An ultra-long part pre-bending feeding positioning device and method

Abstract

The application discloses a kind of super-long part bending front feeding positioning device and method, belong to metal sheet precision forming manufacturing field, method is through high-precision measurement acquisition feeding platform surface full field data, constructs and bending mould center line alignment digitization platform surface model, and generates height constant equal to mould center line global reference line and static platform compensation mapping function.In feeding process, the trajectory deviation of the part bottom surface relative to the reference line is detected in real time using a non-contact displacement sensor array, the static compensation value is combined with feedforward-feedback composite control to drive the active leveling mechanism for vertical adjustment, and the static error and dynamic disturbance of the platform are simultaneously offset.The system continuously monitors the actual trajectory and triggers the instant adjustment mechanism when the deviation exceeds the limit.The application solves the problem of error accumulation in super-long part feeding by establishing a global unified reference, integrating static pre-compensation and dynamic closed-loop control, and constructing multiple verification, achieving high-precision positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of precision forming manufacturing of metal plates, and particularly relates to a feeding and positioning device and method for super-long parts before bending. BACKGROUND

[0002] In the field of metal plate bending processing, especially for super-long parts required in the fields of rail transit and aerospace, the feeding and positioning precision before bending directly determines the accuracy of the bending line, and then affects the subsequent assembly quality. At present, the industry generally follows the technical ideas for conventional short-size parts for the feeding and positioning of such super-long parts, mainly relying on the following ways:

[0003] Mechanical block positioning, rigid blocks are set at the starting end or specific positions of feeding, and the front end of the part is in contact with the block to be considered as positioning. This way is low in cost and simple in operation, but only a single point of contact position can be guaranteed, and the overall attitude change of the part in the super-long stroke due to uneven table surface, self-weight sag and other factors cannot be controlled. The deviation of the middle or tail of the part will directly cause the deflection of the front end after contacting the block, so that there is a difficult-to-measure angle and translation error between the real bending reference line (mold center line) and the theoretical bending line of the part.

[0004] Segmented guide rail guidance, lateral guide rails are set along the feeding path to try to constrain the lateral position of the part. However, super-long parts are often spliced by multiple independent support tables, and the installation straightness and height consistency of each segment of guide rail are difficult to achieve high precision. Small segment misalignment will form a "snake" path, forcing the part to be stuck or generate internal stress in motion, thereby introducing new positioning errors and surface scratch risks.

[0005] Local sensor feedback, photoelectric or contact sensors are installed at the end close to the bending die to detect the instantaneous position of the part and make fine adjustments. The above method only makes "remedial" corrections to the end position at the final moment, and the attitude errors such as distortion and warping accumulated by the part during the feeding process of several meters cannot be completely eliminated at the last moment.

[0006] The prior art indirectly ensures the accurate alignment of the entire super-long part "continuum" with another spatial straight line (mold center line) by controlling the position of one or several discrete "points" or "local line segments". The control strategy of "replacing line with point" and "representing the whole with the part" cannot solve the error accumulation problem under the super-long scale in terms of physical principles.

[0007] The error sources faced by ultra-long parts during the feeding process are continuously distributed and dynamically changing: the multi-segmented feeding table, with its overall flatness, straightness, height differences between segments, and misalignment of joints, constitutes a complex and non-ideal support surface. Traditional methods cannot fully perceive and compensate for this surface error. The elastic deflection of the part itself due to gravity under ultra-long spans, and the dynamic deformation caused by speed changes and inertial forces during movement, mean that its actual axis is not an ideal straight line. Controlling only the positions of its two ends or a few points cannot constrain and correct its overall bending shape. The references relied upon by the aforementioned local positioning methods (such as the position of the stop block, the guide rail mounting surface, and the sensor zero point) are usually not calibrated and associated with the final target, the centerline of the bending die, in a unified, high-precision global coordinate system. This results in the entire positioning process lacking a single, reliable absolute spatial reference throughout.

[0008] As the length of a part increases to a certain extent, the inherent errors of these localized, discrete, and inconsistent control methods will continue to propagate and amplify along the feeding direction. This results in lateral offsets, height deviations, and deflection angles far exceeding the process allowances when the front end of the part reaches the bending die. This not only causes single bending to exceed tolerances, but also makes it impossible to assemble large components that require multiple bends due to the accumulation of errors. Summary of the Invention

[0009] The purpose of this invention is to provide a feeding and positioning device and method for ultra-long parts before bending, which solves the technical problem that the positioning accuracy of ultra-long parts is seriously out of tolerance when the front end of the part reaches the bending die due to the gradual accumulation of geometric errors of the segmented support table and the dynamic deformation of the part itself during the bending feeding process.

[0010] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0011] A method for feeding and positioning ultra-long parts before bending includes the following steps:

[0012] When the equipment is unloaded, full-field geometric data is collected on the entire feeding table to obtain the spatial coordinate point cloud of the table surface; the spatial coordinate point cloud is registered to the absolute coordinate system with the center line of the bending die as the origin to generate a digital table model.

[0013] Based on the digital tabletop model, extract the vertical height coordinate values ​​that are spaced apart along the feeding direction, and construct the original tabletop height distribution function;

[0014] Based on the digital table model and the centerline height of the bending die, a global reference correction algorithm is executed to calculate the static table compensation mapping value at each location point, and a horizontal straight line with a height that is always equal to the centerline height of the bending die is generated as the corrected global reference line.

[0015] In the feeding process, the measured distance value of the bottom surface of the super-long part relative to the fixed reference of the machine tool is detected in real time through a plurality of non-contact displacement sensor arrays arranged above the feeding path; the measured distance value is subtracted from the target value at the corresponding position of the corrected global reference reference line to obtain the trajectory deviation amount;

[0016] The trajectory deviation amount is combined with the static table compensation mapping value of the corresponding position point to generate a positioning correction value sequence on each active leveling mechanism control point in the full path range through a feedforward-feedback composite control. For each control point , the positioning correction value is: ;

[0017] Wherein: : the positioning correction value of the active leveling mechanism control point at time ;

[0018] : the position of the control point of the active leveling mechanism;

[0019] : time variable;

[0020] : the control point corresponding static table compensation mapping function value;

[0021] : proportional control gain coefficient (dimensionless, typical value [5, 20], determined according to the part thickness H and the feeding speed v);

[0022] : integral control gain coefficient (unit 1 / s, typical value [0.1, 1.0], determined according to the part thickness H and the feeding speed v);

[0023] : the trajectory deviation amount (the difference between the part bottom surface and the corrected global reference reference line) of the control point xm at time ;

[0024] : integral variable (characterizing the time integral interval);

[0025] : the integral term of the trajectory deviation amount in the time interval.

[0026] According to the positioning correction value sequence, the active leveling mechanism on each support section is controlled to adjust the position in the vertical direction to offset the influence of the static deviation and dynamic disturbance of the table on the part posture;

[0027] Continuously monitor the actual trajectory of the super-long part during movement, and compare the actual trajectory with the corrected global reference reference line; when the actual trajectory deviates from the corrected global reference reference line by more than a preset threshold, trigger an immediate adjustment mechanism; when the front end of the super-long part approaches the bending die inlet area, capture the part front end edge contour image through a visual recognition module, and extract its geometric center line;

[0028] Calculate the positioning consistency index between the geometric center line and the bending die center line, and start the end fine adjustment program to correct the part front end position when the positioning consistency index does not meet the set accuracy requirement.

[0029] Further, the whole field geometric data acquisition of the whole feeding table surface specifically includes:

[0030] A high-precision laser scanning system or a structured light three-dimensional measurement device is used for data acquisition;

[0031] The measurement device is erected above the table surface perpendicular to the feeding direction and moves along the feeding direction at a constant speed;

[0032] During the acquisition process, ensure that the sampling density meets the requirement of not less than one hundred effective points per unit length.

[0033] Further, the generation of the digitalized table surface model specifically includes:

[0034] The registered spatial coordinate point cloud is reconstructed by a triangulation algorithm;

[0035] The reconstructed table surface model is stored in a regular grid form, and the grid spacing is set to 10 mm;

[0036] Each grid node records its accurate coordinate values in X, Y, and Z three dimensions, wherein the X axis is along the feeding direction, and the Z axis is vertically upward.

[0037] Further, the global reference correction algorithm includes the following steps:

[0038] Read the height value of the pre-calibrated bending die center line in the absolute coordinate system, denoted as ;

[0039] According to the original table height distribution function , calculate the static compensation amount required to adjust each point of the table surface to the height , generate a static table surface compensation mapping function , wherein: ;

[0040] The function The defined horizontal straight line is determined as the corrected global reference line wherein, is a height function of the corrected global reference line.

[0041] Further, the measurement points of the non-contact displacement sensor array correspond one-to-one with the support point positions of the active leveling mechanism in the feeding direction; each group of sensors contains at least three eddy current or laser triangulation units, which are symmetrically arranged in the transverse direction perpendicular to the feeding direction, for synchronously measuring the heights of different transverse positions of the part bottom surface and eliminating single-point measurement errors caused by part transverse yaw or twist by calculating the average value.

[0042] Further, the positioning correction value sequence is generated by combining the trajectory deviation amount and the static table compensation mapping value, specifically:

[0043] For each active leveling mechanism control point , the trajectory deviation amount corresponding to the point is directly obtained and the static table compensation mapping value ;

[0044] The positioning correction value of the control point is calculated based on the following formula :

[0045] ;

[0046] wherein, : time variable;

[0047] : integral term of the trajectory deviation amount in the time interval;

[0048] : integral variable (characterizing the time integral interval);

[0049] (proportional gain) and (integral gain) are control gain coefficients determined according to the part thickness H and the feeding speed v, and the typical value range is: (dimensionless), for example: when the part thickness H = 8 mm and the feeding speed v = 0.5 m / s, take , .

[0050] The positioning correction values of all control points constitute the positioning correction value sequence.

[0051] Further, the instant adjustment mechanism includes:

[0052] The first level local fine-tuning: when the track deviation point is detected, the upstream and downstream groups of active leveling mechanisms closest to the deviation point are activated, and the two groups of mechanisms generate vertical displacements in opposite directions, so that the part at the deviation point receives a corrective torque, thereby suppressing local drift;

[0053] The second level parameter re-tuning: when it is detected that multiple consecutive track points deviate beyond the tolerance band, or a single point deviates more than twice the tolerance threshold, the central cooperative controller starts the online parameter identification algorithm, and based on the real-time track deviation sequence, the static table compensation mapping value is updated, and the updated compensation value is issued to the corresponding active leveling mechanism.

[0054] Further, the front end geometric center line of the part is extracted by a visual recognition module, specifically including:

[0055] An industrial CMOS camera, a telecentric lens and a ring LED light source are used to form a vision system to capture the part front end edge profile image;

[0056] The image is pre-processed by grayscale, Gaussian filtering and contrast enhancement; the sub-pixel edge detection algorithm based on Zernike moment is applied to locate the edge position;

[0057] The extracted edge profile is fitted as a straight line through Hough transform, and the geometric center line is calculated.

[0058] Further, the end fine-tuning program specifically includes:

[0059] The position correction is implemented by controlling the transverse micro-motion platform of the end clamping mechanism of the feeding trolley;

[0060] The transverse micro-motion platform is composed of cross-roller guide and piezoelectric driver, and the stroke range is ±2mm, and the closed-loop control resolution is better than 0.5 microns;

[0061] According to the positioning consistency index, the required transverse translation and rotation are calculated, a composite motion instruction is generated to drive the micro-motion platform; after one adjustment, the visual recognition module is used for retesting again, until the positioning consistency index meets the set accuracy requirement.

[0062] In addition, the application also discloses a super-long part feeding positioning device before bending, which is used for realizing the super-long part feeding positioning method before bending.

[0063] The digital modeling and datum establishing module is configured to collect full-field geometric data of the whole feeding table in the empty state of the equipment, and obtain a spatial coordinate point cloud of the table surface; the spatial coordinate point cloud is registered to an absolute coordinate system with the center line of the bending die as the origin, and a digital table model is generated; and based on the digital table model and the height of the center line of the bending die, a horizontal straight line with a constant height equal to the height of the center line is generated as a corrected global reference line, and a static table compensation mapping function is calculated;

[0064] The real-time feeding positioning compensation module is configured to detect the trajectory deviation of the bottom surface of the super-long part relative to the corrected global reference line in real time through a non-contact displacement sensor array during the feeding process; the trajectory deviation and the static table compensation mapping function value of the corresponding position are combined to generate a positioning correction value sequence, and the vertical position adjustment of the active leveling mechanism is controlled;

[0065] The feeding trajectory monitoring and instant adjustment module is configured to continuously monitor the actual feeding trajectory of the super-long part, and compare it with the corrected global reference line; when the actual trajectory deviates by more than a preset threshold, an instant adjustment mechanism is triggered;

[0066] The end visual fine adjustment module is configured to extract the geometric center line of the super-long part through a visual recognition module when the front end of the super-long part approaches the bending die entrance area; the positioning consistency index between the geometric center line and the center line of the bending die is calculated, and the end fine adjustment program is started for correction when the index does not meet the set accuracy requirement.

[0067] Compared with the prior art, the present application has the following beneficial effects:

[0068] The traditional method relies on multiple local and discrete physical or sensing references for segmented positioning, and the technical essence is to approximate a global path by connecting multiple independent coordinate systems, resulting in inevitable amplification of static deviations such as manufacturing and installation errors, thermal deformation of each support segment, and dynamic disturbances such as part self-weight deflection in the transmission process. The present application discards this kind of indirect logic of approximating the whole from the local, and directly constructs a corrected global reference line strictly aligned with the center line of the bending die and running through the whole feeding path through full-field digital modeling before feeding. The reference line is defined in an absolute coordinate system, providing a unique, continuous and high-precision spatial reference for the whole motion of the super-long part, and completely eliminating the systematic cumulative error caused by reference conversion and segmented splicing from the technical principle level.

[0069] The present application realizes the deep integration of static pre-compensation and dynamic closed-loop control, and changes passive adaptation to active regulation: the prior art mainly adopts passive acceptance or local remedy for the geometric error of the table surface. The present application accurately quantifies and stores the static geometric deviation of the whole path by calculating the static table compensation mapping function, and eliminates it through feedforward control at the start of feeding. At the same time, feedback control is performed in combination with the trajectory deviation obtained by real-time sensing, to form a feedforward-feedback composite control law, thereby suppressing dynamic disturbance in real time. The system is changed from passive response to uncontrollable error to active and accurate regulation based on accurate model and real-time sensing, thereby improving the anti-interference ability and overall precision stability of the system.

[0070] The present application ensures the controlled state of the intermediate process by continuously monitoring the actual feeding trajectory and comparing it with the corrected global reference line in real time and adjusting it in real time. When the front end of the super-long part approaches the bending die inlet area, an independent visual recognition and end fine adjustment program is introduced to measure and correct the absolute position of the final pose, forming a complete verification closed loop from the process to the end. The positioning consistency index between the geometric center line of the part front end and the center line of the bending die is strictly controlled, thereby providing reliable input conditions for the subsequent bending process. The present application reconstructs the reference establishment and error control mode of the feeding positioning of the super-long part, and solves the technical problems caused by the dispersion of the local reference and the accumulation of errors along the way. BRIEF DESCRIPTION OF DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0072] Figure 1 The present application is a method for establishing a reference for feeding positioning of a super-long part.

[0073] Figure 2 The present application is a digital modeling and reference establishment step flowchart.

[0074] Figure 3 The present application is a real-time feeding positioning compensation step flowchart.

[0075] Figure 4 The present application is a device overall architecture block diagram. DETAILED DESCRIPTION

[0076] In the following, only certain exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the embodiments of the present application. Therefore, the drawings and the description are considered to be exemplary in nature rather than limiting.

[0077] Embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0078] Embodiment 1: Referring to Figures 1-4 This embodiment discloses a method for feeding and positioning a super-long part before bending. A digital table geometry model covering the entire feeding path is established, and a global reference line strictly aligned with the center line of the bending die is established on this basis. Through multi-source sensing fusion, dynamic compensation control and closed-loop trajectory verification mechanism, high-precision front-end positioning of the super-long part before entering the bending station is realized.

[0079] The specific implementation of the present application will be described in detail below in connection with system composition, data modeling, reference correction, real-time correction, trajectory monitoring and end fine adjustment.

[0080] First, under the condition that the equipment is in an empty state and the ambient temperature is stable, a high-precision laser scanning system or a structured light three-dimensional measurement device is started to collect geometric data of the entire feeding table. The measurement device is erected on the overhead beam above the table perpendicular to the feeding direction, with a spatial resolution of sub-millimeter and a repeated measurement accuracy of better than ±0.02 millimeters. During the scanning process, the device moves at a constant speed along the feeding direction, and the point cloud data is captured synchronously to ensure that the sampling density meets the requirement of not less than one hundred effective points per unit length. After preliminary filtering and denoising, the obtained spatial coordinate point cloud data is transmitted to the central collaborative controller.

[0081] After receiving the original point cloud data, the central collaborative controller performs coordinate system registration operation. The registration process takes a fixed physical marker point on the center line of the bending die as the origin, defines the X axis to extend along the feeding direction, the Y axis to point horizontally to the die symmetry plane, and the Z axis to be vertical upward to form a right-handed rectangular coordinate system. All table sampling points are converted to this absolute coordinate system through a rigid body transformation matrix to form a registered three-dimensional point cloud set. Subsequently, a triangular subdivision algorithm is used to reconstruct the surface of the point cloud set to generate a digital table model containing height deviation, flatness error and misalignment information at the joint of adjacent support sections. The model is stored in a regular grid form, with a grid spacing of ten millimeters. Each grid node records its accurate coordinate values in X, Y and Z dimensions, and is solidified in the non-volatile memory of the central collaborative controller as a basis for subsequent positioning calculations.

[0082] Further, based on the digitalized table surface model, Z coordinate values at each position are extracted along the X axis direction at intervals of five millimeters, and an original table surface height distribution function is constructed wherein L is the total length of the feeding path. The function completely characterizes the actual spatial coordinates of the physical table surface in the full length range before compensation. On this basis, a global reference correction algorithm is performed to determine an ideal reference line extending through the starting end to the bending end.

[0083] The algorithm specifically includes the following steps:

[0084] S1: mold reference calibration. In the equipment installation stage, the spatial coordinates of feature points on the bending mold center line are measured using a laser tracker, and the Z coordinate values thereof are calibrated and stored as the mold center line height reference. .

[0085] S2: static table surface compensation mapping calculation. The central collaborative controller calculates the static compensation amount of the full path according to the height function in the digitalized table surface model and the calibration value . For any position on the feeding path, the static table surface compensation mapping value is given by the following formula:

[0086] ;

[0087] The physical meaning is: in order to make the table surface at position reach the same height as the mold center line , the active leveling mechanism at this position needs to be adjusted by a distance of ( represents lifting, represents lowering).

[0088] S3: determination of the global reference reference line. The corrected global reference reference line is a horizontally straight line that is clearly defined in space, and its equation is uniquely determined by the mold center line height reference:

[0089] wherein, is the height function of the corrected global reference reference line;

[0090] This straight line is the target trajectory that the super-long part needs to track during feeding.

[0091] Therefore, the corrected global reference reference line is defined as a horizontal straight line with a height equal to , and its function expression is ​This reference line represents the target spatial trajectory that the extra-long part needs to be tracked during the feeding process. Based on this, the system initiates a real-time positioning correction process. It will be stored in the central coordinating controller for subsequent feedforward compensation control.

[0092] On the corrected reference line (Right now After the position is established, the system initiates a real-time positioning correction process. This process is executed by multiple arrays of non-contact displacement sensors arranged above the feeding path. The installation positions of the non-contact displacement sensor arrays are strictly aligned with the support points of each active leveling mechanism in the feeding direction (X-axis), ensuring that there is a corresponding sensor above each support point for measurement, thereby achieving a one-to-one correspondence between measurement points and control points. Each sensor array contains at least three eddy current or laser triangulation units, arranged in a laterally symmetrical layout to eliminate measurement deviations caused by lateral sway of the part. Each sensor continuously detects the distance between the bottom surface of the extra-long part and the machine tool's fixed reference frame at its corresponding X-position. The value is then transmitted to the central co-controller via the industrial real-time Ethernet bus.

[0093] The central coordinating controller has the corresponding location stored internally. Target feeding trajectory value Mapping values ​​of static tabletop compensation The controller first calculates the current trajectory deviation (feedback error):

[0094] All sensor data are aligned via a time synchronization module to ensure that each It has a unified timestamp and a synchronization accuracy better than 100 microseconds.

[0095] The central coordinating controller receives trajectory deviations from each sensor. Then, combined with the pre-stored static table compensation mapping Generate a sequence of location correction values ​​across the entire path. ,in , The locations of the control points corresponding to each active leveling mechanism;

[0096] in:

[0097] The location of the trajectory detection point corresponding to each sensor ( (Sensor serial number).

[0098] t: Real-time time during the feeding process (reflecting the time attribute of dynamic compensation);

[0099] : testing point Real-time trajectory deviation at time t

[0100] Pre-stored static table compensation mapping

[0101] Control point position corresponding to each active leveling mechanism (m is the leveling mechanism number)

[0102] Real-time positioning correction value of control point xm at time t

[0103] Total number of active leveling mechanisms

[0104] Positioning correction value on each active leveling mechanism control point is generated by superimposing the feedforward compensation and the feedback compensation:

[0105]

[0106] , wherein is the feedforward compensation from the static mapping, used to offset the known table geometric error; is the proportional feedback term, is the integral feedback term, both of which constitute the feedback compensation, used to real-time suppress the dynamic trajectory deviation caused by the part weight, inertia and external disturbance. and are pre-set control gain coefficients.

[0107] The generation of this sequence adopts the method of interpolation combined with extrapolation: for the area between the sensors, and are estimated by cubic spline interpolation respectively; for the starting and ending areas, linear extrapolation is carried out based on the boundary conditions.

[0108] This compensation instruction is issued to each segment independent servo drive unit. Each segment of the support area is installed with an active leveling mechanism, which is composed of a precision ball screw pair driven by a servo motor. Its vertical stroke is ±5mm, and the positioning resolution is better than 1 micron, which is used to execute the precise position adjustment instructed by the positioning correction value sequence, and the positioning repeatability is better than ±1 micron. Two kinds of actuators are selected and configured according to the position of the support segment and the expected disturbance characteristics. The output end of the actuator is directly connected to the bottom of the table support plate, and the local table height is changed in real time by adjusting the micron-level vertical displacement, so as to offset the influence of the original manufacturing and installation errors on the part attitude.

[0109] ​​​Meanwhile, the system continuously monitors the actual trajectory of the super-long part during movement. The trajectory is calculated by the inertial measurement unit installed on the feeding trolley combined with the laser tracker measurement system. Specifically, by installing the laser tracker on the machine tool fixed reference, the three-dimensional coordinates of the optical target installed on the feeding trolley are measured in real time, providing absolute position information; the inertial measurement unit provides high-frequency angular velocity and acceleration data. Both data are fused through the extended Kalman filtering algorithm. The state vector contains position, velocity, attitude angle and its bias term. Among them, the angular velocity and acceleration output of the inertial measurement unit is used as the basis for system state prediction; the three-dimensional coordinate data provided by the laser tracker is used as the observation value to correct the predicted state. The fused output is a continuous and stable six-degree-of-freedom pose estimation ;

[0110] wherein:

[0111] : the spatial pose state vector of the part (or the feeding execution mechanism) at time t (contains position and attitude information);

[0112] : the position coordinate of the part in the feeding direction (X axis) at time t (unit: mm, consistent with the definition of the feeding path coordinate x);

[0113] : the position coordinate of the part in the horizontal transverse direction perpendicular to the feeding direction (Y axis) at time t (unit: mm);

[0114] : the height coordinate of the part in the vertical direction (Z axis) at time t (unit: mm, consistent with the height dimension of the table surface);

[0115] : the rotation angle of the part around the X axis (feeding direction) at time t (lateral roll angle, unit: rad / °, representing the torsion deviation of the part along the feeding direction);

[0116] : the rotation angle of the part around the Y axis (horizontal transverse direction) at time t (pitch angle, unit: rad / °, representing the inclination deviation of the part in the vertical direction);

[0117] : the rotation angle of the part around the Z axis (vertical direction) at time t (yaw angle, unit: rad / °, representing the turning deviation of the part in the horizontal direction);

[0118] : the transpose operation of the vector (converts the row vector to the column vector, consistent with the common representation form of the state vector in the control field).

[0119] The pose estimation result is projected to the X-Z plane in real time to extract its longitudinal trajectory , and compared with the corrected global reference line (point by point), i.e. the constant height .

[0120] If at any position satisfies: ;

[0121] wherein:

[0122] : the actual trajectory height of the super-long part at position ;

[0123] : the position of the deviation detection point on the feeding path;

[0124] : the preset tolerance threshold (typical value: ±0.1 mm), which triggers the instant adjustment mechanism.

[0125] The instant adjustment mechanism includes two levels of operations. The first level is local fine-tuning: after the system identifies the deviation point , it activates its nearest neighbor two sets of active leveling mechanisms (positions and respectively), calculates the required reverse compensation torque according to the deviation direction and amplitude, and realizes the torque by coordinating the displacement of the two actuators. For example, if the part is arched upwards at , the actuator at is slightly lowered, and the actuator at is slightly raised to form a downward pressure torque to suppress local drift. The second level is global re-planning: when more than five consecutive sampling points on the intermediate path exceed the tolerance band, or a single point deviates more than twice the tolerance threshold, the central cooperative controller determines that the state of the table has changed significantly, and immediately re-invokes the global reference correction algorithm. At this time, the system can selectively start the fast scanning mode, only for local point cloud resampling in the affected area, or call historical data combined with the current sensor readings for model updating. The updated digital table model is used to recalculate and , and generate a new positioning correction value sequence based on the new , which is issued to all execution units to complete the dynamic reconstruction of the reference line.

[0126] Finally, the system performs a last check of the accuracy of the positioning when the front end of the super-long part approaches the entrance area of the bending die (usually defined as the interval less than 500 mm from the center line of the die). At this stage, a high-resolution visual recognition module is enabled, which is composed of an industrial CMOS camera, a telecentric lens, and a ring-shaped LED light source, installed directly above the entrance of the die, covering a field of view of 50 mm in the front end of the part. The camera captures the edge profile image of the front end of the part at a rate of 30 frames per second, with an image resolution of 5000 pixels per mm.

[0127] The central cooperative controller pre-processes the image, including grayscale, Gaussian filtering, and contrast enhancement. Then the sub-pixel edge detection algorithm is applied: specifically, the edge positioning method based on Zernike matrix is adopted, whose principle is to expand the local area of the image with orthogonal polynomials, and estimate the edge sub-pixel position through the ratio of zero-order and first-order moments. This algorithm can improve the edge positioning accuracy to within 0.1 pixels. The extracted front edge profile is fitted into a straight line by Hough transform, and its geometric center line .

[0128] center line of the bending die The position in the visual coordinate system is determined by the early calibration, and its equation is known. The system calculates the shortest distance and the angle deviation between the two straight lines in the X-Y plane. The positioning consistency index is defined as:

[0129] ;

[0130] wherein is the angle-displacement equivalent coefficient, which is set according to the experience of part thickness and bending radius. For example, the value is 1 / 1000 of the part thickness H (unit: mm / rad), for example: when the part thickness H = 8 mm, k = 0.008 mm / rad; when the bending radius is greater than 10 mm, it can be corrected as k = 0.01 mm / rad.

[0131] If does not meet the set accuracy requirement (for example mm), the end fine-tuning program is started.

[0132] The end fine-tuning program is implemented by controlling the transverse micro-motion platform of the end clamping mechanism of the feeding trolley. The platform is composed of cross-roller guide and piezoelectric driver, with a travel range of ±2 mm and a closed-loop control resolution better than 0.5 microns.

[0133] The controller corrects and Decouple the X-direction translation and the rotation around the Z-axis to generate a composite motion command. After the platform executes the command, the vision recognition module is triggered again for re-measurement until the accuracy requirements are met, ensuring that the part front-end centerline and the bending die centerline are completely aligned in space.

[0134] During the entire feeding process, all sensor data, control commands, and state feedback are transmitted through an industrial real-time Ethernet bus (such as EtherCAT or PROFINET IRT). The network topology adopts a linear or star structure, with the central cooperative controller as the master station and the sensor interface modules, servo drives, and vision processing units as the slave stations. The communication cycle is set to one millisecond, and the time synchronization uses a distributed clock mechanism to ensure that the time synchronization accuracy between subsystems is better than one hundred microseconds, meeting the timing requirements of high-speed dynamic compensation.

[0135] In addition, the central cooperative controller has a built-in fault diagnosis module that uses a hybrid diagnostic strategy based on model and data driving. For sensors, failure is detected through residual analysis: if a sensor reading consistently deviates from its neighborhood mean by more than three standard deviations, it is marked as abnormal; for communication links, interruptions are monitored through a heartbeat packet mechanism; for actuators, stuck or hysteresis faults are identified by comparing current-displacement characteristic curves. Once an abnormal event is detected, the system automatically switches to a redundant channel (such as a backup sensor or communication line) or runs in a degraded mode (such as pausing feeding and maintaining the current position), while sending fault codes and recommended treatment measures to the operation terminal, ensuring the safety and robustness of the positioning process.

[0136] In one specific embodiment, a 2024 aluminum alloy plate with a length of 2800 mm and a thickness of 2 mm is chosen as the processed part. The feeding table is composed of six independent support units, with a total length of thirteen meters. First, a laser scan is performed under no load to generate a digital table model, and the full-length flatness error (peak-to-peak value) of the original table is measured to be 2.7 mm (i.e., the difference between the highest point and the lowest point). After processing by the global reference correction algorithm, the calculated static table compensation mapping has a value range of [-1.8, +0.9] mm, and its peak-to-peak value is also 2.7 mm. During feeding, the displacement sensors arranged at each support point update the trajectory deviation at a ten-millisecond cycle, and the active leveling mechanism (precision ball screw pair) implements dynamic compensation based on the combined feedforward and feedback command. The trajectory monitoring display shows that the maximum deviation of the actual trajectory of the part from the global reference line is controlled within ±0.05 mm. In the end-of-vision verification stage, the initial positioning consistency index is 0.06 mm, which is reduced to 0.03 mm after a slight adjustment, meeting the bending process requirements.

[0137] To verify the technical effect of the present application, the following Comparative Example 1 was conducted.

[0138] Comparative Example 1: A traditional segmented reference transmission method was adopted, that is, only a mechanical stopper was arranged at the starting end as a positioning reference, no active compensation was made in each segment, and the manufacturing accuracy of the table itself was relied on. The same parts were fed on the same equipment. Due to the cumulative error of the table and the deflection of the parts under their own weight, when the front end of the part reached the bending station, the position thereof had a transverse deviation of 0.65 mm and a longitudinal height deviation of 0.42 mm relative to the center line of the die, and the positioning consistency index was 0.42 mm, far beyond the process allowable range (usually required to be within 0.1 mm), resulting in an angle out-of-tolerance of the bent part, which needed to be reworked and corrected.

[0139] Table 1 summarizes the comparison results of the key performance indicators of Example 1 and Comparative Example 1:

[0140] Table 1:

[0141]

[0142] The above data show that, by constructing a full-path digital model, implementing dynamic compensation and closed-loop verification, the present application effectively overcomes the cumulative error problem in the feeding process of the super-long part, and significantly improves the positioning accuracy and process stability.

[0143] The super-long part feeding positioning method according to the present application forms a complete high-precision positioning solution through systematic data acquisition, model construction, reference correction, real-time correction, trajectory monitoring and end fine adjustment. The technical implementation does not depend on idealized table manufacturing conditions, but through active sensing and intelligent control, the uncertainties in the physical world are brought into the controllable range, so as to ensure that the super-long part can still achieve micron-level positioning accuracy in a complex industrial environment, thereby providing reliable technical support for high-quality bending forming of large structural parts.

[0144] The present application solves the problem of cumulative positioning deviation along the way caused by table manufacturing and installation errors, thermal deformation and deflection under self-weight. The method constructs a digital table model covering the full feeding path through high-precision scanning, and establishes a global reference line based on the center line of the bending die; combined with multi-point non-contact displacement sensing and real-time trajectory monitoring, compensation instructions are dynamically generated to drive the active leveling mechanism for micron-level correction; at the same time, multi-source sensing fusion and closed-loop control strategy are adopted to realize high-precision unified reference positioning from the starting end to the bending end, so as to ensure that the front end of the part is strictly aligned with the center line of the bending die.

[0145] Example 2: The present embodiment also discloses a super-long part feeding positioning device before bending, which comprises:

[0146] ​The digital modeling and reference establishing module is configured to collect full-field geometric data of the whole feeding table surface under the no-load state of the equipment, and obtain a spatial coordinate point cloud of the table surface; the spatial coordinate point cloud is registered into an absolute coordinate system with the center line of the bending die as the origin, to generate a digital table surface model; and based on the height of the digital table surface model and the center line of the bending die, a horizontal straight line with a constant height equal to the height of the center line is generated as a corrected global reference line, and a static table compensation mapping function is calculated;

[0147] The real-time feeding positioning compensation module is configured to detect the trajectory deviation of the bottom surface of the super-long part relative to the corrected global reference line in real time through a non-contact displacement sensor array during the feeding process; a positioning correction value sequence is generated by combining the trajectory deviation and the static table compensation mapping function value at the corresponding position, and the vertical position adjustment of the active leveling mechanism is controlled;

[0148] The feeding trajectory monitoring and instant adjustment module is configured to continuously monitor the actual feeding trajectory of the super-long part, and compare it with the corrected global reference line; when the actual trajectory deviates by more than a preset threshold, an instant adjustment mechanism is triggered;

[0149] The end visual fine adjustment module is configured to extract the geometric center line of the super-long part when the front end of the super-long part approaches the bending die entrance area, through a visual recognition module; a positioning consistency index between the geometric center line and the center line of the bending die is calculated, and when the index does not meet the set accuracy requirement, an end fine adjustment program is started to correct.

[0150] Although the preferred embodiments of the present application have been described, those skilled in the art who are familiar with the basic inventive concept can make further changes and modifications to the embodiments. Therefore, the appended claims are intended to include the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0151] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. It should be pointed out that any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for feeding and positioning ultra-long parts before bending, characterized in that, Includes the following steps: Digital modeling and benchmark establishment steps: Under no-load conditions, collect full-field geometric data of the entire feeding table surface to obtain the spatial coordinate point cloud of the table surface; register the spatial coordinate point cloud to the absolute coordinate system with the center line of the bending die as the origin to generate a digital table surface model; based on the height of the digital table surface model and the center line of the bending die, generate a horizontal straight line with a height always equal to the height of the center line as the global benchmark reference line after correction, and calculate the static table surface compensation mapping function; Real-time feeding and positioning compensation steps: During the feeding process, the trajectory deviation of the bottom surface of the extra-long part relative to the corrected global reference line is detected in real time by a non-contact displacement sensor array; combined with the trajectory deviation and the static table compensation mapping function value at the corresponding position, a positioning correction value sequence is generated, and the active leveling mechanism is controlled to adjust the vertical position. Feed trajectory monitoring and real-time adjustment steps: continuously monitor the actual feeding trajectory of the extra-long part and compare it with the corrected global reference line; When the actual trajectory deviates from the preset threshold, an immediate adjustment mechanism is triggered; End-of-line visual fine-tuning steps: When the front end of the extra-long part approaches the entry area of ​​the bending die, its geometric center line is extracted by the visual recognition module; the positioning consistency index between the geometric center line and the center line of the bending die is calculated, and if the positioning consistency index does not meet the set accuracy requirements, the end-of-line fine-tuning program is started for correction.

2. The feeding and positioning method for ultra-long parts before bending according to claim 1, characterized in that, In the digital modeling and benchmark establishment steps, generating the corrected global benchmark reference line specifically includes: Read the pre-calibrated centerline height value of the bending die ; Based on the original tabletop height distribution function obtained from the digital tabletop model Calculate the static platform compensation mapping function. ; Function The defined horizontal straight line is determined as the corrected global reference line. ,in, This is the height function of the corrected global reference line.

3. The feeding and positioning method for ultra-long parts before bending according to claim 1, characterized in that, In the real-time feeding positioning compensation step, the specific steps for generating the positioning correction value sequence are as follows: For each control point of the active leveling mechanism Obtain the trajectory deviation corresponding to that point. Mapping function value with static table compensation ; The positioning correction value of the control point is calculated based on the following formula. : ； in, Time variable; The integral term of the trajectory deviation over the time interval; : Integral variable, representing the time integration interval; The proportional control gain coefficient is determined based on the part thickness H and the feeding speed v. The integral control gain coefficient is determined based on the part thickness H and the feeding speed v; the positioning correction values ​​of all control points constitute the positioning correction value sequence.

4. The feeding and positioning method for ultra-long parts before bending according to claim 1, characterized in that, The positions of the measurement points of the non-contact displacement sensor array in the feeding direction correspond one-to-one with the positions of the support points of the active leveling mechanism.

5. The feeding and positioning method for ultra-long parts before bending according to claim 4, characterized in that, Each sensor group contains at least three ranging units arranged symmetrically in the lateral direction to simultaneously measure the height of different lateral positions on the bottom surface of the part, thereby eliminating measurement errors caused by the lateral sway of the part.

6. The feeding and positioning method for ultra-long parts before bending according to claim 1, characterized in that, Extracting the geometric center line of the front end of the part through the visual recognition module specifically includes: Capture the outline image of the front edge of the part; After image preprocessing, a subpixel edge detection algorithm is applied to locate edges; The extracted edge contours are fitted to a straight line to obtain the geometric center line.

7. The feeding and positioning method for ultra-long parts before bending according to claim 1, characterized in that, The end fine-tuning program is implemented by controlling the lateral micro-motion platform at the end of the feeding trolley. The lateral micro-motion platform performs a composite motion based on the translation and rotation calculated by the positioning consistency index.

8. A method for feeding and positioning ultra-long parts before bending according to any one of claims 1-7, characterized in that, The acquisition of full-field geometric data for the entire feeding platform is accomplished by using high-precision laser scanning or structured light three-dimensional measurement, and by moving and scanning along a direction perpendicular to the feeding direction.

9. A feeding and positioning device for ultra-long parts before bending, used to implement the feeding and positioning method for ultra-long parts before bending according to any one of claims 1-8, characterized in that, include: The digital modeling and benchmark establishment module is configured to collect full-field geometric data of the entire feeding table surface under no-load conditions, and obtain the spatial coordinate point cloud of the table surface; register the spatial coordinate point cloud to an absolute coordinate system with the center line of the bending die as the origin to generate a digital table surface model; and generate a horizontal straight line with a height that is always equal to the height of the center line based on the height of the digital table surface model and the center line of the bending die as a corrected global benchmark reference line, while calculating the static table surface compensation mapping function. The real-time feeding and positioning compensation module is configured to detect the trajectory deviation of the bottom surface of the ultra-long part relative to the corrected global reference line in real time during the feeding process using a non-contact displacement sensor array; combine the trajectory deviation with the static table compensation mapping function value at the corresponding position to generate a positioning correction value sequence, and control the active leveling mechanism to adjust the vertical position. The feeding trajectory monitoring and real-time adjustment module is configured to continuously monitor the actual feeding trajectory of the extra-long part and compare it with the corrected global reference line; when the actual trajectory deviates from the preset threshold, the real-time adjustment mechanism is triggered. The end vision fine-tuning module is configured to extract the geometric center line of the ultra-long part by means of a vision recognition module when the front end of the part approaches the entry area of ​​the bending die. The positioning consistency index between the geometric center line and the bending die center line is calculated, and if the positioning consistency index does not meet the set accuracy requirements, the end fine-tuning program is started for correction.

Images