Wafer component driving detection method and system based on perforated plate and multiple sensors

Abstract

The invention relates to the technical field of semiconductor manufacturing equipment detection, and discloses a wafer component driving detection method and system based on a perforated plate and multiple sensors, and the method comprises the steps: determining a detection strategy for the type of a to-be-detected component, and initializing the working mode of a sensor assembly; carrying out cooperative calibration on the sensor assembly; marking an initial position in the detection coordinate system; starting a driving part to act based on the detection strategy, and acquiring action data of the to-be-detected part by a sensor assembly; fusing the collected motion data, and reconstructing a three-dimensional motion track; and determining a precision index of the to-be-detected component based on the deviation. Standardized installation and positioning of the sensors and the driving parts are achieved through the porous plate, and the problems that a traditional detection method is low in coverage degree and difficult in data space alignment are solved by combining the multi-sensor cooperative calibration and data fusion technology. The method has the advantages that the detection precision and coverage are improved, multi-sensor data space alignment is realized, and the dynamic cooperative detection capability is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor manufacturing equipment detection, and in particular to a wafer component drive detection method and system based on a multi-hole plate and multiple sensors. Background Art

[0002] Wafer processing equipment is a core component of semiconductor manufacturing. It accomplishes complex processes such as wafer transfer, positioning, etching, and deposition through the coordinated operation of multiple precision components. Due to the inherently thin and brittle nature of wafers, typically measuring hundreds of microns thick, the accuracy of the equipment's actuators can easily lead to uneven force distribution and localized stress concentrations on the wafer, potentially leading to wafer breakage or microcracks, significantly reducing product yield. Prior art maintenance of equipment accuracy typically utilizes a selective component inspection strategy. Based on empirical experience or historical failure data, only a few high-risk components are independently calibrated before equipment startup. However, this approach has significant limitations. The stability of the wafer processing process relies on the timing and spatial coupling of multiple component movements. Using a single sensor cannot capture the multi-dimensional nature of component movements, resulting in low detection accuracy and coverage of wafer processing components. Furthermore, existing inspection systems lack a unified reference platform, leading to imprecise sensor placement and difficulty in spatially aligning the data collected by each sensor, further impacting the reliability of inspection results. At the same time, traditional methods cannot achieve dynamic coordination between the driving process and the detection process, making it difficult to accurately capture the actual motion state of components under complex working conditions. To address the above issues, existing technologies urgently need to be improved. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention aims to provide a wafer component drive detection method based on a multi-well plate and multiple sensors. This method has the advantages of improving detection accuracy and coverage, achieving spatial alignment of multi-sensor data, and enhancing dynamic collaborative detection capabilities.

[0004] In order to achieve the above object, the present invention provides the following technical solutions: A wafer component drive detection method based on a porous plate and multiple sensors includes: fixing the component to be detected on the porous plate through a mounting base, and electrically connecting the driving component to the component to be detected; determining a detection strategy based on the type of the component to be detected, and initializing the working mode of the corresponding sensor component based on the detection strategy; collaboratively calibrating the sensor component; marking the initial positions of the installation position of the component to be detected and the position of the sensor after calibration in the detection coordinate system; starting the action of the driving component based on the detection strategy, the driving component drives the component to be detected to move, and at the same time, the sensor component collects the action data of the component to be detected respectively; performing data fusion on the collected action data, and reconstructing the three-dimensional motion trajectory of the component to be detected; calculating the deviation between the three-dimensional motion trajectory and the standard trajectory, and determining the accuracy index of the component to be detected.

[0005] In the present invention, preferably, the porous plate includes n-level through holes in the center, n+i-level through holes on the periphery, 30°, 45°, 60° oblique holes and threaded holes distributed in an array at equal intervals.

[0006] In the present invention, preferably, the detection strategy includes a driving process and process parameters. The driving process is determined based on the actual operation process of the component to be detected. The driving process includes linear motion, steering motion, clamping motion, and pressure motion driving in sequence. The process parameters include corresponding action parameters of each driving process. The linear motion parameters include motion distance, speed, acceleration, and action time; the steering motion parameters include rotation angle, angular velocity, action time, and rotation axis offset; the clamping motion parameters include clamping stroke, clamping force, action time, and clamping parallelism; the pressure motion parameters include pressure value, pressing depth, force control response time, and holding time.

[0007] In the present invention, preferably, the sensor assembly includes a vision array sensor, a strain sensor, a laser interferometer, a capacitance sensor and a laser displacement sensor.

[0008] In the present invention, preferably, the collaborative calibration of the sensor assembly specifically includes static calibration: Start each corresponding sensor in the sensor assembly and restore it to its initial state. In the initial state, each sensor collects initial data from the standard points. The initial data collected by the visual array sensor is matched with the standard points in the overlapping area of ​​the multi-camera field of view to establish a conversion relationship between each camera coordinate system and the multi-hole plate reference coordinate system. The laser interferometer calibrates the beam steering error through the angle calibration block of the oblique hole, and uses the standard points to calibrate the orthogonality of the X / Y / Z axis optical path.

[0009] In the present invention, preferably, the collaborative calibration of the sensor assembly also includes linkage calibration: based on the measurement data of the reference array, the error of each sensor and the influence coefficient of the sensor error on other sensors are calculated to form a coupling error matrix. During the calibration process, based on the coupling error matrix, the multiple sensors in the sensor matrix are linkage-adjusted.

[0010] In the present invention, preferably, the collaborative calibration of the sensor assembly further includes dynamic calibration: When the sensor collects different motion data, the error prediction model outputs the current sensor error compensation value according to the parameters of the current driving process, realizing real-time calibration during motion.

[0011] In the present invention, preferably, the data fusion includes: based on weighted Kalman filtering, the same state data collected by the laser interferometer, capacitive sensor and laser displacement sensor are fused according to certain weights, and the displacement data of the wafer component is obtained after fusion; based on DS evidence theory, the visual sensor and strain sensor data are fused, and the end pressure data of the wafer component is obtained after fusion.

[0012] In the present invention, preferably, the specific implementation steps of the trajectory generation and comparison module include: extracting corresponding quantitative data of wafer components based on the working conditions corresponding to the driving strategy and the data fused based on the data fusion strategy; generating a three-dimensional motion trajectory by mathematical modeling based on the quantitative data in the detection coordinate system; calculating the deviation between the three-dimensional motion trajectory and the standard trajectory to obtain a deviation index.

[0013] The invention discloses a wafer component drive detection system based on a porous plate and multiple sensors, comprising: a test frame, wherein the test frame comprises two side plates and a porous plate, wherein the porous plate is fixed on the side plates, and a top plate is further provided on the top of the porous plate; a driving component, wherein the driving component comprises a power unit and a transmission unit, is fixedly arranged on the test frame, and is used to drive the wafer component to move; a sensor assembly, wherein the sensor assembly comprises a visual array sensor, a laser interferometer, a capacitive sensor, a laser displacement sensor, and a strain sensor; a control unit, wherein the control unit is communicatively connected with the driving component and the sensor assembly, and the control unit comprises a data processing module, a trajectory generation and comparison module, a decision module, and a driving module, wherein the data processing module is used to receive the collected data of the sensor assembly and perform data preprocessing and data fusion on the collected data, and the trajectory generation and comparison module performs motion trajectory restoration and comparison on the fused data to obtain a trajectory deviation index; the decision module determines the driving strategy of the driving component and the acquisition strategy of the sensor assembly according to the type of the wafer component, and sends the driving strategy and sensor strategy to the driving module, and the driving module controls the driving component and the sensor assembly to work according to the corresponding strategy.

[0014] In the present invention, the visual array sensor preferably includes: a front-view camera, mounted on the top plate at the top of the perforated plate, with its lens facing vertically downward; side-view cameras, two of which are mounted on the left and right side plates of the perforated plate, with their lenses facing horizontally; and an oblique camera, mounted behind the perforated plate through a 30° oblique hole, with its lens at a 45° angle to the horizontal plane, to capture the contact area between the actuator and the simulated wafer. N+i level through-holes are provided at the four corners of the perforated plate to capture the movement of the wafer component from different angles.

[0015] In the present invention, preferably, the sensor assembly includes a strain sensor, and the strain sensor is adhered to the surface of the wafer component in contact with the wafer.

[0016] In the present invention, preferably, the laser head of the laser interferometer is fixed on the side plate connected to the porous plate through an adjustable bracket, its light beam is parallel to the X-axis direction of the porous plate, and the reflector is arranged on the end effector of the wafer component; the steering mirror of the laser interferometer is installed on the edge of the porous plate through a 45° inclined hole, the light beam is turned parallel to the Y-axis direction of the porous plate, a vertical optical path bracket is installed on the right side of the side plate, the light beam is vertically upward, and the corresponding reflector is fixed at the bottom of the end effector.

[0017] In the present invention, preferably, the probes of the capacitive sensors are vertically upward and are evenly spaced at eight equal points of the same circle; the laser displacement sensors are respectively installed at the end positions of the X-axis and Y-axis of the porous plate; the strain sensor is pasted on the surface of the wafer component in contact with the wafer; and multiple capacitive sensors are arranged in a ring on the porous plate.

[0018] In the present invention, preferably, the power unit includes a servo motor and a precision cylinder, and the servo motor and the precision cylinder are fixed to the side of the porous plate through a U-shaped frame, the servo motor is used to drive the wafer component to perform steering movement, and the precision cylinder is used to drive the wafer component to perform linear movement; The transmission unit comprises a magnetic coupling transmission, which is arranged at the bottom of the porous plate and connected to the wafer component through the oblique holes of the porous plate.

[0019] The data processing module is provided with a pre-processing strategy, which includes: Time alignment: Unify the data collected by multiple sensors to the same sampling rate, and associate multiple data collected at the same time and store them as a group, with each group corresponding to the sampling time; Spatial unification: Map the data collected by the sensor to the detection coordinate system based on the installation position of the sensor on the porous plate; Noise filtering: Filter high-frequency vibration noise through wavelet transform.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention realizes the standardized installation and positioning of sensors and driving components through a porous plate, and combines multi-sensor collaborative calibration and data fusion technology to solve the problems of low coverage and difficult data space alignment of traditional detection methods. It has the advantages of improving detection accuracy and coverage, realizing multi-sensor data space alignment, and enhancing dynamic collaborative detection capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the wafer component drive detection method based on a multi-hole plate and multiple sensors according to the present invention.

[0022] Figure 2 This is a schematic structural diagram of the test stand described in the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may be a central component. When a component is considered to be "disposed on" another component, it may be directly disposed on the other component or there may be a central component. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0026] See Figure 1A preferred embodiment of the present invention provides a wafer component drive detection method based on a porous plate and multiple sensors, wherein the component to be detected is fixed on the porous plate 1 through a mounting base, and the driving component is electrically connected to the component to be detected; a detection strategy is determined based on the type of the component to be detected, and the working mode of the corresponding sensor component is initialized based on the detection strategy; the sensor component is collaboratively calibrated; the installation position of the component to be detected and the position of the sensor after calibration are marked as initial positions in the detection coordinate system; the driving component is started to act based on the detection strategy, and the driving component drives the component to be detected to act, and at the same time, the sensor component collects the action data of the component to be detected respectively; data fusion is performed on the collected action data to reconstruct the three-dimensional motion trajectory of the component to be detected; the deviation between the three-dimensional motion trajectory and the standard trajectory is calculated to determine the accuracy index of the component to be detected.

[0027] The porous plate 1 refers to a positioning substrate with a multi-level through-hole array. Specifically, it can be made of aluminum alloy material to form a composite structure with a central n-level through-hole 2 and surrounding n+i-level through-holes 3. Different aperture combinations can be used to adapt to the installation requirements of various components. The detection strategy refers to a set of drive parameters customized according to the motion characteristics of the component. Specifically, it can match the component type through a preset parameter database and dynamically generate control instructions including linear motion speed and rotation angle thresholds. Collaborative calibration is the process of eliminating errors in multi-sensor systems. It mainly establishes a coordinate system transformation relationship through static calibration, and then combines dynamic error compensation to achieve sensor data synchronization. Data fusion is mainly for data collected by multi-source heterogeneous sensors.

[0028] Specifically, the component to be inspected is fixed to the porous plate 1 through an adjustable mounting base, and the through-hole array is used to achieve precise adjustment of the mounting position. The driving component outputs a control signal according to the detection strategy to reproduce the complex motion of the component under actual working conditions. After collaborative calibration, the multi-sensor assembly collects displacement, stress, and deformation data from different spatial dimensions. After the collected data is aligned in time and unified in space, a three-dimensional motion trajectory model is generated through multi-algorithm fusion. The model is spatially matched with the standard trajectory, and the trajectory offset is quantitatively calculated as the basis for accuracy assessment. The porous plate 1 is used to achieve rapid reconstruction of the detection tooling, and the collaborative calibration of multiple sensors is combined to eliminate system errors, effectively improving the consistency of the detection environment. Dynamic data fusion technology breaks through the measurement dimension limitations of a single sensor, and can simultaneously obtain multiple physical quantity data such as displacement, stress, and deformation, more accurately reflecting the actual motion state of the component.

[0029] The method of this application enables full-dimensional detection of the motion characteristics of wafer processing components, overcoming the misjudgment problem caused by the single measurement parameter of traditional methods. The multi-sensor collaborative calibration mechanism ensures the uniformity of the spatial benchmark of the detection data and avoids conflicts in multi-source data. The three-dimensional motion trajectory reconstruction technology accurately restores the complex motion state of the components, providing reliable data support for accuracy assessment. The dynamic data fusion method effectively improves the environmental adaptability of the detection system and meets the accuracy detection requirements under different working conditions.

[0030] See Figure 2 In this embodiment, the porous plate 1 includes a central n-level through hole 2, a surrounding n+i-level through hole 3, 30°, 45°, and 60° inclined holes 4, and a threaded hole 5 distributed in an array at equal intervals.

[0031] The central n-level through-holes 2 refer to an equally spaced two-dimensional array of holes formed with the geometric center of the porous plate 1 as the origin. Specifically, this can be achieved using a rectangular grid distribution with a spacing of 5 mm. This serves to provide a reference positioning point for the sensor assembly, ensuring uniform coverage of the core measurement area. The peripheral n+i-level through-holes 3 refer to an extended array with more holes set in the edge area of ​​the porous plate 1 than in the center area. Specifically, this can be achieved using a layout with three holes added to each side. This is used to adapt to the detection requirements of wafer components of different sizes and enhance the adjustability of the peripheral sensor. The 30°, 45°, and 60° oblique holes 4 refer to channel structures that penetrate the porous plate 1 at specific inclination angles. Specifically, this can be achieved using an installation method with a built-in angle calibration block. This is used to guide optical measurement paths in non-orthogonal directions and obtain the three-dimensional motion characteristics of wafer components. The threaded holes 5 refer to fixed holes with an internal thread structure. Specifically, this can be achieved using the M3 standard thread specification. This allows for rapid disassembly and assembly of the sensor bracket and fine-tuning of the position, forming a dynamically reconfigurable detection platform.

[0032] Specifically, the central n-level through holes 2 form a reference positioning network through an equidistant array, providing a stable coordinate reference system for the visual array sensor, while providing flexible options for the installation of wafer components and sensor components. The n+i-level through holes 3 around the periphery expand the number of hole positions, so that the laser displacement sensor 12 can adjust the installation position according to the size of the wafer component to avoid measurement blind spots. The 30°, 45°, and 60° oblique holes 4 are respectively equipped with angle calibration blocks, so that the light beam of the laser interferometer 10 can be turned at non-orthogonal angles to accurately capture the rotational motion trajectory of the wafer component. The threaded holes 5 cooperate with the adjustable bracket to facilitate the installation of various sensors, drive components and wafer components to be measured, as well as to realize the rapid positioning and spacing adjustment of the capacitive sensor 11 in the annular array. Through the synergistic effect of the above-mentioned hole structure, the porous plate 1 can support the synchronous data acquisition of multiple types of sensors in different spatial dimensions, eliminating the problem of loss of three-dimensional motion features caused by the traditional single-plane layout. The composite hole structure design enables the sensor to collect data synchronously from orthogonal and non-orthogonal directions. For example, the 45° oblique hole 4 is used to guide the laser interferometer 10 to measure the axial offset of the wafer holder. Combined with the visual positioning data of the center through-hole array, accurate reconstruction of the three-dimensional motion trajectory can be achieved.

[0033] This application solves the problem of incomplete multi-dimensional data collection caused by limited sensor layout. It realizes reference positioning through a central through-hole array, and the extended through-holes on all sides are adapted to the installation of peripheral sensors. The oblique hole 4 supports three-dimensional motion measurement, and the threaded hole 5 provides dynamic adjustment capability, forming a detection platform covering the full-space motion characteristics of wafer components, which significantly improves the spatial coverage of the detection data and the accuracy of motion trajectory restoration.

[0034] In this embodiment, the detection strategy includes a driving process and process parameters. The driving process is determined based on the actual operation process of the component to be detected. The driving process includes linear motion, steering motion, clamping motion, and pressure motion driving in sequence. The process parameters include the corresponding action parameters of each driving process. The linear motion parameters include motion distance, speed, acceleration, and action time; the steering motion parameters include rotation angle, angular velocity, action time, and rotation axis offset; the clamping motion parameters include clamping stroke, clamping force, action time, and clamping parallelism; the pressure motion parameters include pressure value, pressing depth, force control response time, and holding time.

[0035] The driving process refers to a standardized sequence of motions decomposed from the actual motion pattern of the component to be inspected in the wafer processing equipment. In one embodiment, this is achieved through the coordinated control of a servo motor and a pneumatic cylinder to ensure that the inspection action is consistent with the actual working conditions. Process parameters refer to quantitative indicators corresponding to each type of driving action. In one embodiment, they are dynamically adjusted through sensor feedback and preset thresholds. Acceleration parameters can be calibrated in real time using encoder data. The combination of motion distance and speed in the linear motion parameters is used to characterize displacement accuracy. Acceleration and motion time are associated with dynamic response characteristics. Trapezoidal speed curve control can avoid vibration caused by sudden stops. The rotation axis offset in the steering motion parameters is used to capture mechanical assembly errors. Specifically, the deviation of the rotation center position is measured using a laser interferometer 10. The clamping parallelism in the clamping motion parameters is used to evaluate the uniformity of the contact surface. A dual-axis force sensor is used to monitor the clamping force distribution. The force control response time in the pressure motion parameters is used to evaluate contact stability. The pressure output can be adjusted by a PID algorithm.

[0036] Specifically, the actuation process is broken down into four basic motions to match the actual operating conditions of wafer components. When the inspection component is a robotic arm, the manipulator's transmission process corresponds to a combination of linear and steering motions, while vacuum chuck positioning corresponds to a combination of clamping and pressure motions. A multidimensional parameter system for each motion comprehensively characterizes the motion state from the perspectives of spatial displacement, time series, and mechanical properties. The rotation angle and angular velocity parameters in the steering motion simultaneously monitor the static accuracy and dynamic smoothness of the rotation trajectory. The combination of clamping stroke and clamping force parameters identifies jaw loosening due to wear, and the clamping parallelism parameter calculates flatness error using multi-point strain data. The depression depth and dwell time parameters of the pressure motion evaluate the deformation recovery characteristics of the contact surface, and the force control response time parameter reflects the actuator's ability to suppress pressure fluctuations. Through synergistic analysis of these parameters, complex deviations that cannot be detected by a single sensor can be identified, such as trajectory drift caused by the robot's rotation axis offset and acceleration anomalies in this embodiment. By defining four actuation processes and their corresponding multidimensional parameter systems, this solution achieves comprehensive coverage of the wafer component's motion characteristics. For example, while existing technologies only measure the rotation angle for steering motion, this solution also incorporates the axis offset parameter to detect non-concentric rotation caused by bearing wear. For pressure motion detection, existing technologies focus solely on steady-state pressure values. This solution, by combining the parameters of pressure depth and force-controlled response time, can identify the impact load at the moment of contact.

[0037] This application solves the accuracy and coverage defects caused by insufficient detection dimensions. By strictly matching the driving process with the actual working conditions, the deviation between the simulated action and the real action is avoided; through multi-dimensional parameter definition, the synchronous monitoring of spatial displacement, dynamic response, mechanical distribution and other characteristics is achieved. For example, in the clamping motion detection, the clamping parallelism parameter can identify the plane tilt of 0.1 mm, thereby warning of the risk of uneven force on the wafer; in the pressure movement, the holding time parameter can evaluate the creep characteristics of the contact surface and avoid positioning failure due to material relaxation. This solution enables the detection data to accurately reflect the real state of the component under complex action, providing an accurate benchmark for multi-sensor collaborative calibration.

[0038] In this embodiment, the sensor assembly includes a vision array sensor, a strain sensor, a laser interferometer 10 , a capacitive sensor 11 and a laser displacement sensor 12 .

[0039] In one specific embodiment, the visual array sensor includes: a front-view camera 8, which is mounted on the top plate 7 at the upper end of the porous plate 1, with its lens facing vertically downward; a side-view camera 9, with two side-view cameras 9 mounted on the side plates 6 on the left and right sides of the porous plate 1, respectively, with their lenses facing horizontally; an oblique camera, which is mounted behind the porous plate 1 through a 30° oblique hole 4, with its lens at a 45° angle to the horizontal plane, and photographs the contact area between the actuator and the simulated wafer to capture the spatial posture changes and contact area deformation characteristics during the movement of the wafer component. The laser head of the laser interferometer 10 is fixed to the side plate 6 connected to the porous plate 1 through an adjustable bracket, with its light beam parallel to the X-axis direction of the porous plate 1, and a reflector is set on the end effector of the wafer component; the steering mirror of the laser interferometer 10 is mounted on the edge of the porous plate 1 through a 45° oblique hole 4, and the light beam is turned parallel to the Y-axis direction of the porous plate 1. A vertical optical path bracket is installed on the right side of the side plate 6, and the light beam is vertically upward. The corresponding reflector is fixed to the bottom of the end effector. The capacitive sensors 11, with their probes pointing vertically upward and spaced evenly at eight points on a circle, are used to detect minute changes in the gap between the wafer assembly and the baseplate during movement. Laser displacement sensors 12 are mounted at the X-axis and Y-axis endpoints of the porous plate 1 to supplement displacement information in specific directions. Strain sensors, which can be implemented as resistance strain gauges or fiber Bragg grating sensors, are attached to the contact surface between the wafer assembly and the wafer and are used to monitor the micromechanical distribution at the contact surface between the wafer and the actuator.

[0040] Specifically, by constructing a multimodal sensor combination system, the visual array sensor captures the macroscopic motion trajectory of the wafer component from different angles, the strain sensor simultaneously monitors the stress distribution of the contact surface, the laser interferometer 10 and the capacitive sensor 11 respectively obtain high-precision displacement data and gap changes, and the laser displacement sensor 12 performs redundant measurements of the key axial displacement. The various sensors complement each other in terms of spatial distribution and measurement dimensions. The combination of the visual sensor and the strain sensor can simultaneously obtain the motion trajectory and contact stress, and the combination of the laser interferometer 10 and the capacitive sensor 11 can simultaneously monitor the displacement and gap changes. The collaborative configuration of multi-source heterogeneous sensors covers multi-dimensional physical quantities such as displacement, angle, stress, and gap, providing a complete raw data set for subsequent data fusion. Through the collaborative operation of multiple sensors, this solution not only covers kinematic parameters but also achieves the synchronous monitoring of mechanical properties and gap status, overcoming the limitations of single sensors in the frequency domain, spatial domain, and physical quantity type.

[0041] Through the above technical solution, this application effectively improves the dimensional coverage of wafer component motion detection, enabling the simultaneous acquisition of multi-dimensional data such as displacement, angle, stress, and gap, providing comprehensive data support for subsequent accuracy assessment. This solution addresses the inaccuracy problem of traditional detection methods due to the lack of data dimensions, and is particularly suitable for precision component detection scenarios that require monitoring the coupling relationship between contact mechanical properties and motion trajectory.

[0042] In this embodiment, the collaborative calibration of the sensor assembly specifically includes static calibration: starting the corresponding sensors in the sensor assembly and restoring them to the initial state first; in the initial state, each sensor collects initial data for the standard point; the initial data collected by the visual array sensor is matched with the standard points in the overlapping area of ​​the multi-camera field of view, and a conversion relationship between each camera coordinate system and the reference coordinate system of the multi-hole plate 1 is established; the laser interferometer 10 calibrates the light beam steering error through the angle calibration block of the oblique hole 4, and uses the standard point to calibrate the orthogonality of the X / Y / Z axis optical path.

[0043] Among them, restoring to the initial state refers to eliminating the zero drift error through the sensor reset operation, which can be achieved specifically by sending a calibration instruction or triggering a hardware reset circuit, so that all sensors return to the factory preset reference state. The standard point is set on the porous plate 1, and the central n-level through hole 2 on the porous plate 1 is used as a standard point to provide a unified spatial reference benchmark for multiple sensors. Standard point matching in the overlapping area of ​​the field of view of multiple cameras refers to coordinate alignment using standard points observed by multiple cameras. Specifically, an image feature extraction algorithm can be used to calculate the coordinates of the same-name points under different camera perspectives, and the multi-perspective data space can be unified through a coordinate transformation matrix. The angle calibration block of the oblique hole 4 refers to a physical calibration device with a precise angle scale. Specifically, a reflector bracket installed in the oblique hole 4 of the porous plate 1, such as a mirror holder with a 45° angle scale, can be used to compensate for the beam steering error of the laser interferometer 10.

[0044] Specifically, the static calibration process is divided into four levels: first, each sensor is returned to its initial reference state through hardware reset to eliminate the initial error caused by zero drift or environmental interference; then, the preset standard points on the porous plate 1 are used to trigger all sensors to synchronously collect initial data and establish a physical correlation benchmark for multi-source data; for the visual array sensor, by extracting the target coordinates of the overlapping area of ​​the field of view of multiple cameras, the coordinate system of each camera is mapped to the reference coordinate system of the porous plate 1 using a coordinate transformation algorithm to eliminate the spatial deviation caused by the difference in installation angle; for the laser interferometer 10, the beam steering angle is adjusted in combination with the angle calibration block in the oblique hole 4, and the orthogonality of the X / Y / Z axis optical path is verified through the standard point to compensate for the measurement deviation caused by optical path deflection or assembly error. By integrating the standard target point and the angle calibration block on the porous plate 1, the synchronous calibration of multiple sensors under a unified physical benchmark is realized, solving the coupling problem of cross-sensor coordinate system conversion error and optical path steering error.

[0045] Through the above technical solution, the present application effectively eliminates measurement deviations caused by inconsistent initial states of multiple sensors, establishes a unified spatial coordinate system across sensors, and compensates for the steering angle error of the optical path of the laser interferometer 10. This enables precise alignment of the multi-view data of the visual array sensor, ensures the orthogonality of the optical path of the laser interferometer 10, and provides a highly consistent measurement benchmark for subsequent multi-source data fusion, thereby improving the overall accuracy of wafer component motion detection.

[0046] In this embodiment, the collaborative calibration of the sensor components also includes linkage calibration. Based on the measurement data of the reference array, the error of each sensor and the influence coefficient of the sensor error on other sensors are calculated to form a coupling error matrix. During the calibration process, based on the coupling error matrix, multiple sensors in the sensor matrix are linked and adjusted.

[0047] The measurement data of the reference array refers to a multi-sensor synchronous measurement data set obtained through preset spatial reference points. Specifically, it can be implemented using standard point coordinates generated by a high-precision calibration plate or a laser tracker to establish a multi-sensor error correlation relationship.

[0048] The coupling error matrix refers to a mathematical expression model that reflects the mutual influence relationship between the errors of each sensor. Specifically, it can be implemented by calculating the error transfer coefficient using multivariate regression analysis or principal component analysis, and is used to quantify the degree of cross-interference between sensors.

[0049] Specifically, in this embodiment, a visual array sensor, a laser interferometer 10, a capacitive sensor 11, a laser displacement sensor 12, and a strain sensor are used to collect data from the robotic arm. When collecting data during the movement of the robotic arm, during the linkage calibration process, the robotic arm is controlled to drive the reference target to move along a preset trajectory and collect data synchronously: Static data acquisition: The robotic arm remains in 30 fixed positions for 5 seconds. The forward-view camera 8 collects the coordinates of grid feature points, the prism pixel positions of the side-view camera 9, and the deformation image of the contact area of ​​the oblique camera. The laser interferometer 10 collects the displacement of the robotic arm's end in the X, Y, and Z axes. The eight probes of the capacitive sensor 11 collect the gap between the bottom of the robotic arm and the porous plate 1. The laser displacement sensor 12 collects the end displacement in the X and Y axes. The strain sensor collects the strain value of the suction cup contact area at the end of the robotic arm. The robotic arm performs sinusoidal motion along the X / Y / Z axes at a speed of 50 mm / s for 10 cycles, and the dynamic measurement values ​​of each sensor are recorded as time series data. A coupling error matrix is ​​constructed based on the motion, and the measurement values ​​of the laser interferometer 10 are used as a reference to calculate the error and coupling coefficient of each sensor: Among them, the visual sensor: the grid point coordinate error Δu of the front view camera 8 = measured pixel - true value pixel, the prism positioning error Δp ​​of the side view camera 9 = measured position - positioning value of the laser interferometer 10; the gap error Δd of the capacitive sensor 11 i = measured value - actual height of target boss; displacement error ΔLx of laser displacement sensor 12 = measured value - true value of laser interferometer 10X axis, ΔLy is similar; error Δε of strain sensor i= measured value - known strain on the calibration piece. Analyzing the inter-sensor error correlation, the installation angle error of the front-view camera 8 will cause a change in the prism positioning error of the side-view camera 9. In this embodiment, every 0.1-degree angle error corresponds to a ±2μm positioning error, with an influence coefficient k1 = 2μm / 0.1°. The temperature drift of the capacitive sensor 11, with every 1°C change, results in a ±0.05μm gap error, which simultaneously affects the Z-axis measurement of the laser displacement sensor 12. Every 1°C temperature change results in a ±0.03μm displacement error, with an influence coefficient k2 = 0.03μm / 0.05μm. Contact pressure errors in the strain sensor will cause errors in the deformation image resolution of the oblique camera, with an influence coefficient k3 = 1.5 pixels / 0.1N.

[0050] Taking each sensor as a reference, an 8×8 matrix is ​​formed. From the upper left to the right and bottom, it is the front view camera 8, the left view camera 9, the right view camera 9, the oblique camera, the laser interferometer 10, the capacitive sensor 11, the laser displacement sensor 12, and the strain sensor. The matrix elements are the above-mentioned influence coefficients.

[0051] Based on the coupling error matrix, collaborative parameter corrections were performed on multiple sensors. The extrinsic parameter matrix of the side-view camera 9 was adjusted based on the coupling coefficient k1 between the front-view camera 8 and the side-view camera 9. The installation angles of the left and right-view cameras 9 were corrected by -0.05°. The distortion coefficient of the front-view camera 8 was simultaneously updated from k1=0.02 to k1=0.005, reducing the positioning error between the two cameras from ±5μm to ±1.2μm. Combined with the coupling coefficient k3 between the oblique camera and the strain sensor, the image scaling factor of the oblique camera was adjusted from 1.02 to 0.998. The sensitivity coefficient of the strain sensor was also calibrated from 1.28 to 1.27, reducing the deformation resolution error by 60%.

[0052] In view of the annular distribution characteristics of the capacitance sensor 11, based on the influence of the adjacent capacitance in the coupling error matrix, the error d1 affects d2 by ±0.03μm. At the same time, the zero offsets of the eight capacitance sensors 11 are adjusted: d1 correction + 0.02μm, d2 correction - 0.01μm, d3 correction + 0.018μm, d4 correction -0.005μm, d5 correction + 0.012μm, d6 correction -0.015μm, d7: + 0.008μm, d8 correction + 0.015μm, so that the gap measurement consistency is improved from ±0.2μm to ±0.08μm.

[0053] Incorporating the temperature coupling coefficient k2, a temperature compensation formula is implemented in laser displacement sensor 12: Lx Corrected Value = Lx Measured Value - 0.03 × (Average Capacitor Temperature Drift). This reduces the X / Y axis displacement error from ±3μm to ±0.8μm. Using the Z-axis data from laser interferometer 10 as a reference, the Z-axis coordinate conversion parameters of forward-facing camera 8 are corrected from 0.05mm per pixel to 0.0498mm per pixel. Simultaneously, the 45° angle parameter of the oblique camera is adjusted to 44.98° to ensure multi-sensor consistency in spatial pose calculations.

[0054] In 20 new robot poses, the deviations between the measured values ​​of each sensor and the true values ​​of the laser interferometer 10 were: the positioning error of the vision sensor was ≤±1.5μm, the gap error of the capacitive sensor 11 was ≤±0.1μm, and the error of the strain sensor was ≤±5με. When the robot performed random trajectory motion, the error in the actuator pose calculation using multi-sensor fusion was reduced from ±8μm before calibration to ±2μm, and the time synchronization error in the deformation capture of the contact area was ≤0.5ms. By quantifying the mutual influence between sensors through a coupling error matrix, the "compensation offset" caused by single-sensor calibration was avoided, improving the global consistency of the robot motion measurement by over 70%, meeting the sensor coordination requirements for high-precision wafer component manipulation.

[0055] Through the above technical solution, the present application solves the problem of reduced measurement accuracy caused by error coupling during the multi-sensor collaborative calibration process, ensuring the consistency of the wafer component motion trajectory detection data in the spatial coordinate system. For example, in the dynamic detection of the wafer robot end effector, the measurement deviation of the visual array sensor and the laser interferometer 10 are synchronously corrected, avoiding the distortion of the three-dimensional trajectory reconstruction caused by the error transmission of a single sensor, thereby improving the reliability of the precision assessment of wafer processing equipment.

[0056] In this embodiment, when the sensor collects different motion data, the error prediction model outputs the current sensor error compensation value according to the parameters of the current driving process, thereby realizing real-time calibration during motion.

[0057] Among them, dynamic calibration refers to the real-time adjustment of the sensor error compensation value during the movement of the wafer component. Specifically, it can be achieved by using an adaptive algorithm based on the correlation between the drive parameters and the sensor error. By collecting the motion parameters of the drive component in real time and inputting the error prediction model, the compensation value is dynamically generated to offset the measurement deviation caused by environmental interference or mechanical vibration.

[0058] The error prediction model is a mapping between sensor errors and drive parameters established through machine learning or statistical modeling. Specifically, an improved LSTM-linear hybrid model is employed. For highly dynamic parameters such as acceleration and velocity, a layer of LSTM with 16 neurons captures temporal correlations. Static parameters such as temperature and humidity are modeled using a linear regression module. For example, for every 1°C increase in temperature, the error of the capacitive sensor 11 increases by 0.02μm. The output layer ensures the linkage of multi-sensor compensation values ​​by coupling the weights of the error matrix. The LSTM-linear hybrid model is trained using pre-trained training data. When the drive component performs linear or steering motion, the system acquires the current motion parameters in real time. The robot arm's X / Y / Z axis velocity, acceleration, motor current, temperature, humidity, the average error of the capacitive sensor 11 over the previous three sampling cycles, the error of the laser displacement sensor 12, and the visual positioning error are input into the LSTM-linear hybrid model. After calculation, the LSTM-linear hybrid model outputs real-time error compensation values ​​for the eight capacitive sensors 11, X / Y axis compensation values ​​for the laser displacement sensor 12, and pixel coordinate corrections for the visual array sensor. As a result, the sensor can continuously correct for accumulated errors caused by mechanical vibration or temperature changes during dynamic detection, ensuring that the measured data is consistent with the actual motion state of the wafer component. By dynamically linking drive parameters with sensor errors, this solution achieves real-time error compensation during motion, addressing the problem of sensor accuracy degradation caused by changes in the dynamic characteristics of the mechanical system.

[0059] Through the above technical solution, the present application can automatically correct sensor measurement errors during the continuous movement of wafer components, avoiding trajectory reconstruction deviations caused by error accumulation. For example, when a wafer robot performs high-speed grasping movements, dynamic calibration can effectively suppress the impact of vibration on the laser displacement sensor 12, thereby accurately capturing micron-level displacement changes of the end effector and preventing wafer positioning errors or uncontrolled clamping force caused by measurement distortion.

[0060] In this embodiment, data fusion includes fusing the same state data collected by the laser interferometer 10, the capacitive sensor 11 and the laser displacement sensor 12 according to weights based on weighted Kalman filtering to obtain displacement data, and fusing the visual sensor and strain sensor data based on DS evidence theory to obtain end pressure data.

[0061] Among them, weighted Kalman filtering is a method for optimizing the fusion of multi-source homogeneous data by dynamically adjusting the weight coefficients of each sensor. This can be achieved using an iterative update algorithm for the covariance matrix. By assigning weight coefficients through real-time evaluation of the measurement noise level of each sensor, high-frequency sampling data and contact measurement data complement each other. DS evidence theory is a method for cross-dimensional heterogeneous data fusion through confidence function modeling. This can be achieved by constructing a basic probability distribution function and combining evidence synthesis rules. Reliable decisions are made by eliminating data conflicts between sensors.

[0062] Specifically, in displacement detection scenarios, where the laser interferometer 10, capacitive sensor 11, and laser displacement sensor 12 have the same physical quantities but different measurement errors, a weighted Kalman filter is used to establish a state-space model. By dynamically correcting the predicted and observed values, the high-frequency characteristics of optical measurement are combined with the anti-interference characteristics of contact measurement to eliminate the impact of single-point measurement errors on overall trajectory restoration. To address the mismatch between the visual image features and the dimensionality of the strain sensor scalar data in pressure detection scenarios, a pressure state recognition framework is constructed to convert the deformation characteristics of the contact area captured by the visual sensor into spatially distributed confidence levels. This is then combined with the stress values ​​measured by the strain sensor for evidence synthesis to address misjudgments caused by local occlusion or temperature drift.

[0063] Compared with existing technologies, traditional methods rely on independent calibration of single sensors or simple weighted average fusion, which does not consider the error coupling relationship between optical sensors and contact sensors, nor does it address the dimensional differences between image data and physical quantity data. This solution uses a layered fusion mechanism to achieve dynamic error compensation in the homogeneous data layer and establish a cross-dimensional decision model in the heterogeneous data layer, forming a multi-dimensional data collaborative verification mechanism.

[0064] Through the above technical solution, this application effectively solves the problem of insufficient fusion accuracy of multi-sensor data due to differences in acquisition principles, reducing the restoration error of the three-dimensional motion trajectory of the wafer component to the micron level, and accurately identifying the pressure distribution state of the contact surface between the end effector and the wafer, avoiding the risk of wafer damage caused by local stress concentration.

[0065] In this embodiment, the specific implementation steps of the trajectory generation and comparison module include extracting the corresponding quantitative data of the wafer components based on the working conditions corresponding to the driving strategy and the data fused based on the data fusion strategy; generating a three-dimensional motion trajectory by mathematical modeling based on the quantitative data in the detection coordinate system; calculating the deviation between the three-dimensional motion trajectory and the standard trajectory to obtain a deviation index.

[0066] The detection coordinate system refers to the coordinate system used to unify the spatial reference of multiple sensors. This can be achieved by using a transformation matrix between the reference coordinate system of the multi-well plate 1 and the coordinate systems of each sensor, eliminating the impact of differences in sensor installation positions on the data spatial reference. Quantitative data refers to the displacement, angle, and acceleration parameters extracted from the fused data. Working condition analysis can be used to screen characteristic parameters that are highly correlated with the current drive process, avoiding redundant data processing. Mathematical modeling refers to the algorithm that converts discrete parameters into continuous trajectories. Specifically, a coupled rigid body kinematics and elastic deformation model can be used to reflect the overall motion characteristics of the component and capture micro-deformations of the end effector. The standard trajectory refers to the motion trajectory that wafer processing equipment should achieve under ideal conditions. This can be achieved by extracting spatiotemporal characteristic parameters from the equipment control log and establishing a parameterized equation. The deviation index refers to the overall deviation between the actual trajectory and the standard trajectory. This can be achieved by calculating the weighted root mean square error of spatial position, velocity, and acceleration to generate a quantitative evaluation metric.

[0067] Specifically, this technical solution achieves high-precision trajectory reconstruction through the dual constraints of working condition drive and data fusion. First, based on the working condition type corresponding to the driving strategy, quantitative parameters matching the working condition are screened from the fused data. For example, displacement and acceleration parameters are extracted for linear motion conditions, and angle and angular velocity parameters are extracted for rotation conditions. Subsequently, a mathematical model is established in the detection coordinate system, and the spatial reference differences of multiple sensors are eliminated through coordinate system mapping. The kinematic model is used to convert discrete parameters into continuous three-dimensional trajectories. Finally, the trajectory deviation index is used as a quantitative evaluation indicator. By calculating the comprehensive deviation between the actual trajectory and the standard trajectory in multiple dimensions such as spatial position, motion speed, and acceleration, a quantifiable accuracy evaluation system is formed. In the mathematical modeling stage, a rigid body kinematics and elastic deformation coupling model is used to reflect the overall motion characteristics of the component and capture the micro-deformation of the end effector, thereby improving the fidelity of trajectory reproduction.

[0068] In some specific embodiments, the mathematical modeling of the standard trajectory can establish a parameterized equation based on the spatiotemporal characteristic parameters extracted from the equipment control log. For example, the straight line trajectory is described by the quadratic equation of initial velocity and acceleration, and the arc trajectory is described by the trigonometric function equation of radius and angular velocity. The establishment of the detection coordinate system can be achieved by matching the multi-camera field of view of the visual array sensor, and the data collected by each sensor is uniformly mapped to the reference coordinate system of the porous plate 1. The calculation of the deviation index can be based on the weighted root mean square error algorithm, which assigns different weight coefficients to errors in different dimensions to reflect the degree of influence of the wafer component movement accuracy on the process stability. This solution effectively improves the fidelity of trajectory reconstruction by detecting the coordinate system-multi-sensor data space benchmark and introducing a rigid body and elastic deformation coupling model. In addition, traditional accuracy assessment relies heavily on manual experience and judgment, while this solution uses the deviation index to achieve quantitative calculation of multi-dimensional errors, providing an objective decision-making basis for wafer component maintenance.

[0069] Through the above technical solution, this application solves the problem of inaccurate 3D motion trajectory reconstruction after multi-sensor data fusion. By detecting coordinate system mapping and coupling models to eliminate spatial reference differences and micro-deformation interference, high-fidelity trajectory reproduction is achieved. At the same time, the deviation index based on multi-dimensional error calculation provides a quantitative basis for wafer component motion accuracy assessment, avoiding subjective errors caused by human experience and improving the credibility and operability of detection results.

[0070] Another preferred embodiment of the present invention provides a wafer component drive detection system based on a porous plate 1 and multiple sensors, comprising a test frame, a drive component, a sensor assembly, and a control unit. The test frame comprises two side panels 6 and a porous plate 1, which is fixed to the side panels 6 and topped with a top plate 7. The drive component comprises a power unit and a transmission unit, which are fixed to the test frame. The sensor assembly comprises a visual array sensor, a laser interferometer 10, a capacitive sensor 11, a laser displacement sensor 12, and a strain sensor. The control unit is communicatively connected to the drive component and the sensor assembly and comprises a data processing module, a trajectory generation and comparison module, a decision module, and a drive module.

[0071] Among them, the porous plate 1 refers to a mounting substrate with a multi-level through-hole array, which can specifically adopt a combination structure of central through-holes, edge oblique holes 4 and threaded holes 5 with equal spacing, which is used to provide a modular mounting interface for sensors and drive components. The power unit refers to an actuator that drives the movement of wafer components, which can specifically adopt a combination of servo motors and precision cylinders, and realize precise control of linear and rotational movements through magnetic coupling transmission. The sensor component refers to a multi-physical quantity collaborative perception network, which can specifically achieve multi-angle coverage through visual array sensors, measure spatial displacement through a combination of laser interferometer 10 and laser displacement sensor 12, detect contact pressure through capacitance sensor 11, and monitor surface stress through strain sensor. The control unit refers to an intelligent data processing and decision-making center, which can specifically dynamically match detection strategies through the decision module, control the timing of the drive component movements through the drive module, and fuse multi-source data and generate a three-dimensional trajectory model through the data processing module.

[0072] Specifically, the test frame provides a modular installation foundation for sensors and drive components through the combined structure of the side panels 6 and the porous plate 1. The through-hole layout of the porous plate 1 supports the flexible configuration of sensors at different angles. The drive component adopts a design that separates the power unit and the transmission unit. The power unit is fixed to the side of the porous plate 1 through a U-shaped frame, which can both ensure the drive stability and avoid interference with the sensor layout. The sensor assembly achieves multi-perspective coverage through a visual array sensor. The laser interferometer 10 and the laser displacement sensor 12 are combined to measure spatial displacement, the capacitive sensor 11 detects contact pressure, and the strain sensor monitors surface stress, forming a multi-physical quantity collaborative perception network. The control unit dynamically matches the detection strategy through the decision module. The drive module accurately controls the action timing and amplitude of the drive component according to the strategy parameters. The data processing module adopts time alignment, spatial unification and noise filtering preprocessing strategies to eliminate multi-source data differences. The trajectory generation and comparison module restores the real motion trajectory through data fusion and three-dimensional modeling, and finally quantitatively evaluates the component accuracy through the deviation index.

[0073] Compared with existing technologies, traditional detection systems typically use a single sensor operating independently, only able to acquire single-dimensional motion data. This solution, however, integrates multiple sensor types, such as a visual array and a laser interferometer, through a porous plate 1 structure to construct a multi-dimensional data acquisition network. Existing technologies lack a dynamic calibration mechanism, while this solution implements sensor linkage calibration and error compensation through a control unit, effectively eliminating coupling errors between multiple sensors. Traditional methods rely on manual experience to select detection components, while this solution automatically matches detection strategies through a decision-making module, achieving a fully automated detection process.

[0074] Through the above technical solution, this application solves the accuracy defects caused by the insufficient coverage of a single sensor in wafer component inspection, and realizes the coordinated acquisition and dynamic compensation of multi-physical quantity data. The multi-hole plate structure provides a standardized interface for sensor layout, ensuring the spatial consistency of multi-source data; the multi-module collaborative working mechanism of the control unit effectively improves the accuracy of complex motion trajectory restoration; the automated strategy matching function of the decision module significantly improves the applicability and work efficiency of the inspection system.

[0075] In this embodiment, the power unit comprises a servo motor and a precision cylinder, which are fixed to the vertical side plate 6 at the rear of the perforated plate 1 via a U-shaped bracket. The servo motor drives the wafer assembly for rotational movement, while the precision cylinder drives the wafer assembly for linear motion. The transmission unit comprises a magnetic coupling transmission, which is located at the bottom of the perforated plate 1 and connects to the wafer assembly through the inclined holes 4 of the perforated plate 1, thereby achieving high-precision posture adjustment of the wafer assembly.

[0076] Specifically, the servo motor can be a permanent magnet synchronous servo motor with an absolute encoder. A U-shaped frame is used to rigidly secure the output shaft to ensure coaxiality with the wafer assembly's steering axis. A PID algorithm is then used to correct speed fluctuations in real time. Specifically, the servo motor is mounted horizontally in the upper half of the U-shaped frame. The output shaft passes through a pre-defined through-hole in the side panel 6, with its axis parallel to the Z-axis of the porous plate 1, ensuring that the steering drive force is transmitted along the wafer assembly's rotational axis. Specifically, the precision cylinder can be a compact cylinder with a magnetic scale. The cylinder body is connected to the side of the U-shaped frame via a floating joint to compensate for radial forces caused by installation errors. The precision cylinder is mounted vertically in the lower half of the U-shaped frame, parallel to the side panel 6. The piston rod extends along the X-axis of the porous plate 1 and aligns with the wafer assembly's translation mechanism to avoid radial additional forces during linear motion. The magnetic coupling transmission consists of a driving pulley and a driven pulley with built-in NdFeB magnetic rings. The driving pulley is connected to the drive shaft, while the driven pulley is connected to the wafer assembly's rotating shaft. The driving pulley is fixed to the bottom surface of the porous plate 1 via a bearing seat, with the center of its magnetic ring concentric with the oblique hole 4. The driven pulley is embedded in a groove on the top surface of the porous plate 1, creating a 1mm air gap with the driving pulley through the oblique hole 4 to ensure efficient magnetic field coupling. The axes of both the driving and driven pulleys coincide with the axis of the oblique hole 4, ensuring that power is transmitted to the wafer assembly at a 30° angle, avoiding interference with the detection areas of the capacitive sensor 11 and laser displacement sensor 12 (X / Y axis endpoints) on the base plate.

[0077] Specifically, to meet the wafer assembly's combined "steering and translation" motion requirements, a servo motor drives the steering shaft via a keyed connection, converting electrical pulse signals into continuous rotation within a 0-90° range to meet the angular fine-tuning requirements for wafer alignment. A precision cylinder, connected to the wafer assembly translation mechanism via a piston rod, achieves 0-30mm linear motion within a 0.2-0.6MPa air pressure range, while a buffer valve in the air path mitigates end-of-stroke impact. A magnetically coupled transmission unit transmits power from the servo motor and precision cylinder to the wafer assembly via a magnetic field, avoiding the contact wear and particle contamination associated with traditional mechanical transmissions. It also utilizes an air gap to isolate vibrations between the active and passive ends.

[0078] This application can effectively drive the movement of wafer components and control the steering positioning error within ±0.02°, the linear motion repeatability accuracy reaches ±0.005mm, and no particles are generated during the transmission process, meeting high-precision operation requirements and reducing the impact of inadequate driving on wafer component detection accuracy.

[0079] In this embodiment, the data processing module is equipped with a preprocessing strategy, which includes time alignment, spatial unification, and noise filtering. Time alignment unifies the data collected by multiple sensors to the same sampling rate and associates and stores multiple data collected at the same time as a group, with each group corresponding to a sampling time. Spatial unification maps the data collected by the corresponding sensors to a detection coordinate system based on the sensor's installation position on the porous plate 1. Noise filtering uses wavelet transform to filter high-frequency vibration noise.

[0080] Specifically, time alignment can be achieved through high-precision time-scale synchronization and interpolation algorithms. For devices with different sampling rates, such as laser interferometers 10, capacitive sensors 11, and visual sensors, linear interpolation or cubic spline interpolation is first used, based on the highest sampling rate, to supplement the missing data from the low-frame-rate sensors. Hardware trigger signals are then used to unify the timestamps of all sensors, associating measurement data from the same physical moment into a group. Each data group contains a sampling time tag accurate to the microsecond level.

[0081] Spatial unification is achieved by constructing a spatial transformation matrix based on sensor calibration parameters. A three-dimensional detection coordinate system is established with the geometric center of the porous plate 1 as the origin. The X-axis is along the front-to-back direction of the porous plate 1, the Y-axis is along the left-to-right direction, and the Z-axis is vertically upward. According to the installation parameters of each sensor, such as the vertical installation height of the forward-looking camera 8, the beam direction of the laser interferometer 10, and the annular distribution radius of the capacitive sensor 11, the pixel coordinates (u, v) of the visual sensor are converted into physical coordinates (X, Y, Z) through the translation matrix and the rotation matrix, and the gap value of the capacitive sensor 11 is mapped to the Z-axis height of the corresponding position at the bottom of the wafer, so that all data can be compared in the same spatial dimension.

[0082] Noise filtering is achieved using wavelet transform because it can distinguish signals from noise in both the time domain and the frequency domain: a wavelet basis suitable for high-frequency vibration noise filtering is selected, and 3-5 layers of wavelet decomposition are performed on the dynamic displacement signal of the laser displacement sensor 12 and the stress signal of the strain sensor. The high-frequency coefficients are processed by a soft threshold function and then the signal is reconstructed to effectively filter out the 100-1000Hz high-frequency noise generated by the servo motor operation, cylinder airflow, etc., while retaining useful signal details such as slight deformation of the wafer and sudden change in contact pressure.

[0083] Specifically, in the wafer component motion detection scenario: Time alignment ensures that when the robotic arm moves to X = 50mm, the displacement value of the laser interferometer 10, the gap value of the capacitive sensor 11, and the image features of the visual sensor are accurately correlated, avoiding asynchronous data fusion caused by time deviation, such as matching the current pressure data with the displacement data from 10ms ago. Spatial unification resolves the coordinate difference between the left edge displacement of the wafer measured by the side-view camera 9 and the center displacement measured by the laser interferometer 10. Both are mapped to the global coordinate system through a transformation matrix, enabling multi-dimensional data comparison at the same location. Noise filtering addresses vibration interference during high-speed movement of the robotic arm, reducing the standard deviation of the gap measurement of the capacitive sensor 11 from 0.1μm to 0.03μm, providing more stable observation values ​​for subsequent weighted Kalman filtering.

[0084] This solution uses three levels of collaborative processing: achieving microsecond synchronization in time, establishing global coordinate association in space, and retaining microscale features through noise filtering, laying a data foundation with temporal and spatial consistency and high signal-to-noise ratio for heterogeneous data fusion.

[0085] Through the above preprocessing strategy, the time synchronization error of each sensor data is reduced, the overall spatial coordinates are consistent, and the noise energy is reduced by ≥60%, which effectively supports the displacement fusion accuracy of the subsequent weighted Kalman filter and the pressure recognition accuracy of the DS evidence theory, providing reliable data input for the high-precision motion control of wafer components.

[0086] In some other preferred embodiments of the present invention, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the above embodiment.

[0087] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0088] The above description is a detailed description of the preferred embodiments of the present invention, but the embodiments are not intended to limit the scope of the patent application of the present invention. Any equivalent changes or modifications completed under the technical spirit suggested by the present invention should fall within the patent scope covered by the present invention.

Claims

1. A wafer component drive detection method based on a multi-hole plate and multiple sensors, characterized in that: include: Fixing the component to be detected on the porous plate through a mounting base, and electrically connecting the driving component to the component to be detected; Determine a detection strategy based on the type of the component to be detected, and initialize the working mode of the corresponding sensor component based on the detection strategy; performing a collaborative calibration on the sensor assembly; Mark the initial positions of the installation position of the component to be inspected and the position of the sensor after calibration in the inspection coordinate system; Based on the detection strategy, the driving component is started, and the driving component drives the component to be detected to move, and at the same time, the sensor component respectively collects the movement data of the component to be detected; Perform data fusion on the collected motion data to reconstruct the three-dimensional motion trajectory of the component to be inspected; The deviation between the three-dimensional motion trajectory and the standard trajectory is calculated to determine the accuracy index of the component to be inspected.

2. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 1, characterized in that: The porous plate includes central n-level through holes, surrounding n+i-level through holes, 30°, 45°, 60° oblique holes and threaded holes distributed in an array with equal spacing.

3. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 2, characterized in that: The detection strategy includes a driving process and process parameters. The driving process is determined based on the actual operation process of the component to be detected. The driving process includes linear motion, steering motion, clamping motion, and pressure motion driving in sequence. The process parameters include corresponding action parameters of each driving process. The linear motion parameters include motion distance, speed, acceleration, and action time; the steering motion parameters include rotation angle, angular velocity, action time, and rotation axis offset; the clamping motion parameters include clamping stroke, clamping force, action time, and clamping parallelism; the pressure motion parameters include pressure value, pressing depth, force control response time, and holding time.

4. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 1, characterized in that: The sensor assembly includes a visual array sensor, a strain sensor, a laser interferometer, a capacitance sensor and a laser displacement sensor.

5. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 1, characterized in that: The collaborative calibration of the sensor assembly specifically includes static calibration: Start each corresponding sensor in the sensor assembly and restore it to its initial state first; In the initial state, each sensor collects initial data from the standard point; The initial data collected by the visual array sensor is matched with standard points in the overlapping area of ​​the multi-camera field of view, and the conversion relationship between each camera coordinate system and the multi-well plate reference coordinate system is established; The laser interferometer calibrates the beam steering error through the angle calibration block with oblique holes, and uses standard points to calibrate the orthogonality of the X / Y / Z axis optical path.

6. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 5, characterized in that: The collaborative calibration of the sensor assembly also includes linkage calibration: Based on the measurement data of the reference array, the error of each sensor and the influence coefficient of the sensor error on other sensors are calculated to form a coupling error matrix. During the calibration process, multiple sensors in the sensor matrix are adjusted in a linked manner based on the coupling error matrix.

7. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 5, characterized in that: The collaborative calibration of the sensor assembly also includes dynamic calibration: When the sensor collects different motion data, the error prediction model outputs the current sensor error compensation value according to the parameters of the current driving process, realizing real-time calibration during motion.

8. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 1, characterized in that: The data fusion includes: Based on weighted Kalman filtering, the same state data collected by the laser interferometer, capacitive sensor and laser displacement sensor are fused according to certain weights to obtain the displacement data of the wafer components. The data of visual sensors and strain sensors are fused based on DS evidence theory to obtain the end pressure data of wafer components.

9. The wafer component drive detection method based on a multi-hole plate and multiple sensors according to claim 1, characterized in that: The specific steps for determining the accuracy index of the component to be inspected include: Based on the working conditions corresponding to the driving strategy and the data fused by the data fusion strategy, the corresponding quantitative data of the wafer components are extracted; Generate a three-dimensional motion trajectory by mathematical modeling based on quantitative data in the detection coordinate system; The deviation between the three-dimensional motion trajectory and the standard trajectory is calculated to obtain the deviation index.

10. A wafer component drive detection system based on a multi-hole plate and multiple sensors, characterized in that: include: A test stand, comprising two side plates and a porous plate, wherein the porous plate is fixed to the side plates, and a top plate is further provided on top of the porous plate; A driving component, comprising a power unit and a transmission unit, fixedly disposed on the test frame and configured to drive the wafer component to move; A sensor assembly, comprising a vision array sensor, a laser interferometer, a capacitive sensor, a laser displacement sensor, and a strain sensor; A control unit is communicatively connected with the driving component and the sensor assembly, and includes a data processing module, a trajectory generation and comparison module, a decision module, and a driving module. The data processing module is used to receive the collected data of the sensor assembly and perform data preprocessing and data fusion on the collected data. The fused data is restored and compared by the trajectory generation and comparison module to obtain a trajectory deviation index. The decision module determines the driving strategy of the driving component and the acquisition strategy of the sensor assembly according to the type of wafer component, and sends the driving strategy and sensor strategy to the driving module. The driving module controls the operation of the driving component and the sensor assembly according to the corresponding strategy.

Images