Loading apparatus and control method thereof

Abstract

The application discloses a loading device and a control method thereof, which comprises a monitoring mechanism, a center module arranged on one side of the monitoring mechanism, and a control unit; the center module comprises an upper die assembly and a lower die assembly which belongs to the upper die assembly; the bottom of the upper die assembly and the lower die assembly is respectively provided with a movable rail assembly and a fixed rail assembly; the fixed rail assembly and the movable rail assembly control the operation of the upper die assembly and the lower die assembly to assemble the workpiece; the monitoring mechanism comprises an industrial camera and a moving piece for driving the industrial camera; the moving piece is used for driving the industrial camera to move and detect the upper die assembly and the lower die assembly; the control unit is used for detecting the movement deviation of the upper die assembly and the lower die assembly during the workpiece assembly by controlling the industrial camera, and simultaneously generating a compensation instruction to control the upper die assembly and the lower die assembly to assemble; the total deviation is offset, and the hierarchical compensation eliminates the composite angle deviation caused by the multi-stage motion chain, so that the high-precision alignment of the upper die assembly and the lower die assembly is ensured.

Description

Technical Field

[0001] This invention relates to the field of loading equipment control methods, and more particularly to a loading equipment and its control method. Background Technology

[0002] With the continuous development of the precision manufacturing field, loading equipment, as a key device in automated production lines for achieving precise workpiece positioning, clamping, and transfer, is becoming increasingly important. Especially in industries such as semiconductor packaging, panel display, and high-precision electronic assembly, loading equipment needs to transport wafers, glass substrates, or micro-components to designated workstations at high speed and without damage, achieving precise alignment with target positions at the micron or even sub-micron level. The reliability of this process directly determines the quality of subsequent processes and the yield of the final product.

[0003] Currently, most common control methods for loading equipment rely on preset programmed paths and single-axis servo control. Specifically, the control system typically drives the fixture to move sequentially along linear axes such as X, Y, and Z, as well as rotary axes, based on pre-calibrated theoretical coordinates. During operation, each axis mostly moves independently or performs simple interpolation movements, and the core of its control logic is to make each axis track its command position as accurately as possible. However, this method largely treats the fixture as an idealized rigid body, and its control model is relatively static and isolated.

[0004] A significant limitation of this traditional control mode is its lack of ability to actively and in real-time compensate for dynamic errors between multiple axes. In actual working environments, fixtures and their drive systems are affected by a combination of dynamic factors, including mechanical backlash, transmission chain errors, thermal deformation, and motion inertia. When the fixture moves at high speed or along complex trajectories, these factors cause small, interconnected deviations between the actual motion of each axis and the command. Because existing control methods fail to perceive and compensate for these inter-axis deviations as a whole in real time, their control commands cannot dynamically respond to this coupled error, resulting in a deviation between the actual pose of the fixture end and the theoretical target position. This deviation manifests as "misalignment" in precision alignment scenarios, affecting not only machining and assembly accuracy but also potentially causing workpiece collisions or damage, thus hindering further improvements in production efficiency and product quality. Therefore, developing an intelligent control method capable of actively compensating for multi-axis linkage errors has become an urgent technical requirement for improving the performance of high-end loading equipment. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides a loading device and its control method.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a loading device, comprising: a monitoring mechanism, a central module disposed on one side of the monitoring mechanism, and a control unit;

[0007] The central module includes an upper mold assembly and a lower mold assembly belonging to the upper mold assembly; the bottom of the upper mold assembly and the lower mold assembly are respectively provided with a moving rail assembly and a fixed rail assembly, and the fixed rail assembly and the moving rail assembly control the movement of the upper mold assembly and the lower mold assembly to assemble the workpiece;

[0008] The monitoring mechanism includes an industrial camera and a moving component that drives the industrial camera. The moving component is used to move the industrial camera to detect the upper mold assembly and the lower mold assembly.

[0009] The control unit is used to detect the movement offset of the upper and lower mold components during workpiece assembly by controlling an industrial camera, and at the same time generate compensation commands to control the upper and lower mold components to assemble.

[0010] In a preferred embodiment of the present invention, the upper mold assembly includes a frame, and an upper mold cylinder is fixedly connected to the top of the frame. The upper mold cylinder is used to drive the upper mold to assemble the workpiece.

[0011] In a preferred embodiment of the present invention, the lower mold assembly includes a lower mold, and a linear motor for driving the lower mold to assemble is provided at the bottom of the lower mold. A locking component is provided on the upper mold, and a locking component is also provided on the lower mold. The locking components are used to clamp the workpiece.

[0012] In a preferred embodiment of the present invention, the central module and the monitoring mechanism are arranged in one set or two sets symmetrically arranged to form a single-station loading device or a dual-station loading device.

[0013] In a preferred embodiment of the present invention, the moving rail assembly includes a base and a first slide rail disposed between the base and the frame; a first motor is disposed inside the base, and the first motor is used to drive the frame to move.

[0014] In a preferred embodiment of the present invention, the guide rail assembly includes a second slide rail, which is disposed between the frame and the lower mold, and a second motor is connected to one side of the second slide rail.

[0015] A method for controlling a loading device includes the following steps:

[0016] S1: Drive the loading equipment to run within the working area, acquire the actual position data of each point, and decouple the actual position data based on the preset structured error model to generate positioning error compensation data and straightness error compensation data for each motion axis.

[0017] Among them, the straightness error compensation data of one motion axis is constructed in the form of a two-dimensional table, and the compensation value in the table varies according to the different positions of the other motion axis;

[0018] S2: Receive the target position command, perform inverse kinematics based on the ideal kinematics model, calculate the uncompensated initial axis command, call the positioning error compensation data and straightness error compensation data for feedforward calculation, perform the first correction on the initial axis command, calculate the linkage compensation amount of multiple motion axes, perform the second correction on the axis, offset the total prediction deviation in this direction, and finally generate the drive command.

[0019] S3: Residual visual deviations in workpiece alignment are detected by imaging with an industrial camera;

[0020] If the residual deviation exceeds the allowable range, the detected visual deviation amount is used as a new input, and feedforward calculation is performed again. Fine-tuning is then performed based on the results to eliminate the residual deviation.

[0021] In a preferred embodiment of the present invention, in S1, the construction is specifically as follows:

[0022] The first motion shaft is fixed at different positions. At each fixed position, the second motion shaft is driven to perform a full-stroke motion. The offset of the loading device in the driving direction of the second motion shaft is measured to establish a two-dimensional mapping relationship in which the offset changes together with the positions of the first and second motion shafts.

[0023] In a preferred embodiment of the present invention, in S2, the feedforward calculation includes multiplying a coupling coefficient related to the actual position of the motion axis by the command value of another motion axis to characterize and offset the motion coupling error caused by the non-parallelism between the motion axes.

[0024] In a preferred embodiment of the present invention, in S3, the detected visual deviation is used as a new input for feedforward calculation again, specifically as follows:

[0025] Using the negative of the residual visual deviation as a temporary position correction target, motion planning and multi-level compensation steps are re-executed to calculate fine-tuning instructions.

[0026] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0027] (1) This invention provides a loading device and its control method. By adopting a multi-level feedforward compensation strategy, the initial axis command is generated based on the inverse kinematics model. Then, the error database is called for the first correction, and the linkage compensation amount of multiple motion axes is calculated for the second correction to offset the total prediction deviation. This hierarchical compensation eliminates the compound angle offset caused by the multi-level motion chain, ensuring high-precision alignment of the upper and lower mold components and reducing the risk of assembly stress concentration and workpiece damage. Compared with the prior art, which lacks active multi-axis linkage compensation and often leads to inaccurate alignment, this invention achieves pure correction action through linkage compensation. Furthermore, this strategy enhances the dynamic response capability and reliability of the system, is suitable for high-paced automated assembly scenarios, and improves the overall production efficiency.

[0028] (2) The present invention provides a loading device and its control method. By constructing error data, the comprehensive error of the loading device is decomposed into independent error components with clear physical meaning, so that the error can be accurately quantified and independently compensated, thereby realizing high-precision prediction and active cancellation of the inherent geometric error of the mechanical system, and significantly improving the positioning accuracy. Compared with the existing technology that relies on improving the manufacturing precision of a single guide rail or the overall error mapping, it cannot effectively eliminate the composite angle offset problem caused by the error coupling of multi-level kinematic chains. The present invention reduces the dependence on the hardware limit precision through software modeling.

[0029] (3) This invention provides a loading device and its control method. By integrating a visual closed-loop feedback mechanism, the residual visual deviation of the workpiece alignment is detected by imaging with an industrial camera. The negative number of the residual deviation is used as a temporary position correction target, and the motion planning and multi-level compensation steps are re-executed for fine-tuning. This analysis combines real-time visual measurement and feedforward compensation model to form an adaptive closed-loop control system, which eliminates the small deviation remaining after feedforward compensation, so that the final assembly accuracy reaches the micron-level tolerance requirement. Compared with the existing technology, which may only rely on hardware accuracy or simple feedback and cannot handle dynamic disturbances and random noise, this invention achieves a higher level of precision control through the combination of vision and model compensation. Furthermore, the system has anti-interference and adaptive capabilities, can cope with environmental changes and long-term drift, improves robustness and flexibility, and is particularly suitable for high-precision and flexible manufacturing environments. Attached Figure Description

[0030] 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.

[0031] Figure 1This is a perspective structural diagram of a preferred embodiment of the present invention;

[0032] Figure 2 This is a three-dimensional structural diagram of an industrial camera, representing a preferred embodiment of the present invention;

[0033] Figure 3 This is a three-dimensional structural diagram of a preferred embodiment of the present invention, and a schematic diagram of a dual-station structure.

[0034] In the diagram: 1. Monitoring mechanism; 2. Industrial camera; 3. Moving part; 4. Central module; 5. Upper mold assembly; 6. Frame; 7. Upper mold cylinder; 8. Upper mold; 9. Lower mold assembly; 10. Lower mold; 11. Linear motor; 12. Locking component; 13. Base; 14. First slide rail; 15. First motor; 16. Second slide rail; 17. Second motor. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0037] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the scope of protection of this application. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0038] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0039] Application Overview:

[0040] In the field of precision manufacturing, loading equipment is a key process equipment for achieving high-precision positioning and assembly of workpieces. Traditional loading equipment adopts a rigid positioning method, relying on the machining and installation accuracy of the guide rail to ensure the alignment of the upper and lower molds. However, in actual operation, due to the existence of multi-stage motion chains, the inherent geometric errors will be coupled and superimposed during the motion process, resulting in the so-called compound angular offset problem.

[0041] Specifically, when the moving rail of the loading equipment is installed on the movable fixed rail, the error of the fixed rail will be transmitted to the installation reference of the fixed rail, and the error of the moving rail itself will be further amplified on this reference. This error coupling caused by the series motion chain makes the actual movement trajectory of the lower mold deviate significantly from the theoretical command trajectory, resulting in misalignment of the upper and lower workpieces, stress concentration during assembly, and even damage to the workpiece or loading equipment. Existing technology alleviates this problem by improving the manufacturing precision of a single guide rail, but cannot fundamentally eliminate the influence of multi-level error coupling. Especially after long-term operation and wear, the problem of precision degradation becomes more prominent.

[0042] To address the aforementioned technical problems, this invention provides a loading device and its control method. The loading device includes a monitoring mechanism, a central module, and a control unit. The central module comprises an upper mold assembly and a lower mold assembly, with a moving rail assembly and a fixed rail assembly respectively disposed at its bottom to achieve workpiece clamping and centering movement. Crucially, the monitoring mechanism includes a laterally movable industrial camera for dynamically detecting the workpiece position. The control unit is configured to execute an innovative multi-level compensation control method.

[0043] The core of this invention lies in constructing a structured error database, rather than a traditional overall error mapping. This database decomposes the comprehensive spatial error of the system into independent error components with clear physical meaning, including the positioning error of each motion axis, the straightness error caused in the vertical direction when each axis moves, and the coupling coefficient characterizing the non-parallelism between axes. In particular, the straightness error of the moving track is constructed in the form of a two-dimensional table, and its compensation value changes dynamically according to the different positions of the fixed track, thereby accurately quantifying and capturing the composite angular offset effect.

[0044] In real-time control, upon receiving the target position command, the system first performs inverse kinematics calculation based on the ideal kinematic model. Then, it calls the structured error database for feedforward compensation calculation, performing a first correction to the initial axis commands to eliminate individual axis errors. Next, addressing the predicted total deviation, especially the vertical alignment deviation, the system calculates the cooperative compensation amount by solving the multi-axis linkage kinematic model, performing a second correction to the axis commands. This instructs multiple axes to move collaboratively to produce a clean correction action. Furthermore, the system integrates visual closed-loop feedback, using an industrial camera to detect residual alignment deviations and perform fine-tuning to ensure final accuracy.

[0045] This invention upgrades the control strategy from a passive, single-point hysteresis feedback-based mode to an active, full-field error prediction-based precision compensation mode by introducing structured error modeling and multi-axis linkage feedforward compensation. This fundamentally solves the accuracy problem caused by the compound angular offset of multi-level kinematic chains, significantly improving the dynamic alignment accuracy and repeatability of the loading equipment. Software compensation reduces the dependence on the ultimate precision of mechanical hardware, improving the system's economy and long-term stability. The combination of visual closed loop and compensation model gives the system anti-interference and adaptive capabilities, making it particularly suitable for high-precision, flexible automated assembly scenarios.

[0046] Example 1:

[0047] S1: Drive the loading equipment to run within the working area, acquire the actual position data of each point, and decouple the actual position data based on the preset structured error model to generate positioning error compensation data and straightness error compensation data for each motion axis.

[0048] Among them, the straightness error compensation data of one motion axis is constructed in the form of a two-dimensional table, and the compensation value in the table varies according to the different positions of the other motion axis;

[0049] In a preferred embodiment of the present invention, in S1, the construction is specifically as follows:

[0050] The first motion shaft is fixed at different positions. At each fixed position, the second motion shaft is driven to perform a full-stroke motion. The offset of the loading device in the driving direction of the second motion shaft is measured to establish a two-dimensional mapping relationship in which the offset changes together with the positions of the first and second motion shafts.

[0051] It should be noted that in S1, by constructing a high-precision structured error database, the essence is to decompose the comprehensive geometric error of the complex mechanical system of the loading equipment into error components with clear physical meaning that can be independently measured and compensated.

[0052] Specifically, the control system drives the loading equipment to perform precise grid-based scanning motion within a two-dimensional working plane spanned by the first and second motion axes.

[0053] The first motion axis is denoted as the L-axis, which is the fixed track that controls the lateral displacement of the entire module in the device of the present invention, and the second motion axis is denoted as the D-axis, which is the moving track that controls the relative sliding of the lower module assembly in the device of the present invention.

[0054] For each command position (L) issued by the control system c md i D c md j The system will pause and wait for the motion to fully stabilize. Then, an external high-precision measuring device, namely a laser tracker, will measure the actual three-dimensional spatial coordinates of the endpoint in an absolute world coordinate system. .

[0055] This resulted in a dense set of discrete data mappings: ,

[0056] This provides the most basic data foundation for subsequent error decoupling calculations.

[0057] After acquiring the raw data, decoupling calculations are performed based on a preset structured error model. This structured error model systematically decomposes the data into a linear superposition of error components with clear physical meanings originating from different mechanical parts.

[0058] For an ideal loading device with no geometric errors, the position P of the end effector in the world coordinate system is... i The deal command and axis command (L, D) have a simple linear relationship, for example .

[0059] At the position P obtained by actual measurement a ctual and ideal position P i There are discrepancies between the deals. .

[0060] The structured error model of this invention expresses this total deviation as:

[0061]

[0062]

[0063] In this model, ΔL(L) and ΔD(D) are the positioning error functions of the L-axis and D-axis, respectively, describing the systematic deviation between the commanded position and the actual physical position of each axis.

[0064] δY_L(L) is the straightness error function of the L-axis, which quantifies the undesirable parasitic motion generated in the Y-direction perpendicular to the L-axis when it moves alone.

[0065] The term ε(L) • D characterizes the X-axis coupling error caused by a non-parallelism angle related to the position L due to the non-ideal parallelism of the two axes' motion directions. Its tangent is approximately ε(L). The most critical term is δY. D (D, L) is the straightness error function of the D-axis, but its characteristics change with the position of the L-axis.

[0066] In the above model, the moving track, i.e. the reference direction of the D-axis, is modulated by the pose error of the central module base plate driven by its mounting base, i.e. the L-axis. This causes the parasitic motion in the Y direction generated when the D-axis moves to no longer be a fixed function of D, but a two-dimensional function related to L.

[0067] The process of generating positioning error compensation data and straightness error compensation data for each motion axis involves using a large amount of measured data (L, D, ΔX). t otal,ΔY t The data points (total) are used to inversely solve the various error functions in the above model by curve fitting using the least squares method.

[0068] For example, to extract the positioning error ΔL(L) of the L-axis, the D-axis is fixed at zero (D=0), and the L-axis is allowed to move its full range. In this case, the model simplifies to ΔX. t otal(L,0)≈ΔL(L), by measuring ΔX at position L. t The total value is then fitted to obtain the compensation curve for ΔL(L).

[0069] Similarly, fixing L=0, we iterate along the D-axis to obtain ΔD(D). This is used to obtain the straightness error δY along the L-axis. L (L), also under D=0, measure ΔY t otal(L,0)≈δY L (L) and fitting.

[0070] The straightness error compensation data for one motion axis is constructed in the form of a two-dimensional table, and the specific implementation of the compensation value changing with the position of another motion axis is used to construct δY. D Taking a two-dimensional compensation table of (D, L) as an example, the specific implementation of this construction method is as follows:

[0071] The first motion axis (L-axis) is fixed sequentially at a series of discrete positions L. k (k = 1, 2, ..., m). In each fixed L... kPosition, drives the second motion axis (D axis) from its travel start point to its end point, stepping at certain intervals ΔD.

[0072] At each D-axis command position D p Measure the actual coordinate value Y of the loading device end in the Y direction of the world coordinate system at this time. a ctual(L k D p ).

[0073] Calculate in this specific L k Below, the pure Y-axis offset caused by the movement of the D-axis, i.e. .

[0074] This calculation eliminates the problem in L. k Position is determined by the straightness error δY of the L-axis itself. L (L k The fixed bias introduced by this isolates the Y-axis error purely caused by D-axis motion. Iterate through all preset L... k By retrieving the location and repeating the above process, we will eventually obtain a two-dimensional data table (matrix).

[0075] The row indices of this matrix correspond to different L-axis positions. k The column index corresponds to different D-axis commands. p Each element in the matrix represents the corresponding δY. D (D p ;L k Compensation value.

[0076] During real-time control, when the system needs to compensate for an arbitrary command point (L,D), it will use a bilinear interpolation algorithm to look up the required δY in this two-dimensional table. D Compensation Amount. In this way, the straightness error of the D-axis (moving track) is captured and quantified by the position of the L-axis (fixed track) through this two-dimensional table, thereby characterizing and storing the dynamically changing composite angular offset effect caused by the coupling of geometric errors of the series motion chain, laying a solid data foundation for the subsequent realization of high-precision feedforward compensation.

[0077] S2: Receive the target position command, perform inverse kinematics based on the ideal kinematics model, calculate the uncompensated initial axis command, call the positioning error compensation data and straightness error compensation data for feedforward calculation, perform the first correction on the initial axis command, calculate the linkage compensation amount of multiple motion axes, perform the second correction on the axis, offset the total prediction deviation in this direction, and finally generate the drive command.

[0078] In a preferred embodiment of the present invention, in S2, the feedforward calculation includes multiplying a coupling coefficient related to the actual position of the motion axis by the command value of another motion axis to characterize and offset the motion coupling error caused by the non-parallelism between the motion axes.

[0079] It should be noted that in S2, the target workpiece position is represented in the world coordinate system as... This is converted into actual commands to drive each motion axis, and multi-level feedforward compensation is used to offset the inherent geometric errors of the mechanical system, ultimately achieving high-precision alignment. Specifically:

[0080] Based on the inverse solution of the ideal kinematic model, for an ideal error-free system, its kinematic model simplifies to a linear relationship. Assuming that the two axes of motion, namely the L-axis and the D-axis, are ideally parallel and aligned in the world coordinate system, the end position is [X,Y]. T AND axis commands [L,D] T The relationship is:

[0081]

[0082]

[0083] When the system receives the target location At this time, the inverse kinematics is solved, that is, the axis command to be issued is calculated. This ensures that the end position in the ideal model equals the target position. This is achieved by solving the above system of equations:

[0084]

[0085] Because Y i The deal is always 0, while the target Y t The value is usually not zero, requiring correction in the Y direction. However, the ideal model itself cannot generate Y-direction motion, highlighting its limitations. Therefore, the initial inverse solution ignores the Y-direction deviation and focuses on X-direction positioning. Initial conditions are set, letting D... i The deal value is set to a preset value of 0, thus obtaining the initial axis command without compensation: .

[0086] The calculation is performed in this way, but this calculation instruction only approximately satisfies the target in the X direction and does not take into account mechanical errors and Y-direction alignment requirements, resulting in deviations in the output trajectory.

[0087] The system's first correction involves calling the structured error database for feedforward calculations. Using the error compensation data established in phase S1—namely, the positioning error functions ΔL(L), ΔD(D), the straightness error table δY_D(D, L), and the coupling coefficient ε(L)—pre-compensates the initial axis commands. The correction goal is to make the actual physical position of each axis as close as possible to its command value. The correction formula is:

[0088]

[0089] During this correction phase, ΔL(L) i deal) and ΔD(D i The deal directly compensates for the positioning errors of each axis. The term -ε(L) requires particular explanation. i deal)•D i The deal represents the use of a coupling coefficient related to the actual position of the motion axis, multiplied by the command value of another motion axis, to characterize and offset motion coupling errors caused by non-parallelism between motion axes.

[0090] The coupling coefficient ε(L) here is obtained through calibration in stage S1, quantifying the additional deviation of the L-axis world coordinates caused by the non-parallelism of the two axes when the D-axis moves a unit distance. By subtracting this estimated coupling deviation from the L-axis command in advance, this part of the systematic error is offset at the command level.

[0091] After the first correction, the overall pose of the ends of each axis's actual physical position, especially the Y-axis position perpendicular to the plane of motion, was still not corrected. Since the loading equipment itself does not have an independent Y-axis drive axis, a net displacement in the Y-axis must be synthesized through the coordinated movement of the L-axis and D-axis, utilizing their straightness error effect. Therefore, a second correction is needed to calculate the linkage compensation amount for the coordinated movement of multiple axes.

[0092] Specifically, it generates additional motion. This causes the synthesized end motion to produce a Y-direction displacement ΔY', and ΔY' is equal to the negative of the predicted total Y-direction deviation -δY_total that needs to be offset, i.e. ,

[0093] Simultaneously, a differential kinematic model of the loading equipment, namely the Jacobian matrix J, is established to reduce interference with the X-axis position. Specifically, this involves small changes in the end-effector pose [ΔX, ΔY] near any pose (L, D). T Minor changes in axis commands [ΔL, ΔD] T Approximately satisfies:

[0094]

[0095] The elements of J in the Jacobian matrix can be obtained through the error model.

[0096] For example:

[0097] Rate of change of L-axis straightness error

[0098]

[0099] Rate of change of straightness error along the D-axis.

[0100] Our goal is to solve , so that:

[0101]

[0102] By calculating the inverse operation of the Jacobian matrix and solving this system of equations, the required linkage compensation amounts [ΔL', ΔD'] can be obtained. T This compensation command instructs the L-axis and D-axis to perform coordinated additional motion, actively and precisely generating a Y-axis correction action.

[0103] Finally, the two correction values ​​are summed to generate the final drive command sent to the servo driver:

[0104]

[0105]

[0106] By employing a three-step process of ideal inverse kinematics, error feedforward compensation, and linkage motion synthesis, the abstract target position command is transformed into an axis control command that can actively offset the main geometric errors of the mechanical system. This lays a solid foundation for achieving high-precision dynamic alignment and elevates the control strategy from passive hysteresis feedback to model-predictive feedforward active compensation, greatly enhancing the ability to suppress complex errors.

[0107] S3: Residual visual deviations in workpiece alignment are detected by imaging with an industrial camera;

[0108] If the residual deviation exceeds the allowable range, the detected visual deviation amount is used as a new input, and feedforward calculation is performed again. Fine-tuning is then performed based on the results to eliminate the residual deviation.

[0109] In a preferred embodiment of the present invention, in S3, the detected visual deviation is used as a new input for feedforward calculation again, specifically as follows:

[0110] Using the negative of the residual visual deviation as a temporary position correction target, motion planning and multi-level compensation steps are re-executed to calculate fine-tuning instructions.

[0111] It should be noted that after the loading equipment completes the initial positioning based on feedforward compensation, the system performs visual fine-tuning in this step to eliminate the minor residual deviations that still exist after the modeling and compensation in the first two stages. These deviations originate from incomplete models, unmodeled dynamic disturbances (such as vibration and thermal deformation), or random noise. The specific operation is as follows:

[0112] The industrial camera moves to a preset observation position under system control and simultaneously captures images of specific optical alignment marks located on the upper and lower mold workpieces.

[0113] Machine vision algorithms are used to accurately extract the pixel coordinates of these markers from the acquired images. u p, v u p] and [u d own,v d own).

[0114] Using a pre-calibrated matrix and camera intrinsic parameters, these pixel coordinates are transformed to a unified world coordinate system to obtain the coordinates of the top marker point. and the coordinates of the marked point .

[0115] The final residual visual bias vector δ v The isual is calculated as the positional difference between the lower marker point and the upper marker point: This vector reflects the slight positional deviation that still exists in the alignment direction of the upper and lower workpieces after compensation in step S2.

[0116] The system will use the magnitude |δ| of this residual visual bias vector v isual| with a preset, extremely small tolerance threshold ε t The comparison is performed at 2 micrometers. If the residual deviation exceeds the allowable range, a fine-tuning process is triggered.

[0117] The core logic of fine-tuning is to use the detected visual deviation as new input and perform feedforward calculation again. Specifically, as described in the preferred embodiment, the negative of the residual visual deviation is used as a temporary position correction target. The system sets a new, instantaneous target position. .

[0118] Since the current deviation is δ v In order to eliminate the current deviation, the lower die workpiece is moved by -δ to the upper die workpiece. v The distance of the actual location. Therefore, this temporary position correction target is set to... .

[0119] The system re-executes the motion planning and multi-level compensation steps, minimizing the complex calculation process in step S2 and running it again. However, this time the input is no longer the original assembly target, but this tiny correction amount. Specifically:

[0120] The system utilizes an ideal kinematic model to calculate the desired kinematics. Required ideal axis command increment In a simple linear model, .

[0121] The system calls the complete structured error database established in phase S1, including a two-dimensional table of positioning error, straightness error, and coupling coefficients, to compensate for this ideal command increment, generating an axis command increment that has undergone the first correction. The calculation process is the same as in S2, ensuring that even for minute compensation movements, its execution avoids introducing new mechanical errors.

[0122] The system calculates the amount of correction in the Y direction, -δY, to ensure accurate generation. v The required linkage compensation increment for isual This is achieved by solving the Jacobian matrix equation. This is done to ensure that the Y-axis deviation is precisely offset.

[0123] Finally, the incremental changes from the two corrections are summed to obtain the final fine-tuning instructions. And send it to the servo driver for execution.

[0124] This fine-tuning command drives each axis to perform a very small and rapid coordinated movement. After the movement is complete, the system triggers the industrial camera again to perform imaging detection to verify whether the residual deviation has been reduced to within the tolerance range. If a deviation exceeding the tolerance range still occurs, this fine-tuning process is iterated again until the residual deviation meets the accuracy requirements.

[0125] Example 2:

[0126] A loading device includes: a monitoring mechanism, a central module disposed on one side of the monitoring mechanism, and a control unit;

[0127] The central module includes an upper mold assembly and a lower mold assembly belonging to the upper mold assembly; the bottom of the upper mold assembly and the lower mold assembly are respectively provided with a moving rail assembly and a fixed rail assembly, and the fixed rail assembly and the moving rail assembly control the movement of the upper mold assembly and the lower mold assembly to assemble the workpiece;

[0128] In a preferred embodiment of the present invention, the upper mold assembly includes a frame, and an upper mold cylinder is fixedly connected to the top of the frame. The upper mold cylinder is used to drive the upper mold to assemble the workpiece.

[0129] In a preferred embodiment of the present invention, the lower mold assembly includes a lower mold, and a linear motor for driving the lower mold to assemble is provided at the bottom of the lower mold. A locking component is provided on the upper mold, and a locking component is also provided on the lower mold. The locking components are used to clamp the workpiece.

[0130] In a preferred embodiment of the present invention, the central module and the monitoring mechanism are arranged in one set or two sets symmetrically arranged to form a single-station loading device or a dual-station loading device.

[0131] In a preferred embodiment of the present invention, the moving rail assembly includes a base and a first slide rail disposed between the base and the frame; a first motor is disposed inside the base, and the first motor is used to drive the frame to move.

[0132] In a preferred embodiment of the present invention, the guide rail assembly includes a second slide rail, which is disposed between the frame and the lower mold, and a second motor is connected to one side of the second slide rail.

[0133] It should be noted that the central module includes an upper mold assembly and a lower mold assembly. The upper mold assembly includes a frame, which is located above the base. The frame includes a top plate, a bottom plate, and a connecting shaft that fixes the top plate and the bottom plate. The central module also includes a moving rail assembly and a fixed rail assembly. The moving rail assembly includes a first slide rail and a first slide rail, which are installed between the frame and the base to drive the frame to move and assemble. A first motor is installed inside the base to drive the frame to run on the moving rail assembly. The upper mold cylinder is fixed to the top of the frame, and its output shaft is fixedly connected to the upper mold to drive the upper mold to run.

[0134] The lower mold assembly includes a lower mold and a linear motor. The linear motor is mounted on the base plate of the frame, and the lower mold is located below the upper mold. The lower mold is mounted on the linear motor, so that the output of the linear motor drives the lower mold to move laterally to cooperate with the upper mold for assembly.

[0135] The guide rail assembly includes a second slide rail and a second motor. The second slide rail is longitudinally arranged on the lower mold assembly, and the second motor is connected to the lower mold assembly for transmission, thereby driving the lower mold assembly to move longitudinally on the second slide rail through the output of the second motor.

[0136] Locking components are provided on both the upper and lower molds. The locking components are respectively located on one side of the top plate or bottom plate near the frame. The locking components are dovetail groove clamps, which include cylinders and wedge blocks. When the cylinder is activated, it pushes the wedge blocks to slide along the inclined surface of the dovetail groove, generating a huge lateral clamping force, which rigidly locks the workpiece in the cavity from both sides.

[0137] The monitoring mechanism includes an industrial camera and a moving component that drives the industrial camera. The moving component is used to move the industrial camera to detect the upper mold assembly and the lower mold assembly.

[0138] It should be noted that the monitoring system includes an industrial camera and a moving component that drives the camera. The camera is driven by a linear module, allowing it to move to a specific observation position and image the workpieces clamped on the upper and lower mold assemblies. This movable design enables a single high-resolution camera to cover a large working area and measure multiple key alignment marks. The image data acquired by the camera is the sole source of information for the control system to perform visual servo feedback and residual deviation detection, forming the basis for achieving closed-loop precision compensation.

[0139] A prism is set at the front end of the industrial camera, and a rotary cylinder is set on one side of the prism. Both the rotary cylinder and the prism are relatively stationary with respect to the industrial camera. Only when the industrial camera needs to inspect the upper mold, the rotary cylinder is rotated, which drives the prism to rotate, so that the industrial camera can capture the image of the upper mold.

[0140] The control unit is used to detect the movement offset of the upper and lower mold components during workpiece assembly by controlling an industrial camera, and at the same time generate compensation commands to control the upper and lower mold components to assemble.

[0141] It should be noted that the control unit is used to receive visual feedback signals from the monitoring agency, i.e., the industrial camera. Its main purpose is to identify the movement offset caused by the inherent error of the mechanical system in real time by running the control algorithm based on the visual signals and shaft encoder feedback. It calls the structured error database established in the S1 stage, which includes the positioning error table of each axis and the two-dimensional table of straightness error. Combined with the multi-level compensation strategy from S2 to S3, including feedforward compensation, linkage compensation, and visual fine adjustment, it calculates the compensation command that can actively cancel these offsets in real time.

[0142] The aforementioned compensation commands are sent to the drivers of the moving track and the stationary track, thereby forming a high-frequency closed loop of sensing-computation-compensation, reducing errors during operation.

[0143] Example 3:

[0144] Based on Embodiment 2, the device of the present invention has two sets of central modules and monitoring mechanisms stacked on the left and right sides of the base to achieve a dual-station layout.

[0145] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A loading device, characterized in that, include: The monitoring mechanism includes a central module located on one side of the monitoring mechanism and a control unit. The central module includes an upper mold assembly and a lower mold assembly belonging to the upper mold assembly; the bottom of the upper mold assembly and the lower mold assembly are respectively provided with a moving rail assembly and a fixed rail assembly, and the fixed rail assembly and the moving rail assembly respectively control the operation of the upper mold assembly and the lower mold assembly to assemble the workpiece; The monitoring mechanism includes an industrial camera and a moving component that drives the industrial camera. The moving component is used to move the industrial camera to detect the upper mold assembly and the lower mold assembly. The control unit is used to detect the movement offset of the upper mold assembly and lower mold assembly during workpiece assembly by controlling an industrial camera, and at the same time generate compensation commands to control the upper mold assembly and lower mold assembly to perform assembly.

2. The loading device according to claim 1, characterized in that: The upper mold assembly includes a frame, and an upper mold cylinder is fixedly connected to the top of the frame. The upper mold cylinder is used to drive the upper mold to assemble the workpiece, and a locking component is provided on the upper mold.

3. The loading device and its control method according to claim 1, characterized in that: The lower mold assembly includes a lower mold, and a linear motor for driving the lower mold to assemble is provided at the bottom of the lower mold. The lower mold is also provided with locking components, which are used to clamp the workpiece.

4. The loading device and its control method according to claim 1, characterized in that: The central module and the monitoring mechanism are arranged in one set or two sets symmetrically arranged to form a single-station loading device or a dual-station loading device.

5. The loading device and its control method according to claim 2, characterized in that: The moving rail assembly includes a base and a first slide rail disposed between the base and the frame; a first motor is disposed inside the base, and the first motor is used to drive the frame to move.

6. The loading device and its control method according to claim 1, characterized in that: The track-setting assembly includes a second slide rail, which is disposed between the frame and the lower mold, and a second motor is connected to one side of the second slide rail.

7. A control method for a loading device, based on the loading device according to any one of claims 1-6, characterized in that, Includes the following steps: S1: Drive the loading equipment to run within the working area, acquire the actual position data of each point, and decouple the actual position data based on the preset structured error model to generate positioning error compensation data and straightness error compensation data for each motion axis. Among them, the straightness error compensation data of one motion axis is constructed in the form of a two-dimensional table, and the compensation value in the table varies according to the different positions of the other motion axis; S2: Receive the target position command, perform inverse kinematics based on the ideal kinematics model, calculate the uncompensated initial axis command, call the positioning error compensation data and straightness error compensation data for feedforward calculation, perform the first correction on the initial axis command, calculate the linkage compensation amount of multiple motion axes, perform the second correction on the axis, offset the total prediction deviation in this direction, and finally generate the drive command. S3: Residual visual deviations in workpiece alignment are detected by imaging with an industrial camera; If the residual deviation exceeds the allowable range, the detected visual deviation amount is used as a new input, and feedforward calculation is performed again. Fine-tuning is then performed based on the results to eliminate the residual deviation.

8. The loading device and its control method according to claim 1, characterized in that: In S1, the construction is specifically as follows: The first motion shaft is fixed at different positions. At each fixed position, the second motion shaft is driven to perform a full-stroke motion. The offset of the loading device in the driving direction of the second motion shaft is measured to establish a two-dimensional mapping relationship in which the offset changes together with the positions of the first and second motion shafts.

9. The loading device and its control method according to claim 1, characterized in that: In S2, the feedforward calculation includes multiplying a coupling coefficient related to the actual position of the motion axis by the command value of another motion axis to characterize and offset the motion coupling error caused by the non-parallelism between the motion axes.

10. A loading device and its control method according to claim 1, characterized in that: In step S3, the detected visual deviation is used as a new input for feedforward calculation again, specifically as follows: Using the negative of the residual visual deviation as a temporary position correction target, motion planning and multi-level compensation steps are re-executed to calculate fine-tuning instructions.

Images