An online automatic detection and correction method for flatness defects in aluminum single-panel panels
By using a segmented multi-roller leveling machine and a laser scanner to generate an absolute space zero point and a 2D coordinate system for the material, combined with a servo hydraulic system, the problems of frame vibration and hydraulic delay in aluminum panel leveling equipment are solved, enabling precise detection and correction of aluminum panel surfaces and improving the stability and accuracy of the equipment.
Patent Information
- Application Number
- CN202611072296.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-11-10
AI Technical Summary
Existing multi-roller leveling equipment for aluminum single panels suffers from frame vibration disturbance and material conveying slippage during operation. This causes the three-dimensional topography data of the panel surface collected by the sensor to be superimposed with common mode displacement error, making it difficult to achieve accurate positioning. Furthermore, the response delay of the servo hydraulic actuator leads to mismatch of leveling command position, making it impossible to achieve dynamic closed-loop compensation for the entire panel and easily causing damage to the mechanical components of the equipment.
A segmented multi-roller leveling machine is adopted, combined with inlet and outlet span-width laser line scanners, photoelectric trigger switches, and actual speed measuring wheels to generate an absolute spatial horizontal zero position and a material 2D coordinate system. Data processing is performed by a control computer to generate servo displacement command vectors, compensate for hydraulic response lag, and achieve alignment between action and material position. The static influence coefficient matrix is updated through a recursive algorithm to prevent mathematical accumulation error.
It enables online automatic detection and correction of flatness defects in aluminum single-panel panels, eliminates the influence of frame vibration and conveyor belt slippage, ensures accurate execution of leveling commands, prevents overloading of mechanical structures, and improves the stability and accuracy of the equipment.
Smart Images

Figure CN123065387A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of metal sheet processing technology, specifically to an online automatic detection and correction method for flatness defects on aluminum single-panel surfaces. Background Technology
[0002] Existing multi-roller leveling equipment for aluminum single panels suffers from frame vibration disturbance and material conveying slippage during operation. Frame vibration disturbance causes common-mode displacement errors to be superimposed on the 3D topographic data of the panel surface collected by the sensors, while material conveying slippage causes longitudinal positioning deviation of the panel. The superposition of frame vibration disturbance and material conveying slippage results in the front and rear 3D topographic data being mapped to different spatial references, making it impossible to accurately position them within a unified 2D coordinate system of the material.
[0003] In assessing flatness defects, existing methods struggle to quantitatively convert the local undulations of aluminum panels into the required elongation under pressure for each lateral action zone of a segmented multi-roller leveling machine. Due to the inherent response delay of the servo hydraulic actuator, it cannot align with the moving aluminum panel after the action is issued, resulting in a mismatch in the leveling command execution position. The material's inherent elastoplastic nonlinear deformation characteristics, coupled with nonlinear time-varying disturbances caused by mechanical wear, make it difficult for control systems based on static mapping correlation equations to adapt to time-varying conditions, easily leading to deformation distribution deviations. Existing control models lack boundary projection constraints corresponding to the mechanical stiffness limit during mathematical recursive updates, posing a risk of exceeding the limits of mathematical cumulative errors and generating correlation mapping coefficients beyond the actual mechanical structure's bearing capacity. This prevents the achievement of dynamic closed-loop compensation for the entire panel and easily damages the equipment's mechanical components.
[0004] Therefore, this invention proposes an online automatic detection and correction method for flatness defects on aluminum single-panel panels to address the shortcomings of existing technologies. Summary of the Invention
[0005] The technical problem solved by this invention is that existing multi-roller leveling equipment for aluminum panels suffers from frame vibration disturbance and material conveying slippage during operation. Furthermore, the nonlinear elastic-plastic deformation of the material and the response delay of the hydraulic actuator make it difficult for the static mapping model to adapt to time-varying working conditions, resulting in mismatch of leveling command issuance position and deformation distribution deviation. Consequently, it is impossible to achieve global scanning and global optimal attitude calculation for a single aluminum panel.
[0006] To address the problems existing in the prior art, the present invention provides the following technical solution: This invention provides an online automatic detection and correction method for flatness defects on aluminum single-panel panels. The hardware structure includes a segmented multi-roller leveling machine. The segmented multi-roller leveling machine is equipped with flexible working rollers and multiple sets of horizontally parallel segmented support shoes. Each set of segmented support shoes is connected to a servo hydraulic cylinder on its back. The multiple sets of servo hydraulic cylinders form a servo hydraulic cylinder array, which is connected to and controls the segmented support shoes to push the flexible working rollers to produce deformation.
[0007] A horizontally fixed inlet span laser line scanner is installed at the inlet of the segmented multi-roll leveler, and a horizontally fixed outlet span laser line scanner is installed at the outlet. A left and right solid reference target strip are fixedly installed on the follower bearing seat of the inlet conveyor belt, and a left and right solid reference target strip are fixedly installed on the follower bearing seat of the outlet conveyor belt. This ensures that the solid reference target strips on both sides have common-mode mechanical vibration with the conveyor belt support surface. The scanning field of the inlet span laser line scanner covers the left solid reference target strip, the surface of the aluminum panel, and the right solid reference target strip. The scanning field of the outlet span laser line scanner covers the left solid reference target strip, the surface of the aluminum panel, and the right solid reference target strip.
[0008] The segmented multi-roller leveler is equipped with a photoelectric trigger switch and a measuring speed wheel at the inlet. The photoelectric trigger switch is aligned with the aluminum panel feeding track, and the measuring speed wheel is in contact with the surface of the aluminum panel. The photoelectric trigger switch is connected to the measuring speed wheel and collects the displacement data of the aluminum panel. The control computer is connected to the segmented multi-roller leveler, the inlet span width laser line scanner, the outlet span width laser line scanner, the photoelectric trigger switch, and the measuring speed wheel via cables. The control computer receives height scanning data collected by the inlet span width laser line scanner and the outlet span width laser line scanner, and simultaneously receives displacement data collected by the photoelectric trigger switch and the measuring speed wheel. The control computer then sends servo displacement commands to the servo hydraulic cylinder.
[0009] The online automatic detection and correction method for flatness defects in aluminum single-panel surfaces executes the following control logic in chronological order: The entrance span-width laser line scanner synchronously scans the left solid reference target strip, the surface of the aluminum single panel, and the right solid reference target strip to obtain height scanning data. The control computer generates an absolute spatial horizontal zero point based on the height scanning data of the left and right solid reference target strips. The control computer extracts the difference between the height scanning data and the absolute spatial horizontal zero point to generate the entrance pure panel height matrix. The photoelectric trigger switch detects the leading edge of the aluminum single panel to generate a zero-point trigger signal. The actual measurement speed wheel measures the running displacement of the aluminum single panel. The control computer combines the zero-point trigger signal and the running displacement to generate a material 2D coordinate system. The control computer binds the entrance pure panel height matrix to the material 2D coordinate system. The control computer divides the entrance pure plate height matrix into multiple longitudinal fibers along the aluminum single plate conveying direction. The control computer performs curve discrete integral calculation on each longitudinal fiber to obtain the actual arc length. The control computer extracts the actual arc length with the largest value as the reference base. The control computer combines the actual arc length and the reference base to calculate the required elongation of each longitudinal fiber. The control computer summarizes all the required elongation rates to generate an input vector. The control computer multiplies the input vector with the generalized inverse matrix of the static influence coefficient matrix to generate a servo displacement command vector. The control computer establishes an action delay scheduling queue in conjunction with the material 2D coordinate system. When the coordinate position recorded in the material 2D coordinate system reaches directly below the segmented support shoe, the control computer sends a servo displacement command vector to the servo hydraulic cylinder. The laser line scanner at the exit span simultaneously scans the solid reference target strip on the left side of the exit, the surface of the aluminum single plate, and the solid reference target strip on the right side of the exit to obtain exit height scanning data. The control computer obtains the exit pure plate surface height matrix based on the exit height scanning data. The control computer retrieves the material 2D coordinate system and performs spatial registration between the exit pure plate surface height matrix and the target flat state ideal height matrix at the same coordinate position. The control computer compares the exit pure plate surface height matrix and the target flat state ideal height matrix to generate a topological residual matrix. The control computer extracts the topological residual matrix and the actual execution displacement data fed back by the servo hydraulic cylinder. The control computer uses a recursive algorithm to combine the topological residual matrix and the actual execution displacement data to update the static influence coefficient matrix and generate the dynamic influence coefficient matrix. The control computer applies boundary projection constraints to the dynamic influence coefficient matrix. When the elements inside the dynamic influence coefficient matrix exceed the preset limit, the control computer maintains the static influence coefficient matrix.
[0010] The control computer uses spatial difference operations to subtract the absolute spatial horizontal zero from the height scan data, filtering out low-frequency common-mode displacement disturbances that are rigidly coupled with the frame system to generate the entrance pure plate height matrix; combined with the photoelectric trigger switch to detect the leading edge of the aluminum plate and the actual measured speed wheel to accumulate the running displacement to eliminate conveyor belt slippage, the control computer assembles the lateral pixel index and running displacement into a material 2D coordinate system, so that the front and rear three-dimensional topography data are mapped to the same spatial reference.
[0011] The control computer divides the entrance pure panel height matrix into multiple longitudinal fibers. The control computer performs curve discrete integration on each longitudinal fiber and accumulates the spatial straight-line distance to obtain the actual arc length. The control computer extracts the actual arc length with the largest value as a reference base. The control computer uses the reference base to subtract the actual arc length of each longitudinal fiber to obtain the arc length to be extended variable. The control computer uses the arc length to be extended variable as the dividend and the actual arc length of each longitudinal fiber as the divisor to perform division operation to solve for the required elongation rate of each longitudinal fiber. The flatness defect of the aluminum single panel surface is quantitatively converted into the required elongation rate.
[0012] The control computer assembles the required elongation of multiple longitudinal fibers into an input vector and multiplies it with the generalized inverse matrix of the static influence coefficient matrix to solve for the optimal approximate solution and generate a servo displacement command vector. The control computer calculates the action feedforward advance using the current conveyor line speed and the inherent response delay time of the servo hydraulic system. When the actual travel distance is equal to the target trigger distance, the control computer extracts the servo displacement command vector from the action delay scheduling queue and synchronously sends it to the servo hydraulic cylinder array to compensate for the hydraulic response lag and align the action execution with the position of the material reaching the deformation zone of the segmented support shoe.
[0013] The control computer registers the pure plate height matrix at the outlet with the ideal height matrix representing the absolutely flat target state and subtracts them point by point to generate a topological residual matrix. The control computer uses a recursive algorithm to calculate the error gradient matrix using the actual elongation residual vector and the actual displacement column vector generated by the topological residual matrix and adds it to the static influence coefficient matrix to generate a dynamic influence coefficient matrix. The control computer extracts the mechanical stiffness limit parameters to generate an upper limit constraint matrix and a lower limit constraint matrix and performs boundary projection on the dynamic influence coefficient matrix to prevent the pure mathematical recursive model from accumulating errors and generating correlation mapping coefficients that exceed the actual mechanical structure's bearing capacity.
[0014] This invention provides an online automatic detection and correction method for flatness defects in aluminum single-panel panels. It has the following beneficial effects: 1. This invention obtains height scanning data by simultaneously scanning the left and right solid-state reference target bars using an entrance span-width laser line scanner. The control computer generates an absolute spatial horizontal zero point based on the height scanning data of the left and right solid-state reference target bars. The control computer extracts the difference between the height scanning data and the absolute spatial horizontal zero point to generate an entrance pure plate surface height matrix, filtering out low-frequency common-mode displacement disturbances in the rack system. A photoelectric trigger switch detects the leading edge of the aluminum single plate to generate a zero-point trigger signal. A speed measuring wheel measures the running displacement of the aluminum single plate. The control computer combines the zero-point trigger signal and the running displacement to generate a material 2D coordinate system. The control computer binds the entrance pure plate surface height matrix to the material 2D coordinate system, eliminating the influence of conveyor belt slippage.
[0015] 2. This invention uses a computer to divide the entrance pure plate height matrix into multiple longitudinal fibers. The computer performs discrete integral calculations on each longitudinal fiber to obtain the actual arc length. The computer extracts the actual arc length with the largest value as a reference base to calculate the required elongation of each longitudinal fiber, quantitatively converting flatness defects into the required elongation. The computer summarizes the required elongation to generate an input vector and multiplies it with the generalized inverse matrix of the static influence coefficient matrix to generate a servo displacement command vector. Combining the conveyor line speed and the inherent response delay time, the computer calculates the action feedforward advance. The servo displacement command vector is extracted from the action delay scheduling queue and sent to the servo hydraulic cylinder array to compensate for hydraulic response lag and align the action execution with the material position.
[0016] 3. This invention controls a computer to register the outlet pure plate height matrix with the target flat ideal height matrix in the same coordinates and subtract them to generate a topological residual matrix. A recursive algorithm is then used to combine the actual elongation residual vector generated from the topological residual matrix with the actual displacement column vector corresponding to the actual displacement data to obtain the error gradient matrix. This gradient matrix is accumulated in the static influence coefficient matrix to generate a dynamic influence coefficient matrix. The computer extracts the mechanical stiffness limit parameters to generate upper and lower constraint matrices. Boundary projection constraints are applied to the dynamic influence coefficient matrix to prevent the cumulative error of the mathematical recursive model from generating mapping coefficients that exceed the mechanical structure's bearing capacity. Attached Figure Description
[0017] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a flowchart of the process of the present invention; Figure 3 This is a flowchart illustrating step S10 of the present invention; Figure 4 This is a flowchart illustrating step S20 of the present invention; Figure 5 This is a flowchart illustrating step S30 of the present invention; Figure 6 This is a flowchart illustrating step S40 of the present invention; Figure 7 This is a flowchart illustrating step S50 of the present invention; Figure 8 This is a comparison diagram of the actual arc length of the longitudinal fibers before and after leveling of the aluminum single panel of the present invention; Figure 9 This is a peak evolution diagram of the straightened exit topology residual matrix of the present invention. Detailed Implementation
[0018] The technical solutions in 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.
[0019] See attached document Figure 1 This invention provides an online automatic detection and correction method for flatness defects on aluminum single-panel panels. The hardware structure is equipped with a segmented multi-roller leveling machine. The segmented multi-roller leveling machine is equipped with flexible working rollers and multiple sets of horizontally parallel segmented support shoes. Each set of segmented support shoes is connected to a servo hydraulic cylinder on its back. The multiple sets of servo hydraulic cylinders form a servo hydraulic cylinder array. The servo hydraulic cylinder array is connected to and controls the segmented support shoes to push the flexible working rollers to produce deformation.
[0020] A horizontally fixed inlet span laser line scanner is installed at the inlet of the segmented multi-roll leveler, and a horizontally fixed outlet span laser line scanner is installed at the outlet. A left and right solid reference target strip are fixedly installed on the follower bearing seat of the inlet conveyor belt, and a left and right solid reference target strip are fixedly installed on the follower bearing seat of the outlet conveyor belt. This ensures that the solid reference target strips on both sides have common-mode mechanical vibration with the conveyor belt support surface. The scanning field of the inlet span laser line scanner covers the left solid reference target strip, the aluminum panel surface, and the right solid reference target strip. The scanning field of the outlet span laser line scanner covers the left outlet solid reference target strip, the aluminum panel surface, and the right outlet solid reference target strip.
[0021] The segmented multi-roller leveling machine is equipped with a photoelectric trigger switch and a measuring speed wheel at the inlet. The photoelectric trigger switch is aligned with the aluminum panel feeding track, and the measuring speed wheel is in contact with the surface of the aluminum panel. The photoelectric trigger switch is connected to the measuring speed wheel and collects the aluminum panel conveying displacement data.
[0022] The control computer is connected to the segmented multi-roller leveler, the inlet span width laser line scanner, the outlet span width laser line scanner, the photoelectric trigger switch, and the actual speed measuring wheel via cables. The control computer receives height scanning data collected by the inlet span width laser line scanner and the outlet span width laser line scanner, and simultaneously receives displacement data collected by the photoelectric trigger switch and the actual speed measuring wheel. The control computer then sends servo displacement commands to the servo hydraulic cylinder.
[0023] See attached document Figure 2 This invention provides an online automatic detection and correction method for flatness defects on aluminum single-panel surfaces, comprising the following steps: S10, the entrance span-width laser line scanner synchronously scans the left solid reference target strip, the surface of the aluminum single panel, and the right solid reference target strip to obtain height scanning data. The control computer generates an absolute space horizontal zero point based on the height scanning data of the left and right solid reference target strips. The control computer extracts the difference between the height scanning data and the absolute space horizontal zero point to generate the entrance pure panel height matrix. The photoelectric trigger switch detects the leading edge of the aluminum single panel to generate a zero-point trigger signal. The actual speed measuring wheel measures the running displacement of the aluminum single panel. The control computer combines the zero-point trigger signal and the running displacement to generate a material 2D coordinate system. The control computer binds the entrance pure panel height matrix to the material 2D coordinate system. S20, the control computer divides the entrance pure plate height matrix into multiple longitudinal fibers along the aluminum single plate conveying direction. The control computer performs curve discrete integral calculation on each longitudinal fiber to obtain the actual arc length. The control computer extracts the actual arc length with the largest value as the reference base. The control computer combines the actual arc length and the reference base to calculate the required elongation rate of each longitudinal fiber. S30, the control computer summarizes all the required elongation rates to generate an input vector, the control computer multiplies the input vector with the generalized inverse matrix of the static influence coefficient matrix to generate a servo displacement command vector, the control computer establishes an action delay scheduling queue in combination with the material 2D coordinate system, when the coordinate position recorded in the material 2D coordinate system reaches directly below the segmented support shoe, the control computer sends a servo displacement command vector to the servo hydraulic cylinder. S40, the exit span-width laser line scanner synchronously scans the solid reference target strip on the left side of the exit, the surface of the aluminum single plate, and the solid reference target strip on the right side of the exit to obtain exit height scanning data. The control computer obtains the exit pure plate surface height matrix based on the exit height scanning data. The control computer retrieves the material 2D coordinate system. The control computer performs spatial registration between the exit pure plate surface height matrix and the target flat state ideal height matrix at the same coordinate position. The control computer compares the exit pure plate surface height matrix and the target flat state ideal height matrix to generate a topological residual matrix. S50, the control computer extracts the topological residual matrix and the actual execution displacement data fed back by the servo hydraulic cylinder. The control computer uses a recursive algorithm to combine the topological residual matrix and the actual execution displacement data to update the static influence coefficient matrix and generate the dynamic influence coefficient matrix. The control computer performs boundary projection constraints on the dynamic influence coefficient matrix. When the internal elements of the dynamic influence coefficient matrix exceed the preset limit, the control computer maintains the static influence coefficient matrix.
[0024] The technical solution will be described in detail below with reference to the accompanying drawings.
[0025] See attached document Figure 3 The S10 step includes the following sub-steps: S101, the entrance span-width laser line scanner simultaneously covers the left solid reference target strip, the aluminum panel surface, and the right solid reference target strip to acquire height scanning data within a single scanning cycle. The entrance span-width laser line scanner emits a single-beam laser beam onto the conveyor belt surface, the length of which covers the sum of the lateral widths of the left solid reference target strip, the aluminum panel surface, and the right solid reference target strip. The integrated photosensitive element within the entrance span-width laser line scanner simultaneously receives the laser signals reflected from the left solid reference target strip, the aluminum panel surface, and the right solid reference target strip within a single exposure time, and converts the laser signals into height scanning data.
[0026] The height scan data records the relative height between the aluminum panel surface and the sensor body, as well as the relative height between the reference target strips on both sides and the sensor body. The frame vibration and the low-frequency vibration of the conveyor belt act synchronously on the relative height. For the photoelectric conversion process of the laser line scanner acquiring the 3D point cloud, those skilled in the art can use conventional machine vision calibration procedures. The photoelectric conversion process is well-known in the field and will not be described further here.
[0027] S102, the control computer generates an absolute spatial horizontal zero point based on the height scan data of the left and right solid-state reference target strips. The left and right solid-state reference target strips are fixedly installed on the rigid support frame or bearing seat of the inlet conveyor belt, causing common-mode mechanical vibration with the conveying surface of the aluminum panel in the vertical direction. The control computer identifies the values in the height scan data located within the preset coordinate intervals on both sides, confirming them as the height data of the left and right solid-state reference target strips. The control computer performs a linear fitting operation on the height data of the left and right solid-state reference target strips to obtain the horizontal spatial height connection line. The control computer defines the horizontal spatial height connection line as the absolute spatial horizontal zero point of the current scan frame. The absolute spatial horizontal zero point is synchronously offset with the mechanical vibration of the conveyor belt, thereby filtering out low-frequency common-mode displacement disturbances that are rigidly coupled to the rack system in subsequent differential calculations.
[0028] S103, the control computer extracts the difference between the height scan data and the absolute spatial horizontal zero point to generate the entrance pure panel surface height matrix. The control computer performs spatial difference calculations, subtracting the height values of the lateral point cloud corresponding to the aluminum panel surface from the height values in the height scan data at the same lateral coordinate of the absolute spatial horizontal zero point, thus eliminating common-mode displacement disturbances caused by the sensor body relative to the frame. The control computer uses the difference values output from the spatial difference calculation to arrange and generate the entrance pure panel surface height matrix. The control computer generates the entrance pure panel surface height matrix using the spatial difference calculation formula: ; In the formula, Represents the pure plate height matrix for export; The height of the pure plate surface at the entrance represents the height scan data. Represents the absolute spatial zero position; Represents the horizontal pixel index; Represents the vertical scan frame index.
[0029] S104, the photoelectric trigger switch detects the leading edge of the aluminum panel and generates a zero-point trigger signal. The photoelectric trigger switch is fixedly installed on the side of the feed track at the inlet of the segmented multi-roller leveler. When the leading edge of the conveyed aluminum panel cuts off the detection beam of the photoelectric trigger switch, the photoelectric trigger switch generates a level jump and sends a zero-point trigger signal to the control computer. The control computer sets the current moment of receiving the zero-point trigger signal as the starting zero point of the material's 2D coordinate system in the longitudinal direction.
[0030] For the selection and signal transmission of photoelectric trigger switches, those skilled in the art can use conventional industrial photoelectric sensors connected to switch input interfaces. The selection and signal transmission are well-known technologies in this field and will not be elaborated here.
[0031] S105, the actual speed measuring wheel measures the running displacement of the aluminum panel and generates a material 2D coordinate system by combining it with the zero-point trigger signal. The actual speed measuring wheel is coaxially fixed on the rotating shaft of the upper and lower pinch roller groups at the entrance of the segmented multi-roller leveler. The forced clamping force of the pinch rollers on the aluminum panel eliminates the local slippage caused by the undulation of the material surface. It drives itself to rotate synchronously without slippage as the feeding action proceeds. The rotary encoder, coaxially connected to the actual speed measuring wheel, continuously outputs high-frequency orthogonal pulses to the control computer as it rotates. The control computer accumulates the number of pulses in real time and calculates the actual running displacement of the aluminum panel relative to the initial zero point by combining it with the circumference parameter of the actual speed measuring wheel. This eliminates the longitudinal positioning deviation caused by slippage or speed fluctuation of the inlet conveyor belt of the segmented multi-roller leveler. The control computer finally defines the horizontal pixel index as the horizontal coordinate and the actual running displacement as the vertical coordinate, dynamically assembling a material 2D coordinate system representing the actual surface spatial distribution of the aluminum panel. For pulse counting and displacement conversion of rotary encoders, those skilled in the art can use the encoder counting module of a conventional motion controller to achieve this. Pulse counting and displacement conversion are well-known technologies in this field and will not be described in detail here.
[0032] It is important to note that, in order to avoid the distortion of the three-dimensional shape caused by the flattening force of the pinch rollers, the entrance span laser line scanner is arranged in the free conveying area in front of the upper and lower pinch roller groups, where there is no forced clamping force. In addition, an independent high-frequency auxiliary encoder is installed at the end of the light-load follow-up support shaft directly below the entrance span laser line scanner. This is used to perform early zero-point positioning calculation compensation before the aluminum panel head reaches the main pinch roller, thereby ensuring that the generated material 2D coordinate system accurately maps the most original three-dimensional free surface undulation of the aluminum panel, and ensuring the geometric authenticity of the subsequent actual arc length calculation.
[0033] The moment the leading edge of the aluminum panel is conveyed to the main clamping roller and effective mechanical engagement occurs, the control computer immediately locks the current cumulative longitudinal coordinate value fed back by the high-frequency auxiliary encoder. This cumulative longitudinal coordinate value is used as the new dead reckoning relay origin of the material's 2D coordinate system, and the seamless relay of the displacement signal source is executed synchronously. The control computer maintains a virtual master axis coordinate in the background, and the data from the high-frequency auxiliary encoder and the rotary encoder are respectively used as two independent slave axis mappings. At the moment of engagement, the control computer simultaneously reads the current feedback difference between the two encoders and stores this difference as a fixed offset constant in the internal register. The subsequent longitudinal displacement calculation of the material's 2D coordinate system uses the real-time reading of the rotary encoder + the fixed offset constant for continuous dead reckoning, thereby eliminating the pulse counting step jump or disconnection caused by the instant of physical channel switching.
[0034] S106, the control computer binds the entrance pure plate surface height matrix to the material 2D coordinate system. The control computer retrieves the entrance pure plate surface height matrix, which includes height values corresponding to multiple horizontal pixel indices and vertical scan frame indices. During each scanning cycle of the entrance span-width laser line scanner, the control computer synchronously latches the actual running displacement calculated by the rotary encoder and directly replaces the vertical scan frame index of the current scan frame with the corresponding actual running displacement. This ensures that each height value in the entrance pure plate surface height matrix is accurately mapped to a specific coordinate point in the material 2D coordinate system, completing the mounting of height data on the actual spatial position of the aluminum single plate surface. This enables subsequent elongation calculation and segmented leveling to be dynamically tracked based on a unified material reference.
[0035] See attached document Figure 4 The S20 step includes the following sub-steps: S201, the control computer reads the structural parameters of the segmented support shoes assembled inside the segmented multi-roller leveler. The segmented support shoes have a fixed number of arrangement and a fixed single width in the transverse direction. Based on the fixed number of arrangement of the segmented support shoes, the control computer determines the total number of cuts in the transverse width of the material's 2D coordinate system, so that multiple adjacent longitudinal fibers are summarized and mapped to the transverse action range of the same segmented support shoe, thereby forming a mapping relationship in spatial position.
[0036] S202, the control computer performs lateral discretization interval division on the entrance pure plate height matrix, the control computer retrieves the entrance pure plate height matrix stored in the material 2D coordinate system, and the control computer delineates multiple parallel longitudinal strip regions in the lateral pixel index of the entrance pure plate height matrix according to the lateral center coordinates corresponding to the segmented support shoe. Each longitudinal strip region corresponds to a closed longitudinal material line on the surface of the aluminum single plate.
[0037] S203, the control computer extracts all scanning height values within each longitudinal strip area and generates multiple independent longitudinal fibers. The control computer extracts column vectors of the height values corresponding to the longitudinal scanning frame index under a specific transverse coordinate, and combines them to form discrete height curve data with continuous longitudinal coordinates. Each of the multiple longitudinal fibers independently represents the real fluctuation shape generated by the conveying movement at different transverse positions of the aluminum panel, thereby decoupling the spatial 3D deformation into multiple parallel 2D curves. For the matrix column slicing and vector extraction operations in the computer, those skilled in the art can use array traversal algorithms of conventional programming languages to implement them. The matrix column slicing and vector extraction operations are well-known technologies in this field and will not be described in detail here.
[0038] S204, the control computer performs discrete integral calculations on the curves of each longitudinal fiber to obtain the actual arc length. First, the control computer performs linear detrending processing on the extracted height curve data of a single longitudinal fiber, calculating and subtracting the overall linear slope caused by the overall tilt of the aluminum panel or the feeding pitch, to extract the pure surface high-frequency deformation waveform. Then, the control computer extracts the height difference between adjacent coordinate data points within a single longitudinal fiber, and simultaneously extracts the longitudinal displacement interval between adjacent coordinate data points. The control computer uses the Pythagorean theorem to calculate the spatial straight-line distance between adjacent coordinate data points, and iterates through and accumulates the spatial straight-line distances between all adjacent coordinate data points within a single longitudinal fiber to obtain the curve integral value. The control computer defines the curve integral value as the actual arc length corresponding to a specific longitudinal fiber.
[0039] For the cumulative calculation of the curve distance of discrete data points, those skilled in the art can use the cumulative summation algorithm in conventional numerical analysis. The cumulative summation algorithm is a well-known technology in this field and will not be described in detail here.
[0040] S205, the control computer extracts the actual arc length with the largest value as the reference base. The control computer collects all actual arc lengths calculated from the discrete integral of all longitudinal fibers. It then uses a numerical comparison algorithm to iterate through all actual arc lengths and selects the one with the largest value. Because aluminum panels exhibit uneven elongation in different areas during warping, the actual arc length with the largest value represents the material line with the most severe transverse deformation and the greatest fiber elongation (i.e., the most severe wave or bulge). The control computer sets the actual arc length with the largest value as the reference base and temporarily stores it in storage for later calculations.
[0041] S206, the control computer retrieves the actual arc length of each longitudinal fiber stored in the memory and the actual arc length with the largest value of the reference base, and pairs the actual arc length of each longitudinal fiber with the reference base through the channel index to establish the data basis for differential calculation.
[0042] S207, the control computer uses the reference base to subtract the actual arc length of each longitudinal fiber to perform a subtraction operation, obtains the arc length of each longitudinal fiber relative to the reference base to be extended variable, and clarifies the target extension amount that each shorter longitudinal fiber needs to achieve inside the segmented multi-roller leveling machine, so as to achieve the leveling mechanical goal of making the short fiber flush with the long fiber.
[0043] S208, the control computer uses the arc length to be extended variable as the dividend and the actual arc length of each longitudinal fiber as the divisor to perform a division operation, solving for the required elongation rate of each longitudinal fiber. This quantitatively transforms the flatness defects of the aluminum panel surface into the elongation rate requirement under pressure from the segmented multi-roller leveling machine. The control computer then uses the elongation rate calculation formula to solve for the required elongation rate of each longitudinal fiber. ; In the formula, This represents the required elongation of the i-th longitudinal fiber; Representing the The actual arc length of the longitudinal fiber; The actual arc length representing the largest value (i.e., the reference base). The transverse channel index represents a single longitudinal fiber.
[0044] See attached document Figure 5 The S30 step includes the following sub-steps: In the S301, during the initial operation or offline calibration phase of the equipment, the control system individually drives each group of servo hydraulic cylinders to perform multi-gradient step tests exceeding the material's elastic dead zone, eliminating invalid elastic strokes that do not produce plastic deformation. Simultaneously, an outlet-wide laser line scanner dynamically measures the actual plastic elongation change response of the standard aluminum panel in each transverse segmentation region, thereby constructing and pre-storing a static influence coefficient matrix based on the linear effective working area. From a physical mapping perspective, the static influence coefficient matrix is the Jacobian matrix obtained by performing a locally linearized Taylor expansion of the material's constitutive nonlinear relationship near the elastic-plastic critical yield point of a specific plate thickness and material in a segmented multi-roller leveler. Subsequently, the control computer retrieves the pre-stored static influence coefficient matrix. This matrix establishes a mapping relationship between the actuation displacement of each group of servo hydraulic cylinders corresponding to each segment support shoe in the segmented multi-roller leveler and the required elongation of multiple longitudinal fibers. The number of transverse row elements within the static influence coefficient matrix equals the total number of segmented longitudinal fibers, and the number of longitudinal column elements equals the total number of arrangement elements of the segmented support shoes.
[0045] Since the total number of longitudinal fibers cut out is much greater than the total number of segmented support shoes, the static influence coefficient matrix is a slender non-square matrix. Each matrix element in the static influence coefficient matrix clearly represents the change in elongation of the longitudinal fibers at the corresponding positions caused by a single set of servo hydraulic cylinders generating a unit displacement, thus reflecting the mechanical characteristics of the segmented multi-roller leveler with multiple points of force coupled together in the transverse structure.
[0046] S302, the control computer arranges the required elongation of multiple longitudinal fibers in the order of the transverse channel index, and combines them to generate an input vector in the form of a column vector. The numerical values of the elements inside the input vector correspond one-to-one with the deformation degree of different transverse areas of the aluminum panel. The input vector serves as the input control target of the multi-input multi-output control system.
[0047] S303, the control computer performs a generalized inverse operation on the static influence coefficient matrix to obtain the generalized inverse matrix of the static influence coefficient matrix. The control computer then performs matrix multiplication between the generalized inverse matrix of the static influence coefficient matrix and the input vector, thereby solving the optimal approximate solution of the overdetermined control equation in the least squares sense. This generates a servo displacement command vector in column vector form, which internally contains the hydraulic stroke target values corresponding to each of the multiple segmented support shoes. The control computer generates the servo displacement command vector through a reverse mapping calculation formula. ; In the formula, Represents the servo displacement command vector; The generalized inverse matrix representing the static influence coefficient matrix; This represents the input vector.
[0048] Numerical calculations of generalized matrix inverse operations and matrix multiplication can be implemented using conventional linear algebra software algorithm libraries by those skilled in the art. Numerical calculations of generalized matrix inverse operations and matrix multiplication are well-known technologies in this field and will not be elaborated upon here.
[0049] S304: The control computer establishes a first-in-first-out (FIFO) motion delay scheduling queue by combining the material's 2D coordinate system with the mechanical positions of the segmented support shoes. The control computer extracts the servo displacement command vector sequence corresponding to the continuous longitudinally segmented micro-elements within the entire plate area. Considering the mechanical frequency response limitations of the hydraulic system, the control computer performs moving smoothing filtering and sliding window envelope extremum extraction on the servo displacement command values of several adjacent micro-elements along the longitudinal channel, reasonably reducing the dimensionality of the high-frequency theoretical calculation data to generate a dynamic timing control command package that matches the dynamic response frequency of the hydraulic system. The control computer stores the dynamic timing control command package, consisting of multiple sets of continuous nodes, into the motion delay scheduling queue, awaiting triggering.
[0050] S305, the control computer retrieves the assembly parameters of the segmented multi-roller leveling machine, specifying the fixed installation distance between the beam projection surface of the entrance span-width laser line scanner and the center line of the first set of force-bearing segment support shoes corresponding to the servo hydraulic cylinder array. To meet the timing requirements for performing a full-panel global scan and global optimal attitude calculation on a single aluminum panel, the fixed installation distance is configured to be greater than the maximum longitudinal length of the aluminum panel under test, ensuring that the physical leading edge of the aluminum panel only reaches the first set of force-bearing segment support shoes after the entire aluminum panel has completely passed through the scanning field of view of the entrance span-width laser line scanner.
[0051] The control computer extracts the real-time total longitudinal displacement fed back by the measured speed-measuring wheel. It also extracts the longitudinal coordinate values recorded in the execution control data packet at the top of the action delay scheduling queue. The control computer subtracts the longitudinal coordinate values from the real-time total longitudinal displacement to calculate the actual travel distance of the target aluminum panel cross-section since the completion of height data acquisition. Finally, the control computer retrieves the preset inherent response delay time of the servo hydraulic system and, combined with the current conveyor linear speed calculated from the measured speed-measuring wheel, calculates the action feedforward using a formula. ; In the formula, Represents the advance of the action feed-in; Represents the current conveyor line speed; This represents the inherent response delay time of the servo hydraulic system. The control computer further utilizes the fixed installation distance minus the motion feed-in advance. Generate the target trigger distance and control the computer to compare the actual travel distance with the target trigger distance in real time.
[0052] S306, when the actual travel distance equals the target trigger distance, the control computer determines to trigger the advance action mechanism to compensate for hydraulic response lag, ensuring that when the servo hydraulic cylinder actually presses down to its position, the coordinate position recorded in the material's 2D coordinate system accurately reaches the deformation action zone directly below the first set of force-bearing segment support shoes inside the segmented multi-roll leveler (i.e., the starting point of the elasto-plastic continuous deformation zone formed by the interlocking of multiple work rolls). The control computer triggers the synchronization control mechanism, retrieving the execution control data packet corresponding to the arrival position from the action delay scheduling queue. The control computer parses the execution control data packet to obtain the servo displacement command vector, and synchronously sends the servo displacement command vector to the servo hydraulic cylinder array.
[0053] During the passage of the entire aluminum panel through the segmented multi-roller leveling machine, multiple sets of servo hydraulic cylinders continuously generate dynamic timing control command packets according to the action delay scheduling queue. This allows for dynamic compensation of deformation differences in different longitudinal regions of the same aluminum panel while meeting the physical stiffness limit of the mechanical continuous envelope. The multiple sets of servo hydraulic cylinders drive corresponding segmented support shoes to push the flexible working rollers according to the servo displacement command vector, generating local deformation correction in the coordinate-aligned material areas.
[0054] For signal processing of digital command to analog quantity driving hydraulic actuators, those skilled in the art can use conventional digital-to-analog converter boards. Signal processing of digital command to analog quantity driving hydraulic actuators is a well-known technology in this field and will not be described in detail here.
[0055] See attached document Figure 6 The S40 step includes the following sub-steps: S401, the exit span-width laser line scanner simultaneously covers the solid reference target strip on the left side of the exit, the surface of the aluminum single panel, and the solid reference target strip on the right side of the exit to acquire exit height scanning data within a single scanning cycle. The exit span-width laser line scanner projects a transverse laser beam onto the conveyor belt surface, the length of which covers the sum of the transverse widths of the solid reference target strip on the left side of the exit, the surface of the aluminum single panel, and the solid reference target strip on the right side of the exit. The photosensitive element integrated inside the exit span-width laser line scanner simultaneously receives the optical signals reflected from the solid reference target strip on the left side of the exit, the surface of the aluminum single panel, and the solid reference target strip on the right side of the exit, and converts the optical signals into exit height scanning data. The exit height scanning data records the three-dimensional topography information of the aluminum single panel surface after deformation correction by the segmented multi-roller leveling machine, as well as the relative height information of the two reference target strips. For the photoelectric conversion process of the laser line scanner capturing reflected light and generating height data, those skilled in the art can use the triangulation calculation program inside a conventional three-dimensional vision sensor to achieve this. The photoelectric conversion process is a well-known technology in the field and will not be described in detail here.
[0056] S402, the control computer generates the absolute spatial horizontal zero point of the exit based on the scan height values of the solid-state reference target strips on the left and right sides of the exit. The solid-state reference target strips on the left and right sides of the exit are fixedly installed on the rigid support frame or bearing seat of the exit conveyor belt, causing them to generate common-mode mechanical runout in the vertical direction with the conveyor surface of the leveled aluminum panel. The control computer extracts the height values of the corresponding two edge positions from the exit height scan data, using them as the height data of the solid-state reference target strips on the left and right sides of the exit, respectively. The control computer performs a linear fitting operation on the height data of the solid-state reference target strips on the left and right sides of the exit to obtain the horizontal spatial height connection line. The control computer defines the horizontal spatial height connection line as the absolute spatial horizontal zero point of the exit scan frame, thereby filtering out the low-frequency common-mode vibration offset coupled to the rigid base at the exit end of the leveling machine in subsequent spatial differential calculations.
[0057] S403, the control computer obtains the pure surface height matrix of the aluminum panel at the exit based on the exit height scan data. The control computer performs spatial difference calculations to extract the lateral point cloud height values corresponding to the aluminum panel surface from the exit height scan data. These lateral point cloud height values are then subtracted sequentially from the height values corresponding to the zero horizontal coordinates at the exit, filtering out the jump displacement disturbances caused by the segmented multi-roller leveling machine's exit conveyor belt. The control computer uses the difference values obtained from the spatial difference calculations to generate the pure surface height matrix of the exit, completing the same-frame vibration removal processing of the exit end data. The control computer generates the pure surface height matrix of the exit using the exit spatial difference calculation formula: ; In the formula, Represents the pure plate height matrix for export; Represents the exit height scan data; This represents the absolute level of exports at zero. Represents the horizontal pixel index; Represents the vertical scan frame index.
[0058] S404, the control computer retrieves the material's 2D coordinate system and extracts the assembly parameters of the segmented multi-roller leveler, clarifying the fixed mechanical span between the inlet and outlet span-width laser line scanners in the longitudinal conveying direction. The control computer simultaneously extracts the real-time measured displacement of the actual speed measuring wheel. The control computer extracts the real-time physical running linear velocity of the segmented multi-roller leveler's outlet-side conveying mechanism; simultaneously, the outlet span-width laser line scanner generates an outlet-side zero-point trigger signal by capturing the contour jump characteristics of the aluminum single-panel's physical leading edge. The control computer forcibly sets the moment corresponding to the outlet-side zero-point trigger signal as the absolute zero-starting point of the outlet longitudinal mapping coordinates. From this zero-starting point, the control computer performs time integration on the real-time physical running linear velocity, directly calculating the longitudinal mapping coordinates of the corresponding outlet scanning frame within the material's 2D coordinate system. The control computer calculates the longitudinal mapping coordinates using the outlet integral mapping formula. ; In the formula, Represents the vertical mapping coordinates; This represents the zero-point trigger signal on the export side, indicating the zero-point start time. Represents the sampling time of the current exit scan frame; The representative of the export-side conveyor mechanism is Real-time physical linear velocity at any given moment.
[0059] When the exit-width laser line scanner captures the physical trailing edge contour jump characteristics of the aluminum panel, the control computer extracts the integrally calculated total longitudinal length of the entire panel at the exit. The control computer retrieves the required elongation rate of each longitudinal fiber calculated in step S20, and calculates the theoretical total plastic elongation of the aluminum panel under leveling action through weighted averaging. The control computer subtracts the sum of the original total length at the entrance and the theoretical total plastic elongation from the total longitudinal length of the entire panel at the exit, separating the pure displacement residual caused solely by factors such as belt slippage. Subsequently, the control computer performs closed-loop deviation allocation and compensation on each longitudinal mapping coordinate extracted from the entire panel based solely on this pure displacement residual, thereby eliminating the cumulative positioning error caused by conveyor belt slippage while preserving the true physical elongation deformation of the material.
[0060] For the synchronous acquisition and processing of displacement sensor signals, those skilled in the art can use conventional multi-channel data acquisition cards. The synchronous acquisition and processing of displacement sensor signals is a well-known technology in this field and will not be elaborated here.
[0061] S405, the control computer assigns the longitudinal mapping coordinates and corresponding lateral pixel indices to all height values in the exit pure plate surface height matrix, and establishes a position association label between the exit height data and the material's 2D coordinate system. The control computer uses this position association label to mount the exit pure plate surface height matrix into the material's 2D coordinate system framework, completing the conversion of the exit height data to the global material reference coordinate domain, ensuring that the morphology data before and after leveling correction are on the same reference base.
[0062] S406, the control computer establishes the parameter setting standard for the ideal height matrix of the target flat state. The control computer extracts the data grid arrangement dimension stored within the material's 2D coordinate system. Based on this data grid arrangement dimension, the control computer creates a 2D basic array in memory with the same number of rows and columns as the output pure panel height matrix. The control computer retrieves the system's preset absolute flatness reference surface parameters and globally writes the constant values of these parameters into all data nodes of the 2D basic array, generating the ideal height matrix of the target flat state, representing the aluminum panel in a state where internal stress warping is completely eliminated and the surface is absolutely flat. This ideal height matrix of the target flat state constitutes a quantitative benchmark for evaluating the degree of residual deformation after flattening.
[0063] S407, the control computer spatially registers the exit pure plate surface height matrix with the target flat state ideal height matrix at the same coordinate position. The control computer traverses within the material's 2D coordinate system, searching for data nodes with the same horizontal pixel index and the same vertical mapping coordinates. The control computer extracts elements from the exit pure plate surface height matrix and the target flat state ideal height matrix with the same data node coordinates, merging them to form comparative micro-element pairs, thus completing the data dimensionality reduction from a global matrix view to local point-to-point analysis. The control computer completes the alignment and extraction of data nodes using a spatial registration mapping formula: ; In the formula, A pair of spatially registered micro-elements representing the same coordinate position; This represents the height value of the pure plate surface height matrix at the corresponding coordinates. The ideal reference value of the target's ideal height matrix in a flat state at the corresponding coordinates; Represents the horizontal pixel index; Represents the vertical mapping coordinates.
[0064] S408, the control computer compares the height matrix of the pure plate surface at the exit with the ideal height matrix of the target flat state and generates a topological residual matrix. The control computer then compares the differential elements after spatial registration. The height values of each coordinate node within the pure surface height matrix of the exit are extracted and compared with the ideal reference values of the same coordinate nodes within the ideal height matrix of the target flat state. The control computer performs point-by-point subtraction on the height values with the same horizontal pixel index and vertical mapping coordinates and the ideal reference values to obtain the vertical deviation of the actual surface morphology of each micro-element region from the ideal flat state. For the low-level traversal operation of subtracting corresponding elements in the matrix, those skilled in the art can use parallel array calculation functions in conventional computer instruction sets. The low-level traversal operation of subtracting corresponding elements in the matrix is a well-known technique in the art and will not be described in detail here.
[0065] S409, the control computer recombines the vertical deviations calculated from all nodes according to the original material 2D coordinate system arrangement, generating a topological residual matrix characterizing the remaining deformation of the aluminum panel after physical leveling. The sign and magnitude of the elements in the topological residual matrix directly reflect the under- or over-adjustment of the leveling operation at a specific coordinate position by the segmented multi-roller leveler. The control computer calculates and generates the topological residual matrix using a difference comparison formula: ; In the formula, Represents the topological residual matrix; This represents the height value of the pure plate surface height matrix at the corresponding coordinates. The ideal reference value of the target's ideal height matrix in a flat state at the corresponding coordinates; Represents the horizontal pixel index; Represents the vertical mapping coordinates.
[0066] See attached document Figure 7 The S50 step includes the following sub-steps: S501 specifies the monitoring conditions for the control computer to extract the topological residual matrix and the actual displacement data fed back by the servo hydraulic cylinders under specific operating conditions. The control computer monitors the conveyor belt spindle speed and hydraulic pump station pressure of the segmented multi-roller leveler in real time. When the conveyor belt spindle speed is within a preset steady-state speed range and the hydraulic pump station pressure is maintained within a preset rated pressure range, the control computer determines that the current equipment has reached a specific operating condition. Furthermore, the control computer further determines whether all servo hydraulic cylinders remain within the static command dead zone without sudden changes during the processing of the aluminum panel. Only when the steady-state pressure holding condition of the panel is met is the extraction of the current panel's topological residual matrix allowed for model updating.
[0067] Under specific operating conditions, the control computer retrieves the topological residual matrix corresponding to the same aluminum single-panel generated in memory. Simultaneously, the control computer extracts the actual displacement data returned in real-time from the displacement sensor integrated within the servo hydraulic cylinder. The topological residual matrix and the actual displacement data are timestamped using the control computer's internal clock, establishing a causal alignment benchmark between the deformation assessment result and the mechanical motion execution. For high-frequency data acquisition and filtering / interference suppression processing from the displacement sensor, those skilled in the art can implement it using a high-speed analog input board of a conventional programmable logic controller. High-frequency data acquisition and filtering / interference suppression processing are well-known technologies in this field and will not be elaborated upon here.
[0068] In S502, the control computer uses a recursive algorithm to update the static influence coefficient matrix and generate a dynamic influence coefficient matrix by combining the topological residual matrix and the actual displacement data. The control computer performs discrete integral calculations on the coordinate nodes corresponding to each longitudinal fiber in the topological residual matrix to obtain the actual exit arc length after leveling correction. The control computer calculates the difference between the actual exit arc length and the actual arc length with the largest value used as the reference base. This difference is divided by the reference base value to generate an actual elongation residual vector that conforms to the physical definition of engineering tensile deformation. This actual elongation residual vector accurately represents the degree of under- or over-adjustment in the local leveling area. The control computer extracts and combines the actual displacement data according to the arrangement channels of the segmented support shoes to form an actual displacement column vector.
[0069] The control computer uses a recursive algorithm to calculate the outer product of the actual elongation residual vector and the actual displacement column vector, deriving the error gradient matrix characterizing the current system state deviation. The control computer then adds the error gradient matrix, along with a set iteration step size, to all corresponding elements of the static influence coefficient matrix, completing the adaptive correction of the mechanical mapping correlation model of the segmented multi-roll leveler. The dynamic influence coefficient matrix can track and absorb nonlinear time-varying disturbances caused by wear of mechanical components and batch-to-batch yield strength fluctuations in aluminum panels in real time, and replaces the static influence coefficient matrix in the inverse mapping calculation in the next control loop.
[0070] S503, the control computer calculates and generates a dynamic influence coefficient matrix using a matrix adaptive recursive formula: ; In the formula, Representing the The dynamic influence coefficient matrix for each iteration cycle; The dynamic influence coefficient matrix represents the dynamic influence coefficient matrix of the previous iteration cycle, and the dynamic influence coefficient matrix of the initial operating cycle of the control system is forcibly assigned the static influence coefficient matrix. This represents the recursive learning rate coefficient set internally by the control computer. This represents the actual elongation residual vector generated by the transformation of the topological residual matrix; This represents a row vector (i.e., the transpose of the actual displacement column vectors) consisting of the actual displacement data. An integer index representing the discrete control time step.
[0071] S504, the control computer retrieves the mechanical stiffness limit parameters recorded in the assembly configuration table of the segmented multi-roller leveler. These parameters include the maximum local bending limit of the flexible work roll and the maximum allowable displacement offset of the servo hydraulic cylinders corresponding to each segmented support shoe. The control computer maps these mechanical stiffness limit parameters to the system's matrix parameter space, generating an upper constraint matrix and a lower constraint matrix with dimensions identical to the dynamic influence coefficient matrix, based on the total number of segmented support shoe arrangements and the total number of longitudinal fiber cuts. These upper and lower constraint matrices establish the physically valid boundaries for parameter recursion updates, preventing the purely mathematical recursive model from generating correlation mapping coefficients that exceed the actual mechanical structure's bearing capacity due to accumulated errors.
[0072] S505: The control computer performs a boundary projection operation on the dynamic influence coefficient matrix obtained in the current iteration cycle. The control computer extracts matrix elements at each coordinate position within the dynamic influence coefficient matrix one by one. Simultaneously, the control computer extracts the numerical boundaries of corresponding coordinate positions within the upper and lower constraint matrices. The control computer determines whether a matrix element has exceeded the boundary. If a matrix element is greater than the value of the upper constraint matrix, the control computer forcibly modifies the matrix element to the value of the upper constraint matrix; if a matrix element is less than the value of the lower constraint matrix, the control computer forcibly modifies the matrix element to the value of the lower constraint matrix; when a matrix element is between the upper and lower constraint matrices, the control computer retains the original matrix element unchanged.
[0073] For the underlying logic processing of matrix element extreme value judgment and forced assignment, those skilled in the art can use conditional branch statements in conventional programming logic. Conditional branch statements are well-known technology in this field and will not be elaborated here.
[0074] S506: After the control computer completes the traversal, comparison, and replacement of all matrix elements, it assembles and generates a constrained dynamic influence coefficient matrix. This constrained dynamic influence coefficient matrix is directly used to replace the static influence coefficient matrix for the next control cycle in the inverse mapping calculation of multi-input multi-output control, ensuring that subsequent servo hydraulic cylinder displacement commands can be safely executed by the physical mechanism. The control computer generates the constrained dynamic influence coefficient matrix using the boundary projection constraint formula: ; In the formula, The first one in the constrained dynamic influence coefficient matrix Line number Column elements; The first element in the dynamically generated influence coefficient matrix, generated by recursive updates, represents the... Line number Column elements; Represents the lower bound constraint matrix of the th Line number The lower limit value of the column; Represents the upper limit constraint matrix of the first Line number The upper limit value of the column; The horizontal row index represents a single longitudinal fiber, corresponding to the row index of the matrix; This represents the vertical column index corresponding to the segmented support boot.
[0075] Specific application examples: Initial operating conditions and equipment parameter settings: The material to be tested is a 3003 series aluminum single panel with dimensions of 2000mm (length) × 1500mm (width) × 3mm (thickness). Due to uneven release of rolling stress, the aluminum single panel exhibits a typical mid-wave defect (i.e., the middle of the panel is wavy due to excessive extension, while the two side edges are relatively straight).
[0076] The segmented multi-roll leveling machine is internally equipped with 7 sets of transversely arranged segmented support shoes (i.e., the number of longitudinal column elements is 7). The system's servo hydraulic system has an inherent response delay time. Second.
[0077] Data acquisition and elongation calculation: Aluminum single-panel The speed passes through the entrance. The entrance span-width laser line scanner acquires height scanning data, and the control computer, in conjunction with the photoelectric trigger switch and the actual speed measuring wheel, generates a material 2D coordinate system, and binds the entrance pure plate surface height matrix to it.
[0078] The control computer divides the slab surface laterally into 7 longitudinal fibers. By performing discrete integral calculations on the height matrix of the pure slab surface at the entrance, the actual arc lengths of the 7 longitudinal fibers are obtained. to .
[0079] The actual arc length of the fibers on both sides was calculated. The middle fiber, due to its undulating shape, has the longest actual arc length, and this largest actual arc length serves as the reference base. .
[0080] Substitute into the elongation calculation formula Taking the first longitudinal fiber at the outermost edge as an example: (That is, a required extension rate of 0.25% is needed to catch up with the middle wave pattern).
[0081] This generates an input vector containing 7 elements. .
[0082] Feedforward control and command issuance: The computer controls the input vector The generalized inverse matrix of the static influence coefficient matrix obtained through offline calibration Multiplication, calculated using the inverse mapping formula Generate servo displacement command vector .
[0083] Simultaneously, control the computer to calculate the action feed-forward amount: .
[0084] The target trigger distance is calculated by subtracting 50mm from the fixed installation distance of the first set of force-bearing segment support shoes. When the actual travel distance equals the target trigger distance, the action is triggered in advance to overcome hydraulic lag, and the servo hydraulic cylinder array precisely pushes the elastic flexible work roller to produce deformation.
[0085] Loop closure detection and adaptive update: The leveled aluminum panel passes through the exit, and the exit width laser line scanner acquires the exit height scanning data. The control computer spatially registers the pure panel height matrix at the exit with the ideal height matrix of the target flat state (the absolute height difference between the reference plane and the target plane is 0).
[0086] Substitute into the difference comparison formula The topological residual matrix was generated. In the initial batch, a maximum protrusion of approximately +0.8 mm was found in the topological residual matrix (indicating wear in the hydraulic system or that the material yield strength was slightly higher than the calibration value, resulting in under-adjustment).
[0087] The control computer extracts the topological residual matrix and the actual displacement data fed back from the servo hydraulic cylinder, and then uses the matrix adaptive recursive formula: (Set the learning rate coefficient) =0.1) Generate a dynamic influence coefficient matrix.
[0088] The computer then executes the formula using the upper and lower constraint matrices. Boundary projection constraints are applied to prevent the correction range from exceeding the limit of the flexible work roll, and a constrained dynamic influence coefficient matrix is generated for the control of the next sheet.
[0089] Experimental verification and effect comparison: To verify the ability of the dynamic influence coefficient matrix recursive algorithm in this invention to correct the mechanical mapping deviation of the segmented multi-roller leveling machine, as well as the improvement of the actual morphology of the aluminum single panel before and after leveling, an online automatic detection and correction experiment was conducted on 50 consecutive aluminum single panels of the same specification.
[0090] The conclusions are as follows: Reference Appendix Figure 8 The figure shows the actual arc length distribution of the seven longitudinal fibers cut from a single aluminum sheet under steady-state conditions before entering the equipment (original inlet data) and after correction (actual outlet data). It can be seen that the actual arc length of the central longitudinal fiber (channel 4) at the inlet is greater than that of the fibers on both sides (channels 1 and 7), indicating a mid-wave defect. However, after precise pressing correction by the servo hydraulic cylinder array for the required elongation, the actual arc lengths of each longitudinal fiber at the outlet become basically consistent and close to the value of the reference base, proving the effectiveness of using the pure panel height matrix to calculate the elongation and issue servo displacement command vectors.
[0091] Reference Appendix Figure 9 This figure records the maximum absolute value of the vertical deviation (i.e., the maximum residual deformation) within the topological residual matrix of each of 50 aluminum panels processed in the same batch. It can be seen that in the initial few iterations (the first 5 panels), due to the use of a static influence coefficient matrix, there is a slight under-adjustment in the leveling, and the peak value of the topological residual matrix remains around 0.8 mm. As the system triggers specific operating conditions, the control computer uses a recursive algorithm to continuously update and generate a dynamic influence coefficient matrix by combining the topological residual matrix with actual displacement data. The error is absorbed and converged. After approximately the 15th panel, the peak value of the topological residual matrix steadily decreases and remains below 0.15 mm. This fully demonstrates that the adaptive recursive model of this invention can effectively overcome mechanical wear and material nonlinear fluctuations, ensuring long-term flatness correction accuracy.
Claims
1. An online automatic detection and correction method for flatness defects in aluminum single-panel panels, applied to a system including a segmented multi-roller leveling machine, an entrance span laser line scanner, and a control computer, wherein the segmented multi-roller leveling machine is internally equipped with multiple segmented support shoes controlled by a servo hydraulic cylinder array, characterized in that... Includes the following steps: The entrance span-width laser line scanner acquires the height scanning data of the aluminum single panel, and the control computer generates the entrance pure panel height matrix based on the height scanning data. The control computer divides the entrance pure plate height matrix longitudinally into multiple longitudinal fibers corresponding to each of the segmented support boots, and calculates the required elongation rate of each longitudinal fiber; The control computer extracts the required elongation of each of the longitudinal fibers to form an input vector, and performs mapping calculations in combination with the generalized inverse matrix of the preset static influence coefficient matrix to generate a servo displacement command vector corresponding to the servo hydraulic cylinder array. The control computer acquires the running displacement of the aluminum panel, determines the coordinate position of the aluminum panel based on the running displacement, and when the aluminum panel area corresponding to the required elongation reaches the segmented support shoe, it sends the servo displacement command vector to the servo hydraulic cylinder array to drive the segmented support shoe to move.
2. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 1, characterized in that, The system also includes a left solid reference target bar and a right solid reference target bar installed on the inlet side of the segmented multi-roll leveler; The process of acquiring height scanning data of the aluminum single panel and generating an entrance pure panel height matrix based on the height scanning data by the control computer includes: The entrance span-width laser line scanner simultaneously scans the left solid reference target bar, the surface of the aluminum single plate, and the right solid reference target bar to obtain height scanning data; The control computer fits the horizontal spatial height line based on the height scanning data of the left solid reference target bar and the right solid reference target bar, and defines it as the absolute spatial horizontal zero point; Spatial difference operation is performed to extract the difference between the surface height of the aluminum single panel and the absolute spatial horizontal zero point, and the entrance pure panel height matrix is generated by combining the horizontal pixel index.
3. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 2, characterized in that, The system also includes a photoelectric trigger switch and a speed measuring wheel; The process of obtaining the running displacement of the aluminum panel includes: The photoelectric trigger switch detects the leading edge of the aluminum panel and generates a zero-point trigger signal, and the actual speed measuring wheel measures the actual running displacement of the aluminum panel as the running displacement. The control computer defines the horizontal pixel index as the horizontal coordinate and the actual running displacement as the vertical coordinate, and dynamically assembles a material 2D coordinate system representing the actual surface spatial distribution of the aluminum single panel by combining the zero-point trigger signal.
4. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 1, characterized in that, The step of dividing the entrance pure plate height matrix longitudinally into multiple longitudinal fibers corresponding to each of the segmented support boots, and calculating the required elongation of each longitudinal fiber, includes: The control computer defines multiple longitudinal strip regions in the entrance pure plate height matrix according to the transverse center coordinates corresponding to the segmented support boot, and extracts the height values of each longitudinal strip region to generate multiple independent longitudinal fibers. The control computer performs discrete integral calculations on each longitudinal fiber to obtain the actual arc length.
5. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 4, characterized in that, The control computer extracts the actual arc length with the largest value as the reference base. The control computer uses the reference base to subtract the actual arc length of each longitudinal fiber to obtain the arc length to be extended variable, and performs a division operation with the arc length to be extended variable as the dividend and the actual arc length of each longitudinal fiber as the divisor to solve for the required elongation rate of each longitudinal fiber.
6. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 1, characterized in that, The step of generating a servo displacement command vector corresponding to the servo hydraulic cylinder array by mapping calculation based on the required elongation rate and a preset influence coefficient matrix includes: The control computer combines the desired elongation rates of the multiple longitudinal fibers to generate an input vector in the form of a column vector; Perform a generalized inverse matrix operation on the static influence coefficient matrix to obtain the generalized inverse matrix; The input vector is multiplied by the generalized inverse matrix to generate the servo displacement command vector that covers multiple hydraulic stroke target values.
7. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 3, characterized in that, The step of sending the servo displacement command vector to the servo hydraulic cylinder array to drive the segmented support shoe to move includes: An action delay scheduling queue is established, which stores an execution control data packet containing the servo displacement command vector and the corresponding longitudinal coordinate value. The action feedforward is calculated by combining the current conveyor line speed obtained from the measured speed measuring wheel with the preset inherent response delay time of the servo hydraulic system. The target trigger distance is generated by subtracting the action feed-forward amount from the fixed installation distance from the beam projection surface of the inlet wide-span laser line scanner to the center line of the cross-section of the first group of force-bearing segment support shoe corresponding to the servo hydraulic cylinder array. When the actual running displacement minus the vertical coordinate value at the top of the action delay scheduling queue equals the target trigger distance, the corresponding servo displacement command vector is extracted and synchronously sent out.
8. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 3, characterized in that, The system also includes an exit span-width laser line scanner, an exit left-side solid-state reference target bar, and an exit right-side solid-state reference target bar, after the action is executed: The exit span-width laser line scanner simultaneously covers and scans the solid reference target bar on the left side of the exit, the surface of the aluminum single plate, and the solid reference target bar on the right side of the exit to obtain the pure plate surface height matrix at the exit. The control computer performs time integration on the real-time physical running linear velocity of the exit-side conveyor mechanism to obtain the longitudinal mapping coordinates of the corresponding exit scan frame. Within the material's 2D coordinate system, data nodes with the same horizontal pixel index and the same vertical mapping coordinate are searched. The outlet pure plate height matrix and the target flat state ideal height matrix are spatially registered at the same coordinate position, and a topological residual matrix is generated by comparison.
9. The online automatic detection and correction method for flatness defects of aluminum single-panel panels according to claim 8, characterized in that, The control computer extracts the topological residual matrix and the actual displacement data fed back by the servo hydraulic cylinder array under steady-state speed and rated pressure range conditions. The topological residual matrix is transformed into an actual elongation residual vector, and the actual executed displacement data is combined into an actual displacement column vector. The error gradient matrix is obtained by taking the outer product of the actual elongation residual vector and the actual displacement column vector using a recursive algorithm. The error gradient matrix is then added to the static influence coefficient matrix along with a preset iteration step size to update and generate a dynamic influence coefficient matrix to replace the next control cycle.
10. The online automatic detection and correction method for flatness defects of aluminum single-panel according to claim 9, characterized in that, The control computer retrieves the mechanical stiffness limit parameters and maps them to the system's matrix parameter space, generating an upper limit constraint matrix and a lower limit constraint matrix that are consistent with the dimensions of the dynamic influence coefficient matrix. A boundary projection operation is performed on the dynamic influence coefficient matrix. When a matrix element is greater than the upper limit constraint matrix value or less than the lower limit constraint matrix value, it is forcibly modified to the corresponding boundary value, and a restricted dynamic influence coefficient matrix is assembled to replace the static influence coefficient matrix of the next control cycle.
