into efficient matrix operations and achieve near-instantaneous response to temperature field changes. The load disturbance parameters are decoupled through time domain correlation analysis, which can separate the process force from the complex coupled signal and accurately quantify the load influence from different sources.

[0069] In a specific implementation, as one example, a multi-error-source coupling prediction framework is constructed by combining the core structural features of the high-speed wire bonding equipment motion platform: a gantry-type dual-axis structure, direct drive of X / Y axis voice coil motors, piezoelectric ceramic drive of the Z axis, and dual control of welding head force and position. This framework leverages the platform's high-speed, high-acceleration, micron-level positioning, and multi-error-source coupling characteristics. Independent prediction sub-models are established for the physical causes and dynamic characteristics of each error source. A time-varying weighted matrix is ​​introduced to dynamically adapt the influence weights of the error sources. Through rolling time-domain extrapolation, the displacement deviation sequence within a predetermined future time window is predicted, quantifying the deviation under the coupling effect of multiple error sources. The specific implementation is as follows:

[0070] Step S310: For the servo tracking error sub-model, based on the driving characteristics of the X / Y axis voice coil motors and Z axis piezoelectric ceramics of the high-speed wire bonding equipment motion platform, a second-order linear dynamic model of the servo system is established. The model input consists of the position, velocity, and acceleration command sequence of the motion platform and the dynamic parameters of the servo system obtained through parallel identification. The dynamic parameters of the servo system include equivalent inertia, viscous damping coefficient, and Coulomb friction coefficient. The angular displacement of the motor is converted into the linear displacement of the platform through coordinate transformation. A hysteresis compensation term is introduced to quantify the servo response hysteresis effect. The hysteresis compensation term is calibrated based on the step response characteristics of the servo system and is associated with the real-time velocity of the platform and the servo response time constant. A disturbance observation term is introduced to characterize the influence of random disturbances and is associated with the position command and the real-time position feedback of the grating ruler. The Runge-Kutta numerical integration algorithm is used to solve the dynamic model. The integration step size is consistent with the servo cycle of the equipment motion controller. Through iterative calculation, the predicted value sequence of tracking errors caused by servo response hysteresis and disturbances within a predetermined time window is output. The length of the predetermined time window is set according to the single-step duration of the bonding process and covers the complete cycle from the current moment to the next process switch.

[0071] Step S320: For the vibration-induced error sub-model, based on the mechanical characteristics of the gantry-type dual-axis structure of the motion platform, the guide rail slider, and the welding head mounting base, a vibration-induced error sub-model is established. The model inputs are the acceleration and jerk commands of the motion platform and the mechanical structure vibration modal parameters obtained through parallel identification. The mechanical structure vibration modal parameters include the first and second order dominant vibration frequencies, damping ratios, and mode shape participation factors. The sub-model is constructed using the modal superposition method. First, the acceleration and jerk commands are converted into structural vibration excitations. The excitation amplitude is calibrated through the excitation transfer coefficient, which is derived from the stiffness matrix and mass matrix of the platform structure. For the first and second order dominant vibration frequencies, the model is constructed using the modal superposition method. For the second dominant vibration modes, single-degree-of-freedom vibration equations are established, which include parameters such as modal mass, modal damping, and modal stiffness. Modal damping is obtained by relating damping ratio, modal mass, and modal stiffness, while modal stiffness is obtained by relating dominant vibration frequency and modal mass. The vibration equations are solved using the complex exponential method to obtain the time-domain response of each mode. Multimodal response superposition is achieved through time-domain synthesis. Modal weighting coefficients are introduced in the synthesis process, and these coefficients are allocated according to the magnitude of the mode participation factor. Modes with larger mode participation factors correspond to higher weighting coefficients. Finally, a sequence of predicted displacement oscillation values ​​caused by structural vibration within a predetermined future time window is output.

[0072] Step S330: For the thermally induced deformation error sub-model, based on the theories of heat conduction and thermoelasticity, and combined with the structural characteristics of the platform's linear motor stator, guide rail pair, and frame, a thermally induced deformation error sub-model is established. The model input consists of array-type temperature distribution data from multi-dimensional real-time monitoring data and thermally induced deformation parameters obtained through parallel identification. The thermally induced deformation parameters represent the platform's positioning reference drift. First, spatial interpolation processing is performed on the temperature distribution data. The Kriging interpolation algorithm is used to complete the missing temperature values ​​in key structural regions to obtain a continuous temperature field, and then the temperature gradient of each key structure is calculated. The core of the sub-model is described by thermoelasticity equations. The relationship between temperature and deformation is established through equations that incorporate the relationships between strain tensor, displacement components, coefficient of thermal expansion, temperature difference, stress tensor, shear modulus, and Poisson's ratio. The model boundary conditions are corrected using thermally induced deformation parameters obtained through parallel identification. The drift of the positioning reference is substituted into the equations as the initial deformation. The solution is obtained through finite element discretization, and the structure is meshed using tetrahedral elements to obtain the thermally induced displacement components at each node of the platform. Displacement data at the end of the moving platform is extracted, and a sequence of predicted values ​​for thermal expansion / contraction of the platform structure due to temperature gradients within a predetermined future time window is output. The model update frequency is synchronized with the temperature acquisition data update frequency.

[0073] Step S340: For the load disturbance error sub-model, a load disturbance error sub-model is established based on the contact mechanical characteristics of the bonding head with the chip and substrate in the bonding process. The model input consists of the process force signal from multi-dimensional real-time monitoring data and the load disturbance parameters obtained through parallel identification. The process force signal includes bonding pressure and ultrasonic vibration reaction force, and the load disturbance parameters include dynamic ultrasonic force amplitude, inertial coupling coefficient, and pure process force amplitude. First, the transmission law of the process force is quantified by modeling the mechanical transmission path. The transmission path includes the bonding head, the Z-axis piezoelectric ceramic drive end, and the platform body. A transfer function is established to correlate the process force input and the platform end displacement output. The process force signal is decomposed into static and dynamic components through offline hammer impact tests and frequency response function analysis. The static component corresponds to the pure process force amplitude, while the dynamic component is related to the dynamic ultrasonic force amplitude and the ultrasonic transducer drive frequency. The displacement response is calculated by substituting these components into the transfer function. An inertial coupling term is introduced to quantify the coupling effect between the process force and the platform's motion inertia. The inertial coupling term is associated with the inertial coupling coefficient and the platform's real-time acceleration command. The final output of the sub-model is the superposition result of each component, which yields a sequence of predicted values ​​for the additional platform deformation caused by the dynamic force of the bonding process within a predetermined time window. A load disturbance coefficient is introduced during the superposition process, and the load disturbance coefficient is calibrated according to the bonding process type.

[0074] Step S350: Perform vector standardization on the above four types of error prediction value sequences to unify data dimensions and units. The standardization process is associated with the mean and standard deviation of historical error sequences to ensure the rationality of weight allocation for each error component during fusion. Construct a time-varying weighted matrix, with matrix elements representing the weight coefficients of each error component. The weight coefficients are calibrated through offline experiments and a parameter library is established. The dynamic adjustment logic is determined based on the current motion state, process stage, and temperature gradient: During the high-speed, high-acceleration motion stage of the platform, increase the weight coefficients of vibration-induced error and servo tracking error; during the bonding execution stage, increase the weight coefficient of load disturbance error; and during the temperature gradient... During the stage of significant temperature change, the weighting coefficient of thermally induced deformation error is increased. The weighting adjustment is achieved by looking up a table, calling the weight values ​​in the parameter library according to the current real-time status. The standardized error prediction value is multiplied by the corresponding weighting coefficient, and the coupled displacement deviation prediction value is obtained through vector superposition operation. The deviation prediction within a predetermined time window is completed by using a rolling time domain extrapolation method. The window length is set according to the process single-step duration. During the extrapolation process, after each step of prediction is completed, the current collected data and identification parameters are updated, the input conditions of each sub-model are corrected, and the prediction results are rolled for optimization. Finally, a complete predicted displacement deviation sequence is output.

[0075] In this embodiment, each prediction sub-model is constructed based on the physical causes of the error sources and the structural characteristics of the equipment. The model input is directly related to multi-dimensional monitoring data and parallel identification parameters. The equation form, parameter association, and solution algorithm are all precisely matched with the characteristics of the error sources to ensure the pertinence of the prediction of each error component. The time-varying weighted matrix is ​​calibrated through offline experiments and the weight coefficients are dynamically adjusted according to the motion state, process stage, and temperature gradient. This fully considers the dynamic changes in the degree of influence of different error sources and their mutual coupling effects, avoiding fusion bias caused by fixed weights. The dynamic time-domain extrapolation method uses the servo cycle as the step size and combines real-time data updates to achieve rolling optimization of the prediction results, outputting the deviation sequence within the future predetermined time window, which meets the real-time compensation requirements of high-speed wire bonding equipment.

[0076] In a specific implementation, as one example, considering the process execution characteristics of high-speed wire bonding equipment, the solder joint forming requirements of different bonding steps, and the operational needs of equipment micron-level positioning and dynamic process switching, based on the equipment process execution controller and process parameter library, a dedicated accuracy level for each process step is preset, and a quantitative conversion rule is established for the accuracy level to a tolerance range centered on the target position. Differentiated tolerance range widths are set for the first solder joint of ball bonding and the second solder joint of wedge bonding to achieve matching and retrieval of position accuracy tolerance ranges; the specific implementation is as follows:

[0077] Step S410: Analyze the entire process of high-speed wire bonding, including core operations such as ball bonding first solder joint formation, wedge bonding second solder joint formation, wire traction, and point switching. Based on the process characteristics, solder joint function, and bonding yield requirements of each step, a specific accuracy level is preset for each step. Accuracy levels are divided according to the allowable positioning error range. The ball bonding first solder joint formation step is set to the highest accuracy level, while the wedge bonding second solder joint formation step has a lower accuracy level. Non-solder joint formation steps such as wire traction and point switching are set to a basic accuracy level. The allowable positioning error range corresponding to all accuracy levels is calibrated through offline process testing. The calibration basis is the minimum positioning accuracy requirement for bonding yield to meet the process standard under each step. The correlation between process steps and corresponding accuracy levels is stored in the equipment's process parameter library, establishing a one-to-one corresponding call index.

[0078] Step S420: The equipment's process execution controller collects the execution instructions of the current bonding process in real time. The process step identification module parses the process code, operation type, solder joint type and other information in the execution instructions. Based on the parsing results, it matches the call index in the process parameter library to identify the bonding process step that is currently being executed in real time. The identification process is carried out synchronously with the parsing of the process execution instructions. Based on the identification results, the preset accuracy level corresponding to the step is retrieved from the process parameter library to ensure that the accuracy level is accurately matched with the current process step.

[0079] Step S430: The process execution controller parses the point coordinate information in the current bonding process execution command and extracts the three-dimensional spatial coordinates of the target bonding position. These coordinates are used as the center coordinates of the position accuracy tolerance range. The coordinate parsing accuracy is consistent with the positioning accuracy of the equipment motion platform, which is adapted to the technical requirements of the equipment micron-level positioning, and ensures that the spatial position of the tolerance range corresponds precisely to the actual bonding target position.

[0080] Step S440: Establish a quantization conversion rule from accuracy level to position accuracy tolerance range. Map the allowable positioning error range corresponding to the retrieved accuracy level to a three-dimensional tolerance range centered on the three-dimensional spatial coordinates of the target bonding position in a positive-negative symmetrical manner. The converted tolerance range has clear upper and lower limit thresholds. The upper and lower limit thresholds in each direction are the target position coordinates superimposed with the allowable positive and negative values ​​of the positioning error in the corresponding direction. This completes the quantization conversion from accuracy level to a tolerance range centered on the target position with upper and lower limits.

[0081] Step S450: Based on the process characteristics of the ball-bonded first solder joint and the wedge-bonded second solder joint, during offline process test calibration, a differentiated allowable range of positioning error is set so that the tolerance interval width corresponding to the ball-bonded first solder joint is smaller than the tolerance interval width corresponding to the wedge-bonded second solder joint. The tolerance interval width is the difference between the upper and lower threshold values ​​in each direction. The calibration basis for the differentiated width is the forming process requirements of the two solder joints, the mechanical property compliance conditions of the solder joints, and the accuracy requirements of the actual bonding process.

[0082] Step S460: The converted position accuracy tolerance range is transmitted in real time to the subsequent multi-objective optimization module as the basis for the accuracy constraint of the multi-objective optimization problem. When the process steps of the equipment are switched or the target bonding position is changed, the process execution controller re-executes the complete process of process step identification, accuracy level retrieval, target position coordinate analysis, and tolerance range conversion, and updates the position accuracy tolerance range in real time to ensure dynamic adaptation between the tolerance range and the current process execution state.

[0083] In this embodiment, a dedicated accuracy level is preset according to the actual accuracy requirements of each bonding process step, so that the position accuracy tolerance range is precisely matched with the process characteristics and solder joint forming requirements of each step, avoiding the problem of insufficient or excessive compensation caused by a uniform tolerance standard; the real-time identification of process steps and the synchronous retrieval of accuracy levels ensure that the matching of accuracy levels and process execution have no time delay, adapting to the high-speed process execution characteristics of the equipment and meeting the technical requirements of real-time compensation of the motion platform; the tolerance range is constructed with the three-dimensional spatial coordinates of the target bonding position as the center, so that the tolerance range corresponds precisely to the spatial position of the actual bonding point, avoiding the problem that the tolerance range with a fixed numerical range cannot adapt to different bonding points; different tolerance range widths are set for the first solder joint of spherical bonding and the second solder joint of wedge bonding to match the different positioning accuracy requirements of the two solder joints. The narrow tolerance range of the first solder joint of spherical bonding ensures the positioning accuracy of the core solder joint, while the wide tolerance range of the second solder joint of wedge bonding avoids over-compensation, taking into account both the accuracy requirements of the bonding process and the operating efficiency of the equipment.

[0084] In a specific implementation, as one example, regarding step S500, by combining the micron-level positioning compensation requirements of the high-speed wire bonding equipment and the physical operating characteristics of the compensation actuator, and relying on the predicted displacement deviation coupled by multiple error sources and the position accuracy tolerance range of dynamic matching, the optimization objective is quantitatively modeled, the priority and quantitative representation methods of the primary and secondary objectives are clarified, the objective dimensions are unified and optimization constraints are established, and a complete multi-objective optimization problem model is integrated; the specific implementation is as follows:

[0085] Step S510: Construct the primary optimization objective function to quantitatively characterize the difference between the predicted displacement deviation and the position accuracy tolerance interval. The function input is the predicted displacement deviation sequence output by S300 and the position accuracy tolerance interval determined by S400, with minimizing this difference as the core objective. The predicted displacement deviation is decomposed into components in three-dimensional space, and the deviation of each component from the upper and lower limits of the tolerance interval in the corresponding direction is calculated. The deviation exceeding the tolerance interval is assigned a basic weight, and the deviation fluctuation within the tolerance interval is assigned a correction weight. The correction weight is less than the basic weight. The quantitative value of the primary objective function is obtained by weighted summation. The smaller the quantitative value, the smaller the difference between the predicted displacement deviation and the tolerance interval. The weight coefficients are calibrated through offline process experiments, and the calibration basis is the degree of influence of the positioning accuracy in each direction on the bonding yield.

[0086] Step S520: Construct a secondary optimization objective function to quantitatively characterize and fuse the energy consumption and mechanical impact of the compensation action, with minimizing this fused value as the secondary objective. For energy consumption, the quantitative characterization is the product of the driving power of the compensation actuator and the duration of the compensation action, with the driving power correlated with the amplitude of the trajectory feedforward correction and vibration suppression. For mechanical impact, the quantitative characterization is the rate of change of acceleration of the compensation action, with the rate of change of acceleration correlated with the adjustment rate of each compensation parameter. A fixed-weight fusion method is used to integrate the quantitative values ​​of energy consumption and mechanical impact into a single secondary objective function. The weight coefficients are calibrated through offline testing, based on the equipment's energy consumption control standards and structural impact resistance requirements.

[0087] Step S530: Determine the optimization variables for the multi-objective optimization problem. The optimization variables are the controllable parameters of the compensation actuator, including the trajectory feedforward correction for the motion controller, the vibration suppression for the active damping device, and the parameter adjustment for the force-position control loop. Each optimization variable is represented in three-dimensional component form. Determine the value range of each optimization variable. The value range is calibrated by the physical performance constraints of the compensation actuator. The value range of the trajectory feedforward correction matches the maximum compensation stroke of the motion platform, the value range of the vibration suppression matches the output limit of the active damping device, and the value range of the parameter adjustment for the force-position control loop matches the stable adjustment range of the control loop.

[0088] Step S540: Establish priority constraints for multi-objective optimization, clarifying that the primary optimization objective has a higher priority than the secondary optimization objective. During the constraint solution process, the optimization of the secondary objective is based on the premise that the optimization result of the primary objective meets the accuracy constraint. That is, after the quantized value of the primary objective function drops below the threshold of the process calibration, the secondary objective is then optimized. At the same time, establish physical boundary constraints for the optimization variables, constraining the values ​​of each optimization variable to not exceed the preset physical performance range, so as to avoid the failure of the compensation actuator or abnormal operation of the equipment due to parameter over-limit.

[0089] Step S550: Normalize the primary and secondary objective functions to eliminate dimensional differences between different objectives and ensure the rationality of the optimization solution; statistically analyze the historical extreme values ​​and calibration extreme values ​​of each objective function within the constraints of the process and equipment, and map the quantized values ​​of each objective function to the same numerical range based on the extreme values. The normalized objective function values ​​are all dimensionless, and the numerical change trend is consistent with the original function, that is, the smaller the quantized value of the original function, the smaller the normalized value.

[0090] Step S560: Integrate the normalized primary and secondary objective functions, optimization variables, and constraints to form a complete multi-objective optimization problem model. The model takes minimizing the normalized primary objective function as the core and minimizing the normalized secondary objective function as an auxiliary. The solution input of the model is determined to be the predicted displacement deviation sequence and the position accuracy tolerance interval. The solution output of the model is the set of compensation parameters to be optimized, that is, the specific values ​​of each optimization variable.

[0091] In this embodiment, the primary optimization objective function quantifies and weights the difference between the predicted displacement deviation and the tolerance range, ensuring that the optimization direction precisely aligns with the positional accuracy constraints. The secondary optimization objective function quantitatively integrates energy consumption and mechanical impact, incorporating equipment operating efficiency and structural stability into the optimization system. This avoids ineffective energy consumption and excessive impact caused by single-precision optimization, ensuring that the optimization results take into account both the equipment's compensation performance and operational performance. Clear optimization variables and physical boundary constraints ensure that the solution range of the multi-objective optimization problem aligns with the actual operating capabilities of the compensation actuator, avoiding ineffective optimization solutions that exceed physical performance. The priority constraints of the primary and secondary objectives establish the core position of accuracy optimization, ensuring that the compensation action first meets the positioning accuracy requirements of the bonding process, while simultaneously defining reasonable boundaries for the optimization of secondary objectives.

[0092] In some embodiments of the present invention, regarding step S600, based on the multi-objective optimization problem model established in step S500, the real-time predicted displacement deviation output in S300, and the dynamic position accuracy tolerance range determined in S400, a model predictive control optimizer is used to construct a rolling time-domain solution framework, synchronously incorporating the physical limit constraints of the actuator, and generating time-synchronized collaborative compensation instructions. The specific implementation is as follows:

[0093] Step S610: Configure the model predictive control optimizer. Combining the dynamic response characteristics and compensation execution requirements of the high-speed wire bonding equipment motion platform, determine the prediction time domain and control time domain of the optimizer. For example, the prediction time domain covers 1.5 times the duration of the future predetermined time window, and the control time domain is consistent with the equipment servo cycle. The parameters of the prediction time domain and control time domain are calibrated through offline experiments. The calibration basis is the dynamic response time of the platform and the action response delay of the compensation execution mechanism to ensure that the solution rhythm of the optimizer is accurately matched with the platform operating status and compensation execution speed.

[0094] Step S620: Set the rolling time domain solution rules. The optimizer's solution process starts synchronously with the device servo cycle. A complete optimization solution is completed in each servo cycle. During the solution, the latest predicted displacement deviation sequence output by S300 at the current time, the current position accuracy tolerance range determined by S400, and the real-time operating status parameters of the motion actuator are imported. After the solution is completed, only the compensation parameters corresponding to the first control step in the control time domain are extracted as the execution parameters of the current cycle. The compensation parameters of the remaining control steps are discarded. The updated input parameters are re-imported in the next servo cycle, and the solution process is repeated to realize the dynamic solution in the rolling time domain.

[0095] Step S630: Import constraints into the model predictive control optimizer. The constraints include the physical limit constraints of the motion actuator and the optimization constraints set in S500. The physical limit constraints of the motion actuator cover the maximum compensation stroke of the motion platform corresponding to the trajectory feedforward correction, the output limit of the active damping device corresponding to the vibration suppression, and the control loop adjustment limit corresponding to the force position control loop parameter adjustment. Each limit parameter is calibrated through the factory parameters and offline performance tests of the actuator to ensure that each compensation parameter does not exceed the operating capacity of the mechanism during the optimization solution process, and to avoid actuator failure.

[0096] Step S640: The optimizer calls the built-in multi-objective solution algorithm. Based on the normalized multi-objective optimization problem model established in S500, it solves the optimization variables according to the constraint rule that the primary objective has higher priority than the secondary objective. During the solution process, the quantized values ​​of the primary objective function and the secondary objective function corresponding to each optimization variable are calculated in real time. The optimal compensation parameter set that satisfies all constraints is found through iterative calculation. The number of solution iterations is calibrated through offline experiments. The calibration basis is the balance between solution accuracy and solution time, ensuring that the solution results meet the accuracy requirements without causing time delay.

[0097] Step S650: Integrate and generate collaborative compensation instructions. The compensation parameter set obtained from the optimization solution is decomposed into corresponding instruction components according to the type of compensation execution component. These components include at least the trajectory feedforward correction amount for the motion controller, the vibration suppression amount for the active damping device, and the parameter adjustment amount for the force-position control loop. The timing synchronization of each instruction component is calibrated, specifying the issuance time, execution start time, and execution duration of each component. This ensures that each compensation instruction component starts synchronously and executes collaboratively, avoiding superimposed interference between different compensation actions. The timing parameter calibration is based on the action response speed of each execution component.

[0098] Step S660: The integrated collaborative compensation command is transmitted to the corresponding actuator in real time via the high-speed bus. After receiving the trajectory feedforward correction amount, the motion controller corrects the motion trajectory command in real time. After receiving the vibration suppression amount, the active damping device adjusts the damping output parameters. After receiving the parameter adjustment amount, the force position control loop optimizes the bonding pressure and ultrasonic vibration parameters. The action response of all actuators and the issuance of commands are synchronized.

[0099] In this embodiment, the rolling time-domain solution method of the model predictive control optimizer is synchronized with the equipment servo cycle. Each cycle imports updated predicted displacement deviations and position accuracy tolerance ranges, enabling real-time updates of compensation parameters. This ensures precise matching between the collaborative compensation commands and the platform's actual operating state and error evolution patterns. The solution process incorporates the physical limit constraints of the motion actuator, combined with offline calibrated limit parameters, to avoid generating compensation commands exceeding the actuator's operating capabilities. This ensures the executability of the collaborative compensation commands, protects the actuator from damage, and guarantees equipment operational stability. Each component of the collaborative compensation command undergoes time-series synchronous calibration, enabling simultaneous execution of trajectory feedforward correction, vibration suppression, and force-position parameter adjustment, eliminating superimposed interference between different compensation actions. The optimizer's built-in multi-objective solution algorithm follows the priority constraints of primary and secondary objectives. The solution results satisfy both the accuracy requirements corresponding to the primary objective and the energy consumption and mechanical shock control corresponding to the secondary objectives, achieving a synergistic balance between accuracy compensation and equipment operating performance.

[0100] Based on the same inventive concept as the dynamic compensation method for displacement error of a high-speed wire bonding equipment motion platform in the foregoing embodiments, this invention also provides a dynamic compensation device for displacement error of a high-speed wire bonding equipment motion platform, such as... Figure 2 As shown, the device includes:

[0101] The real-time acquisition module is used to collect the status parameters of the motion platform in real time and obtain multi-dimensional real-time monitoring data.

[0102] The parallel identification module is used to independently estimate the platform's real-time characteristic parameters based on the multi-dimensional real-time monitoring data through multiple preset identification algorithms executed in parallel.

[0103] The multi-error-source prediction module is used to input the real-time characteristic parameters of the platform into a pre-stored multi-error-source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined future time window.

[0104] The tolerance range determination module is used to determine the corresponding position accuracy tolerance range based on the positioning accuracy requirements of the current bonding process step.

[0105] The optimization problem construction module is used to establish a multi-objective optimization problem with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objectives.

[0106] The control optimization module is used to solve the multi-objective optimization problem in the rolling time domain under the constraint of satisfying the physical limits of the motion actuator, generate cooperative compensation instructions in real time, and control the actuator to execute the instructions.

[0107] The system described above in this invention can effectively realize the dynamic compensation method for displacement error of the motion platform of high-speed wire bonding equipment, and the technical effects it can achieve are as described in the above embodiments, which will not be repeated here.

[0108] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for dynamic compensation of displacement error of a motion platform in a high-speed wire bonding device, characterized in that, include: The motion platform's status parameters are collected in real time to obtain multi-dimensional real-time monitoring data; Based on the multi-dimensional real-time monitoring data, the platform's real-time characteristic parameters are independently estimated by multiple preset identification algorithms executed in parallel. The real-time characteristic parameters of the platform are input into a preset multi-error-source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined time window in the future. Based on the positioning accuracy requirements of the current bonding process steps, determine the corresponding position accuracy tolerance range; A multi-objective optimization problem is established with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objectives. A model predictive control optimizer is used to solve the multi-objective optimization problem in the rolling time domain under the constraint of satisfying the physical limits of the motion actuator, generate cooperative compensation instructions in real time, and execute them.

2. The method for dynamic compensation of displacement error of the motion platform of high-speed wire bonding equipment according to claim 1, characterized in that, The multi-dimensional real-time monitoring data includes position feedback, vibration acceleration, temperature distribution, and process forces.

3. The method for dynamic compensation of displacement error of the motion platform of high-speed wire bonding equipment according to claim 2, characterized in that, The platform's real-time characteristic parameters include servo system dynamic parameters, mechanical structure vibration mode parameters, thermal deformation parameters, and load disturbance parameters.

4. The dynamic compensation method for displacement error of the motion platform of the high-speed wire bonding equipment according to claim 3, characterized in that, The method of independently estimating the platform's real-time characteristic parameters through multiple preset identification algorithms executed in parallel includes: For the dynamic parameters of the servo system, the recursive least squares algorithm is used to estimate the equivalent inertia, viscous damping coefficient and Coulomb friction coefficient of the servo loop in real time based on the deviation between the current command and position feedback of the servo motor. For the vibration modal parameters of the mechanical structure, the vibration acceleration of the motion platform is analyzed in real time to extract the first and second dominant vibration frequencies, damping ratio and mode participation factor under the current motion state; For thermal deformation parameters, based on the temperature distribution on the linear motor stator, guide rail and frame, and combined with the pre-calibrated transfer matrix between the thermal expansion coefficient and structural deformation, the platform positioning reference drift caused by thermal deformation is calculated in real time. For the load disturbance parameters, the temporal correlation between the process force of the welding head and the motion trajectory command is analyzed, and the load disturbance components caused by bonding pressure and ultrasonic vibration reaction force are separated and quantified in real time.

5. The method for dynamic compensation of displacement error of the motion platform of a high-speed wire bonding device according to claim 4, characterized in that, The method for constructing the multi-error-source coupled prediction algorithm includes: A servo tracking error sub-model is established, with motion commands and dynamic parameters of the servo system as inputs, and the output being the predicted tracking error caused by servo response lag and disturbances. A vibration-induced error sub-model is established, whose inputs are the acceleration / jump command of the motion platform and the vibration mode parameters of the mechanical structure, and whose output is the predicted value of displacement oscillation caused by structural vibration. A thermal deformation error sub-model is established, with the temperature distribution and thermal deformation parameters as inputs and the predicted values ​​of thermal expansion / contraction of the platform structure caused by the temperature gradient as outputs. A load disturbance error sub-model is established, with the process force and load disturbance parameters as inputs and the predicted value of additional platform deformation caused by the dynamic force of the bonding process as output. The predicted tracking error, displacement oscillation, thermal expansion / contraction, and additional deformation are vector-superimposed and fused using a preset time-varying weighting matrix to output the predicted displacement deviation.

6. The method for dynamic compensation of displacement error of the motion platform of a high-speed wire bonding device according to claim 1, characterized in that, The position accuracy tolerance range includes: Different precision levels are preset for different bonding process steps; Based on the currently executing step, the corresponding precision level is invoked and converted into a tolerance range centered on the target position with upper and lower limits; Specifically, for the first solder joint of a spherical bond, the width of the tolerance interval is smaller than that for the second solder joint of a wedge bond.

7. The method for dynamic compensation of displacement error of the motion platform of a high-speed wire bonding device according to claim 1, characterized in that, The collaborative compensation command includes at least one of the following: trajectory feedforward correction for the motion controller, vibration suppression for the active damping device, and parameter adjustment for the force-position control loop.

8. The method for dynamic compensation of displacement error of the motion platform of high-speed wire bonding equipment according to claim 2, characterized in that, The position feedback is acquired using a dual-source synchronous acquisition and real-time calibration method with a grating ruler displacement sensor and a laser interferometer. The laser interferometer is symmetrically arranged on the outer ends of the crossbeams on both sides of the motion platform gantry mechanism, and the measuring optical axis is collinear with the measuring axis of the grating ruler.

9. The method for dynamic compensation of displacement error of the motion platform of a high-speed wire bonding device according to claim 2, characterized in that, The vibration acceleration is acquired using a piezoelectric triaxial accelerometer, which is connected to the platform via a flexible connection assembly. The flexible connection assembly includes a silicone damping pad and a metal fixing plate, which are used to isolate the transmission of non-target vibrations.

10. A dynamic compensation device for displacement error of a motion platform in a high-speed wire bonding device, characterized in that, include: The real-time acquisition module is used to collect the status parameters of the motion platform in real time and obtain multi-dimensional real-time monitoring data. The parallel identification module is used to independently estimate the platform's real-time characteristic parameters based on the multi-dimensional real-time monitoring data through multiple preset identification algorithms executed in parallel. The multi-error-source prediction module is used to input the real-time characteristic parameters of the platform into a pre-stored multi-error-source coupled prediction algorithm to calculate the predicted displacement deviation of the motion platform within a predetermined time window in the future. The tolerance range determination module is used to determine the corresponding position accuracy tolerance range based on the positioning accuracy requirements of the current bonding process step. The optimization problem construction module is used to establish a multi-objective optimization problem with minimizing the difference between the predicted displacement deviation and the position accuracy tolerance range as the primary optimization objective and minimizing the energy consumption and mechanical impact of the compensation action as the secondary optimization objectives. The control optimization module is used to solve the multi-objective optimization problem in the rolling time domain under the constraint of satisfying the physical limits of the motion actuator, generate cooperative compensation instructions in real time, and control the actuator to execute the instructions.

Images