An Adaptive Hydraulic Servo Leveling Method and System Based on Point Cloud Data

Abstract

This invention discloses an adaptive hydraulic servo leveling method and system based on point cloud data, belonging to the field of industrial control technology. The invention acquires three-dimensional point cloud data by scanning the bottom surface of the workpiece, and after denoising and filtering, accurately registers it with the workpiece's CAD model. It calculates the flatness deviation and generates a three-dimensional warping deformation map. Based on the map and the hydraulic cylinder position distribution, an optimal collaborative lifting path is generated through a multi-objective path planning algorithm. A model predictive control algorithm drives the collaborative lifting of multiple hydraulic cylinders, and accuracy is verified through a second scan, forming a closed-loop control. After achieving the target, the workpiece is hydraulically locked and delivered for processing. This invention achieves accurate perception and adaptive leveling of warping deformation in large workpieces, improving leveling accuracy and effectively overcoming the problems of low efficiency and poor accuracy in traditional manual leveling, significantly improving processing quality and efficiency.

Description

Technical Field

[0001] This invention belongs to the field of industrial control technology, and specifically relates to an adaptive hydraulic servo leveling method and system based on point cloud data. Background Technology

[0002] In the precision machining of large wear-resistant parts (such as vertical mill liners and roller sleeves), the workpiece is prone to warping deformation due to its own weight and internal stress. Traditional clamping methods rely on manual "dimension alignment - copper shims" experience, which is inefficient and has poor accuracy (flatness is usually >0.1mm), resulting in vibration and chatter, out-of-tolerance accuracy, and tool wear during machining. Existing leveling technologies mostly use mechanical shims or simple hydraulic adjustments, which lack self-adaptive capabilities and cannot compensate for deformation in real time.

[0003] Although some high-end equipment has introduced laser measurement, the data processing is simple and the leveling strategy is fixed, making it difficult to deal with complex deformations. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the problems in related technologies, this invention provides an adaptive hydraulic servo leveling method and system based on point cloud data, thereby overcoming the aforementioned technical problems in existing related technologies.

[0006] (II) Technical Solution

[0007] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0008] An adaptive hydraulic servo leveling method based on point cloud data includes the following steps:

[0009] S1. Based on the workpiece CAD model, scan the bottom surface of the workpiece to obtain the three-dimensional point cloud data of the bottom surface of the workpiece;

[0010] S2. Denoise and filter the 3D point cloud data of the bottom surface of the workpiece to obtain the processed 3D point cloud data of the bottom surface of the workpiece; register the processed 3D point cloud data of the bottom surface of the workpiece using the iterative nearest point algorithm to obtain the registered 3D point cloud data of the bottom surface of the workpiece.

[0011] S3. Calculate the flatness deviation of the three-dimensional point cloud data of the workpiece bottom surface after registration, and generate a three-dimensional map of the workpiece bottom surface warping deformation; based on the three-dimensional map of the workpiece bottom surface warping deformation and according to the distribution of hydraulic cylinder positions, obtain the optimal collaborative lifting path through a path planning algorithm.

[0012] S4. Based on the optimal collaborative lifting path and combined with the model predictive control algorithm, multi-hydraulic cylinder collaborative lifting is performed to obtain the lifted workpiece.

[0013] S5. Scan the bottom surface of the lifted workpiece to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece; perform accuracy judgment based on the three-dimensional point cloud data of the bottom surface of the second workpiece; if the accuracy judgment meets the standard, proceed to S6; otherwise, return to S3 based on the three-dimensional point cloud data of the bottom surface of the second workpiece until the accuracy judgment meets the standard.

[0014] S6. Use a hydraulic cylinder to hydraulically lock the workpiece, release the safety lock, and deliver it.

[0015] This invention acquires 3D point cloud data of the workpiece's bottom surface through scanning. After denoising and registration, it analyzes flatness deviations and generates a warping deformation map. Based on this, it plans the optimal collaborative lifting path and achieves precise lifting with multiple hydraulic cylinders using model predictive control. Finally, it verifies accuracy through a second scan and completes hydraulic locking for delivery. This invention achieves improved leveling accuracy while realizing a technological leap from traditional manual experience-based leveling to fully automated, data-driven leveling. It achieves efficient handling of complex deformation patterns while maintaining an optimal balance between leveling accuracy and efficiency. It also achieves stable and reliable leveling quality control, forming a complete closed-loop control system of measurement, planning, execution, and verification, providing a reliable guarantee for the precision machining of workpieces.

[0016] Preferably, step S1 includes the following steps:

[0017] S11. Retrieve the CAD model of the workpiece and align the coordinate system of the model with the physical coordinate system of the tooling platform to obtain the aligned coordinate system.

[0018] Based on the aligned coordinate system and the bottom geometry of the workpiece's CAD model, a scanning path is planned to obtain the scanning motion trajectory.

[0019] S12. Scan the bottom surface of the workpiece by scanning the motion trajectory to obtain the initial three-dimensional point cloud data of the bottom surface of the workpiece;

[0020] Set coverage threshold and point cloud density threshold; detect the coverage and point cloud density of the initial workpiece bottom surface three-dimensional point cloud data, and determine the accuracy by combining the coverage threshold and point cloud density threshold;

[0021] If the coverage and point cloud density of the initial workpiece bottom surface 3D point cloud data mentioned in the accuracy determination are both greater than the coverage threshold and the point cloud density threshold, then the accuracy is determined to be up to standard, and the initial workpiece bottom surface 3D point cloud data is used as the workpiece bottom surface 3D point cloud data; otherwise, the causes of insufficient coverage or point cloud density in the initial workpiece bottom surface 3D point cloud data are collected and detected to obtain data hole data; a second scan is performed based on the data-controlled data until the accuracy is up to standard.

[0022] This invention performs workpiece bottom surface scanning through coordinate alignment and intelligent path planning, and intelligently judges data quality by setting point cloud coverage and density thresholds, automatically identifying data holes and triggering supplementary scanning; thus achieving the effect of ensuring the integrity and accuracy of 3D point cloud data, realizing the automation and intelligence of the scanning process, effectively eliminating the data loss problem caused by factors such as workpiece surface reflection and occlusion, and laying a reliable data foundation for subsequent precise leveling.

[0023] Preferably, step S2 includes the following steps:

[0024] S21. Perform outlier removal and smoothing filtering on the three-dimensional point cloud data of the bottom surface of the workpiece to obtain the processed three-dimensional point cloud data of the bottom surface of the workpiece.

[0025] S22. Find the nearest point to each point in the distance-processed 3D point cloud data of the bottom surface of the workpiece on the surface of the workpiece CAD model, establish point pairs, and obtain the point pair set of the bottom surface of the workpiece.

[0026] S23. By continuously rotating and translating the point pairs in the workpiece bottom surface point pair set, and using the iterative nearest point algorithm, calculate the overall distance error between the point cloud and the model to obtain the distance error set; select the rigid body transformation matrix corresponding to the smallest distance error in the distance error set to obtain the optimal rigid body transformation matrix; move the point cloud in the processed workpiece bottom surface three-dimensional point cloud data according to the optimal rigid body transformation matrix to obtain the initially registered workpiece bottom surface three-dimensional point cloud data.

[0027] S24. Set the error threshold; repeat S22 and S23. When there are distance errors in the distance error set that are less than the error threshold, stop the iteration and obtain the registered three-dimensional point cloud data of the bottom surface of the workpiece.

[0028] This invention effectively eliminates measurement noise and preserves the true shape by performing outlier removal, smoothing filtering, and iterative nearest-point registration on the three-dimensional point cloud data of the workpiece bottom surface. It achieves high-precision automatic alignment between the measured point cloud and the CAD model, laying a reliable benchmark for subsequent accurate flatness deviation calculation.

[0029] Preferably, step S3 includes the following steps:

[0030] S31. For each data point in the 3D point cloud data of the bottom surface of the registered workpiece, calculate the shortest distance from the data point to the theoretical surface of the CAD model along the Z-axis of the coordinate system to obtain the set of height deviations.

[0031] S32. Take the height deviation of each data point in the height deviation set as the Z-axis coordinate, and keep the X and Y coordinates of each data point unchanged to obtain the three-dimensional map of the warping deformation of the bottom surface of the workpiece.

[0032] S33. Based on the three-dimensional map of the warping deformation of the workpiece bottom surface, and according to the position distribution of the hydraulic cylinders, a multi-objective function is constructed and the optimal collaborative lifting path is obtained through a path planning algorithm.

[0033] This invention generates a three-dimensional warping deformation map by calculating the height deviation between the point cloud and the theoretical model, and realizes accurate quantification of workpiece deformation and intelligent optimization decision-making of leveling strategy by planning the lifting path based on the map and multi-objective optimization algorithm.

[0034] Preferably, step S33 includes the following steps:

[0035] S331. Construct a multi-objective optimization problem; the multi-objective optimization problem is to find a target lifting displacement for each hydraulic cylinder so that the leveling accuracy is the highest and the efficiency is the greatest, that is, the sum of the absolute values ​​of the residual deviations at all support points after leveling is the smallest and the total leveling time is the shortest.

[0036] Define weighting coefficients for accuracy and efficiency in a multi-objective optimization problem; based on these weighting coefficients, transform the multi-objective optimization problem into a multi-objective function.

[0037] S332. Set the maximum number of optimization iterations; construct a chromosome population; each chromosome in the chromosome population is a leveling scheme, including displacement gene segments and sequence gene segments, the displacement gene segments include the target lifting displacement of each hydraulic cylinder, and the sequence gene segments include the lifting sequence of each hydraulic cylinder;

[0038] S333. Start iteration. In each iteration, calculate the fitness value of each chromosome in the chromosome population according to the multi-objective function to obtain the set of chromosome fitness values. Select chromosomes in the chromosome population according to the fitness values ​​in the set of chromosome fitness values ​​to obtain the selected chromosome population.

[0039] Crossover and mutation operations are performed on the chromosomes in the selected chromosome population to obtain the manipulated chromosome population; in each iteration, the current best leveling scheme is selected based on the chromosome fitness values ​​in the set of chromosome fitness values;

[0040] S334, repeat S332 and S333, and when the maximum number of iterations is reached, the optimal leveling scheme is obtained; the optimal leveling scheme is used as the optimal collaborative lifting path;

[0041] This invention constructs a multi-objective optimization function that simultaneously considers leveling accuracy and efficiency, and uses a genetic algorithm to collaboratively optimize the hydraulic cylinder displacement and lifting sequence. This achieves the effect of rapidly generating the optimal leveling scheme under complex deformation conditions, while realizing the best balance between accuracy requirements and time costs during the leveling process, significantly improving the system's adaptive capability and overall leveling efficiency.

[0042] Preferably, step S4 includes the following steps:

[0043] S41. The optimal collaborative lifting path is converted into control commands; all hydraulic cylinders execute the control commands, and after execution, the real-time position, movement speed, and drive current of each hydraulic cylinder are collected to obtain the real-time data of the hydraulic cylinders.

[0044] S42. Input the real-time data of the hydraulic cylinder into the model predictive control algorithm to obtain real-time dynamic hydraulic cylinder state prediction data;

[0045] The hydraulic cylinder is adjusted by feedforward compensation based on real-time dynamic hydraulic cylinder status prediction data and PID control system.

[0046] S43. Set the hydraulic cylinder position accuracy threshold; repeat S42 to calculate the distance between all hydraulic cylinders and the target position in the optimal collaborative lifting path, and obtain the hydraulic cylinder position accuracy value.

[0047] When the position accuracy value of the hydraulic cylinder is less than the position accuracy threshold of the hydraulic cylinder, the iteration stops and the lifted workpiece is obtained.

[0048] This invention transforms the optimal path into control commands and employs a model predictive control combined with a PID feedforward compensation strategy to adjust the hydraulic cylinder action in real time. This achieves precise compensation for the nonlinear characteristics of the hydraulic system and high-precision closed-loop control of the lifting process, effectively eliminating overshoot and oscillation problems and ensuring that the workpiece reaches the target position smoothly and accurately.

[0049] Preferably, the model predictive control algorithm in S42 is obtained through the following steps:

[0050] S421. Collect historical hydraulic cylinder data and label the historical hydraulic cylinder data to obtain labeled historical hydraulic cylinder data; set the training data ratio and the test data ratio; divide the labeled historical hydraulic cylinder data into training set and test set according to the training data ratio and the test data ratio.

[0051] S422. Construct an initial BP neural network model; train and test the initial BP neural network model using a training set and a test set to obtain a model predictive control algorithm; adjust the model parameters according to the training results of each round during training and testing, and obtain the model predictive control algorithm when the test standard is met; the test standard is set according to specific requirements.

[0052] This invention trains a BP neural network model by collecting historical hydraulic cylinder data, and constructs a data-driven model predictive control algorithm. It improves control accuracy by utilizing historical system operation data, realizes autonomous optimization and adaptive adjustment of controller parameters, and significantly reduces reliance on manual debugging experience.

[0053] Preferably, step S5 includes the following steps:

[0054] S51. Scan the bottom surface of the workpiece after it has been lifted to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece.

[0055] S52. Calculate the shortest distance from the data points in the three-dimensional point cloud data of the bottom surface of the second workpiece to the theoretical surface of the CAD model, and obtain the set of residual height deviations.

[0056] S53. Calculate the actual flatness of the current workpiece based on the residual height deviation value in the residual height deviation set, and obtain the real-time workpiece flatness.

[0057] Set a flatness threshold; based on the flatness threshold and the real-time workpiece flatness, determine the accuracy; if the real-time workpiece flatness is less than the flatness threshold, the accuracy is determined to meet the standard, and S6 is executed; otherwise, based on the three-dimensional point cloud data of the bottom surface of the second workpiece, return to execute steps S3 and S4 until the accuracy is determined to meet the standard.

[0058] This invention acquires point cloud data of the leveled workpiece through secondary scanning, calculates the residual height deviation, and judges the flatness accuracy. It achieves the effect of objectively and quantitatively verifying the leveling results, forming a closed-loop control mechanism of measurement-adjustment-verification, ensuring that the flatness of the workpiece accurately meets the standard, and effectively preventing defective products from flowing into subsequent processing stages.

[0059] Preferably, step S6 includes the following steps:

[0060] S61. Once the accuracy is determined to be up to standard, switch all hydraulic cylinders from position control mode to pressure holding mode to hydraulically lock the workpiece.

[0061] S62. After hydraulic locking, the safety pre-lock of the auxiliary clamping device is released and a signal indicating leveling completion and start of machining is issued, and the machine tool can then execute the subsequent finishing program.

[0062] This invention uses hydraulic locking technology to firmly fix the leveled workpiece and automatically sends a processing ready signal to maintain the workpiece in a precise leveling state. At the same time, it achieves seamless connection between leveling and processing, significantly improving the automation level of the production process.

[0063] An adaptive hydraulic servo leveling system based on point cloud data is used to implement the aforementioned adaptive hydraulic servo leveling method based on point cloud data. It includes a data acquisition and initialization module, a point cloud processing and registration module, a deviation analysis and path planning module, a lifting control and execution module, a precision verification and iteration module, and a locking and delivery module.

[0064] The data acquisition and initialization module is used for system initialization and workpiece positioning. This includes the control unit performing status diagnostics on devices such as hydraulic servo cylinders, displacement sensors, and 3D laser scanners, and retracting all hydraulic cylinder piston rods to a unified initial position. After the workpiece is hoisted onto the tooling platform using an overhead crane, it is initially limited by an auxiliary clamping device to prevent slippage. The workpiece CAD model is retrieved and aligned with the physical coordinate system, and the scanning path is planned to ensure that the laser line fully covers the bottom surface of the workpiece to avoid collisions. Finally, the 3D point cloud data of the bottom surface of the workpiece is obtained by scanning, and the data quality is judged by a threshold. If necessary, additional scanning is performed to ensure accuracy.

[0065] The point cloud processing and registration module preprocesses the collected 3D point cloud data of the workpiece bottom surface, including using a statistical filtering algorithm to remove outliers and smoothing the data using Gaussian filtering or median filtering to preserve the true shape features; it then registers the processed point cloud with the CAD model using an iterative nearest point algorithm, continuously rotating and translating the point pairs to calculate the minimum distance error until the point cloud and the model are in the same precise coordinate system, laying the foundation for subsequent deviation calculations.

[0066] The deviation analysis and path planning module calculates the height deviation between each point in the registered point cloud data and the theoretical surface of the CAD model under a unified coordinate system, and generates a three-dimensional map of the workpiece bottom surface warping deformation visualized by color mapping. Based on this map and the hydraulic cylinder position distribution, a multi-objective function (considering leveling accuracy and efficiency) is constructed, and the lifting path is optimized by path planning methods such as genetic algorithms, including determining the target lifting displacement and lifting sequence of each hydraulic cylinder to minimize residual deviation and total leveling time.

[0067] The lifting control and execution module converts the optimal collaborative lifting path into control commands and adjusts the hydraulic cylinder actions in real time through a model predictive control algorithm. The algorithm predicts the hydraulic cylinder state based on a BP neural network trained on historical data and combines it with a PID control system for feedforward compensation to ensure that the hydraulic cylinders are lifted accurately according to the planned path. Iterative adjustments are made until the position accuracy of all hydraulic cylinders meets the threshold, and the lifted workpiece is obtained.

[0068] After lifting, the accuracy verification and iteration module scans the bottom surface of the workpiece again to obtain the three-dimensional point cloud data of the second workpiece bottom surface, calculates the residual height deviation and the real-time workpiece flatness, and judges the accuracy by comparing the flatness with a preset threshold: if it meets the standard, it proceeds to the next module; otherwise, it returns to the deviation analysis and path planning module based on the new point cloud data, re-optimizes the lifting path and executes the lifting until the accuracy meets the standard or the maximum number of iterations is reached, and then an alarm is triggered.

[0069] Once the required precision is achieved, the locking and delivery module switches all hydraulic cylinders from position control mode to pressure holding mode to hydraulically lock the workpiece, ensuring it is firmly supported in the leveling posture. Then, it releases the safety pre-lock of the auxiliary clamping device and sends a leveling completion signal, allowing the machine tool to execute subsequent finishing procedures and ultimately deliver the workpiece.

[0070] (III) Beneficial Effects

[0071] The present invention has the following beneficial effects:

[0072] This invention achieves precise perception and quantitative analysis of warping deformation of large workpieces through high-precision 3D point cloud acquisition and intelligent processing. It utilizes 3D laser scanning technology to acquire complete 3D topographic data of the workpiece's bottom surface, and combines this with an iterative nearest-point algorithm to accurately register the measured point cloud with the CAD model, enabling accurate identification of microscopic concave-convex deformation of the workpiece's bottom surface. By generating a visualized 3D warping deformation map, it transforms abstract flatness deviations into intuitive color distribution maps, providing a reliable data foundation for subsequent leveling strategy formulation. This overcomes the subjectivity and inaccuracy of traditional manual methods and improves leveling accuracy.

[0073] This invention innovatively applies a multi-objective optimization algorithm to hydraulic leveling path planning, achieving the best balance between accuracy and efficiency. Based on a genetic algorithm, a chromosome model containing displacement gene segments and sequence gene segments is constructed. Through iterative optimization, the optimal collaborative lifting path that minimizes residual deviation and shortens leveling time is calculated. It not only considers the lifting displacement of each hydraulic cylinder but also optimizes the action sequence, effectively avoiding common problems in traditional leveling such as over-adjustment and oscillation, significantly improving leveling efficiency, while ensuring the ability to adaptively handle complex deformation modes.

[0074] This invention constructs a complete intelligent leveling control system through model predictive control and closed-loop verification mechanisms. The system transforms the optimized lifting path into specific control commands, uses a model predictive control algorithm based on a BP neural network to predict changes in the hydraulic cylinder state in real time, and combines PID control for feedforward compensation, effectively overcoming the nonlinearity and time-varying nature of the hydraulic system. Through re-scanning and accuracy verification after lifting, a complete closed-loop control is formed to ensure that the leveling results are reliable and meet the standards. This adaptive leveling method significantly reduces the reliance on operator experience, improves the consistency and reliability of the processing, and provides strong technical support for the precision machining of large workpieces.

[0075] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0076] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0077] Figure 1 This is a flowchart illustrating an adaptive hydraulic servo leveling method based on point cloud data according to the present invention.

[0078] Figure 2 This is a schematic diagram of a module of an adaptive hydraulic servo leveling system based on point cloud data according to the present invention. Detailed Implementation

[0079] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0080] Example 1:

[0081] To resolve the above technical issues, please refer to [link / reference]. Figure 1 This invention discloses an adaptive hydraulic servo leveling method based on point cloud data, comprising the following steps:

[0082] S1. Based on the workpiece CAD model, scan the bottom surface of the workpiece to obtain the three-dimensional point cloud data of the bottom surface of the workpiece;

[0083] S1 includes the following steps:

[0084] Before scanning the workpiece, system initialization and workpiece positioning are required. Specifically: the control unit performs status diagnostics on the hydraulic servo cylinders, displacement sensors, 3D laser scanner, and communication bus; all servo hydraulic cylinder piston rods retract to a unified initial position (zero position) to prepare for receiving the workpiece; the workpiece (such as a vertical mill liner) is hoisted onto the tooling platform using an overhead crane and placed in the center area of ​​the support pad; the peripheral light auxiliary clamping device is activated to initially limit the workpiece and prevent accidental slippage during scanning and leveling.

[0085] S11. Retrieve the CAD model of the workpiece and align the coordinate system of the model with the physical coordinate system of the tooling platform to obtain the aligned coordinate system.

[0086] Based on the aligned coordinate system and the bottom geometry of the workpiece's CAD model, a scanning path is planned to obtain the scanning motion trajectory. Specifically, the scanning motion trajectory enables the scanner's laser line to cover the entire bottom surface of the workpiece vertically or at the optimal angle, as well as the potential contact areas of each support pad. Collision interference simulation is performed during path planning to ensure that the robotic arm does not collide with the workpiece, hydraulic cylinder, or platform during the movement.

[0087] S12. Scan the bottom surface of the workpiece by scanning the motion trajectory to obtain the initial three-dimensional point cloud data of the bottom surface of the workpiece;

[0088] Set coverage threshold and point cloud density threshold; detect the coverage and point cloud density of the initial workpiece bottom surface three-dimensional point cloud data, and determine the accuracy by combining the coverage threshold and point cloud density threshold;

[0089] If the coverage and point cloud density of the initial workpiece bottom surface 3D point cloud data mentioned in the accuracy determination are both greater than the coverage threshold and the point cloud density threshold, then the accuracy is determined to be up to standard, and the initial workpiece bottom surface 3D point cloud data is used as the workpiece bottom surface 3D point cloud data; otherwise, the causes of insufficient coverage or point cloud density in the initial workpiece bottom surface 3D point cloud data (such as workpiece surface reflection, deep grooves or occlusion) are collected and detected to obtain data voids; a second scan is performed based on the data-controlled data until the accuracy is up to standard.

[0090] S2. Denoise and filter the 3D point cloud data of the bottom surface of the workpiece to obtain the processed 3D point cloud data of the bottom surface of the workpiece; register the processed 3D point cloud data of the bottom surface of the workpiece using the iterative nearest point algorithm to obtain the registered 3D point cloud data of the bottom surface of the workpiece.

[0091] S2 includes the following steps:

[0092] S21. Perform outlier removal and smoothing filtering on the three-dimensional point cloud data of the bottom surface of the workpiece to obtain the processed three-dimensional point cloud data of the bottom surface of the workpiece.

[0093] Specifically: The outlier removal uses a statistical filtering algorithm to analyze the distance distribution between each point and its nearest neighbor in the three-dimensional point cloud data of the workpiece bottom surface; discrete noise points (such as those caused by reflection or dust) whose distance exceeds a certain multiple (such as 2 times) of the standard deviation are automatically identified and deleted;

[0094] The smoothing filtering process uses Gaussian filtering or median filtering algorithms to smooth the three-dimensional point cloud data of the workpiece bottom surface, suppress random measurement noise, and retain the true macroscopic morphological features of the workpiece.

[0095] S22. Find the nearest point to each point in the distance-processed 3D point cloud data of the bottom surface of the workpiece on the surface of the workpiece CAD model, establish point pairs, and obtain the point pair set of the bottom surface of the workpiece.

[0096] S23. By continuously rotating and translating the point pairs in the workpiece bottom surface point pair set, and using the iterative nearest point algorithm, the overall distance error between the point cloud and the model is calculated to obtain the distance error set; the rigid body transformation matrix corresponding to the smallest distance error in the distance error set is selected to obtain the optimal rigid body transformation matrix; the point cloud in the processed workpiece bottom surface 3D point cloud data is moved to a new position closer to the model according to the optimal rigid body transformation matrix to obtain the initially registered workpiece bottom surface 3D point cloud data.

[0097] S24. Set the error threshold; repeat S22 and S23. When there are distance errors in the distance error set that are less than the error threshold, stop the iteration and obtain the registered three-dimensional point cloud data of the bottom surface of the workpiece.

[0098] After registration, the registered 3D point cloud data of the workpiece bottom surface and the ideal CAD model are in the same precise coordinate system, laying the foundation for the next step of deviation calculation.

[0099] S3. Calculate the flatness deviation of the three-dimensional point cloud data of the workpiece bottom surface after registration, and generate a three-dimensional map of the workpiece bottom surface warping deformation; based on the three-dimensional map of the workpiece bottom surface warping deformation and according to the distribution of hydraulic cylinder positions, obtain the optimal collaborative lifting path through a path planning algorithm.

[0100] S3 includes the following steps:

[0101] S31. Under a unified coordinate system, for each data point in the 3D point cloud data of the registered workpiece bottom surface, calculate the shortest distance from the data point to the theoretical surface of the CAD model (i.e., the ideal plane or curved surface) along the Z-axis (vertical direction) of the coordinate system to obtain the height deviation set; a positive height deviation value Δh for each point in the height deviation set indicates that the data point is higher than the theoretical surface of the CAD model, while a negative value indicates that it is concave.

[0102] S32. Using the height deviation of each data point in the height deviation set as the Z-axis coordinate, while keeping the X and Y coordinates of each data point unchanged, a three-dimensional map of the workpiece bottom surface warping deformation is obtained. The three-dimensional map of the workpiece bottom surface warping deformation is visualized through color mapping, such as using blue to represent negative deviation (concavity), red to represent positive deviation (convexity), and green to represent zero deviation. This colored three-dimensional map intuitively shows the unevenness distribution, amplitude, and shape of the workpiece bottom surface.

[0103] S33. Based on the three-dimensional map of the warping deformation of the workpiece bottom surface, and according to the position distribution of the hydraulic cylinders, a multi-objective function is constructed and the optimal collaborative lifting path is obtained through a path planning algorithm.

[0104] S33 includes the following steps:

[0105] S331. Construct a multi-objective optimization problem; the multi-objective optimization problem is to find a target lifting displacement for each hydraulic cylinder so that the leveling accuracy is the highest and the efficiency is the greatest, that is, the sum of the absolute values ​​of the residual deviations at all support points after leveling is the smallest and the total leveling time is the shortest.

[0106] In a multi-objective optimization problem, weight coefficients a1 and a2 are set for accuracy and efficiency. Based on these weight coefficients, the multi-objective optimization problem is transformed into a multi-objective function. The formula for the multi-objective function F is as follows.

[0107] ;

[0108] Among them, (∑|d i -x i |) represents the residual deviation, d i x represents the current height deviation of the area corresponding to the i-th hydraulic cylinder. i This represents the target lifting displacement of the i-th hydraulic cylinder; Let v represent the total leveling time, v represent the average lifting speed of the hydraulic cylinders, Z represent the set of lifting sequences, i.e., the activation sequence of the hydraulic cylinders, Δt represent the switching time during sequential lifting of the hydraulic cylinders, and max(x) represent the total leveling time. i ) / v represents the time of the parallel lifting phase (determined by the cylinder with the largest displacement). This indicates the time cost of the sequential jacking phase;

[0109] S332. Set the maximum number of optimization iterations; construct a chromosome population, setting the size of the chromosome population to p, then the chromosome population is represented as... q i Represents the first chromosome in the population. Each chromosome in the chromosome population is a leveling scheme, including displacement gene segments and sequence gene segments. The displacement gene segments include the target lifting displacement of each hydraulic cylinder, and the sequence gene segments include the lifting sequence of each hydraulic cylinder.

[0110] Taking the chromosome 1 of four hydraulic cylinders as an example, chromosome 1 = [2.5, 1.8, 3.2, 0.9; 3, 1, 4, 2]; where the displacement scheme is cylinder 1 lifting 2.5mm, cylinder 2 lifting 1.8mm, cylinder 3 lifting 3.2mm, and cylinder 4 lifting 0.9mm, and the lifting sequence is cylinder 3, cylinder 1, cylinder 4, cylinder 2;

[0111] S333. Start iteration. In each iteration, calculate the fitness value of each chromosome in the chromosome population according to the multi-objective function to obtain the set of chromosome fitness values. Select chromosomes in the chromosome population according to the fitness values ​​in the set of chromosome fitness values ​​to obtain the selected chromosome population. The smaller the fitness value, the better the leveling scheme.

[0112] Crossover and mutation operations are performed on the chromosomes in the selected chromosome population to obtain the manipulated chromosome population; in each iteration, the current best leveling scheme is selected based on the chromosome fitness values ​​in the set of chromosome fitness values;

[0113] S334, repeat S332 and S333, and when the maximum number of iterations is reached, the optimal leveling scheme is obtained; the optimal leveling scheme is used as the optimal collaborative lifting path;

[0114] S4. Based on the optimal collaborative lifting path and combined with the model predictive control algorithm, multi-hydraulic cylinder collaborative lifting is performed to obtain the lifted workpiece.

[0115] S4 includes the following steps:

[0116] S41. Convert the optimal coordinated lifting path into control commands; the control commands include a target position array, motion curve parameters, and a coordinated action group; the motion curve parameters include the maximum speed, acceleration, and deceleration of the hydraulic cylinders; the coordinated action group is used to define which hydraulic cylinders move synchronously and which move sequentially.

[0117] All hydraulic cylinders execute control commands, and after execution, real-time data of the hydraulic cylinders is obtained by collecting the real-time position, movement speed, and drive current (indirectly reflecting the load force) of each hydraulic cylinder.

[0118] S42. Input the real-time data of the hydraulic cylinder into the model predictive control algorithm to obtain real-time dynamic hydraulic cylinder state prediction data; the real-time dynamic system prediction data includes predicting uneven load pressure caused by the offset of the workpiece center of gravity, predicting the coupling interference when multiple hydraulic cylinders move, and predicting the overshoot or oscillation trend that the system may have.

[0119] The hydraulic cylinder is adjusted by feedforward compensation based on real-time dynamic hydraulic cylinder status prediction data and PID control system.

[0120] S43. Set the hydraulic cylinder position accuracy threshold; repeat S42 to calculate the distance between all hydraulic cylinders and the target position in the optimal collaborative lifting path, and obtain the hydraulic cylinder position accuracy value.

[0121] When the position accuracy value of the hydraulic cylinder is less than the position accuracy threshold of the hydraulic cylinder, the iteration stops and the lifted workpiece is obtained.

[0122] The model predictive control algorithm in S42 is obtained through the following steps:

[0123] S421. Collect historical hydraulic cylinder data and assign labels to the historical hydraulic cylinder data to obtain tagged historical hydraulic cylinder data; set the training data ratio and the test data ratio; divide the tagged historical hydraulic cylinder data into a training set and a test set according to the training data ratio and the test data ratio; the historical hydraulic cylinder data includes historical control command data, historical hydraulic cylinder status data, and historical operating condition data; the label is the hydraulic cylinder data of the historical hydraulic cylinder data in the next cycle;

[0124] S422. Construct an initial BP neural network model; train and test the initial BP neural network model using a training set and a test set to obtain a model predictive control algorithm; adjust the model parameters according to the training results of each round during training and testing, and obtain the model predictive control algorithm when the test standard is met; the test standard is set according to specific requirements.

[0125] S5. Scan the bottom surface of the lifted workpiece to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece; perform accuracy judgment based on the three-dimensional point cloud data of the bottom surface of the second workpiece; if the accuracy judgment meets the standard, proceed to S6; otherwise, return to S3 based on the three-dimensional point cloud data of the bottom surface of the second workpiece until the accuracy judgment meets the standard.

[0126] S5 includes the following steps:

[0127] S51. Scan the bottom surface of the workpiece after lifting to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece. The process of obtaining the three-dimensional point cloud data of the bottom surface of the second workpiece is the same as that of S1 and S2, but no re-registration is required because the coordinate transformation relationship between the workpiece and the CAD model has been established during the first scan in S1. During the fine-tuning process, the macroscopic position of the workpiece has not changed, so the coordinate system established in the first scan can be used directly.

[0128] S52. Calculate the shortest distance from the data points in the three-dimensional point cloud data of the bottom surface of the second workpiece to the theoretical surface of the CAD model (i.e., the ideal plane or curved surface), and obtain the set of residual height deviations;

[0129] S53. Calculate the actual flatness of the current workpiece (i.e., the distance between the highest and lowest points on the normal vector of the reference plane) based on the residual height deviation values ​​in the residual height deviation set, and obtain the real-time workpiece flatness; the actual flatness can be calculated by the least squares method or the minimum region method.

[0130] Set a flatness threshold; based on the flatness threshold and the real-time workpiece flatness, determine the accuracy; if the real-time workpiece flatness is less than the flatness threshold, the accuracy is determined to be up to standard, and S6 is executed; otherwise, based on the 3D point cloud data of the bottom surface of the second workpiece, return to execute steps S3 and S4 until the accuracy is determined to be up to standard; if the real-time workpiece flatness cannot be less than the flatness threshold, the maximum number of fine-tuning iterations (e.g., 5 times) is preset to prevent getting stuck in an infinite loop; at this time, the system will alarm and prompt the operator to check the system or workpiece status;

[0131] S6. Use a hydraulic cylinder to hydraulically lock the workpiece, release the safety lock, and deliver it.

[0132] S6 includes the following steps:

[0133] S61. Once the accuracy is determined to be up to standard, switch all hydraulic cylinders from position control mode to pressure holding mode to hydraulically lock the workpiece. At this time, the hydraulic cylinders are equivalent to a set of high-rigidity mechanical supports, which firmly support the workpiece in the leveled ideal posture.

[0134] S62. After hydraulic locking, the safety pre-lock of the auxiliary clamping device is released and a signal indicating leveling completion and start of machining is issued, allowing the machine tool to execute the subsequent finishing program.

[0135] Example 2:

[0136] Please see Figure 2 An adaptive hydraulic servo leveling system based on point cloud data is used to implement the aforementioned adaptive hydraulic servo leveling method based on point cloud data. The system includes a data acquisition and initialization module, a point cloud processing and registration module, a deviation analysis and path planning module, a lifting control and execution module, a precision verification and iteration module, and a locking and delivery module.

[0137] The data acquisition and initialization module is used for system initialization and workpiece positioning. This includes the control unit performing status diagnostics on devices such as hydraulic servo cylinders, displacement sensors, and 3D laser scanners, and retracting all hydraulic cylinder piston rods to a unified initial position. After the workpiece is hoisted onto the tooling platform using an overhead crane, it is initially limited by an auxiliary clamping device to prevent slippage. The workpiece CAD model is retrieved and aligned with the physical coordinate system, and the scanning path is planned to ensure that the laser line fully covers the bottom surface of the workpiece to avoid collisions. Finally, the 3D point cloud data of the bottom surface of the workpiece is obtained by scanning, and the data quality is judged by a threshold. If necessary, additional scanning is performed to ensure accuracy.

[0138] The point cloud processing and registration module preprocesses the collected 3D point cloud data of the workpiece bottom surface, including using a statistical filtering algorithm to remove outliers and smoothing the data using Gaussian filtering or median filtering to preserve the true shape features; it then registers the processed point cloud with the CAD model using an iterative nearest point algorithm, continuously rotating and translating the point pairs to calculate the minimum distance error until the point cloud and the model are in the same precise coordinate system, laying the foundation for subsequent deviation calculations.

[0139] The deviation analysis and path planning module calculates the height deviation between each point in the registered point cloud data and the theoretical surface of the CAD model under a unified coordinate system, and generates a three-dimensional map of the workpiece bottom surface warping deformation visualized by color mapping. Based on this map and the hydraulic cylinder position distribution, a multi-objective function (considering leveling accuracy and efficiency) is constructed, and the lifting path is optimized by path planning methods such as genetic algorithms, including determining the target lifting displacement and lifting sequence of each hydraulic cylinder to minimize residual deviation and total leveling time.

[0140] The lifting control and execution module converts the optimal collaborative lifting path into control commands and adjusts the hydraulic cylinder actions in real time through a model predictive control algorithm. The algorithm predicts the hydraulic cylinder state based on a BP neural network trained on historical data and combines it with a PID control system for feedforward compensation to ensure that the hydraulic cylinders are lifted accurately according to the planned path. Iterative adjustments are made until the position accuracy of all hydraulic cylinders meets the threshold, and the lifted workpiece is obtained.

[0141] After lifting, the accuracy verification and iteration module scans the bottom surface of the workpiece again to obtain the three-dimensional point cloud data of the second workpiece bottom surface, calculates the residual height deviation and the real-time workpiece flatness, and judges the accuracy by comparing the flatness with a preset threshold: if it meets the standard, it proceeds to the next module; otherwise, it returns to the deviation analysis and path planning module based on the new point cloud data, re-optimizes the lifting path and executes the lifting until the accuracy meets the standard or the maximum number of iterations is reached, and then an alarm is triggered.

[0142] Once the required precision is achieved, the locking and delivery module switches all hydraulic cylinders from position control mode to pressure holding mode to hydraulically lock the workpiece, ensuring it is firmly supported in the leveling posture. Then, it releases the safety pre-lock of the auxiliary clamping device and sends a leveling completion signal, allowing the machine tool to execute subsequent finishing procedures and ultimately deliver the workpiece.

[0143] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0144] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. An adaptive hydraulic servo leveling method based on point cloud data, characterized in that, Includes the following steps: S1. Based on the workpiece CAD model, scan the bottom surface of the workpiece to obtain the three-dimensional point cloud data of the bottom surface of the workpiece; S2. Denoise and filter the three-dimensional point cloud data of the bottom surface of the workpiece to obtain the processed three-dimensional point cloud data of the bottom surface of the workpiece. The processed 3D point cloud data of the bottom surface of the workpiece is registered by the iterative nearest point algorithm to obtain the registered 3D point cloud data of the bottom surface of the workpiece. S3. Calculate the flatness deviation of the three-dimensional point cloud data of the workpiece bottom surface after registration, and generate a three-dimensional map of the workpiece bottom surface warping deformation; based on the three-dimensional map of the workpiece bottom surface warping deformation and according to the distribution of hydraulic cylinder positions, obtain the optimal collaborative lifting path through a path planning algorithm. S4. Based on the optimal collaborative lifting path and combined with the model predictive control algorithm, multi-hydraulic cylinder collaborative lifting is performed to obtain the lifted workpiece. S5. Scan the bottom surface of the workpiece after lifting to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece; Accuracy is determined based on the three-dimensional point cloud data of the bottom surface of the second workpiece; if the accuracy is satisfactory, proceed to step S6; otherwise, return to step S3 based on the three-dimensional point cloud data of the bottom surface of the second workpiece, until the accuracy is satisfactory. S6. Use a hydraulic cylinder to hydraulically lock the workpiece, release the safety lock, and deliver it.

2. The adaptive hydraulic servo leveling method based on point cloud data according to claim 1, characterized in that, S1 includes the following steps: S11. Retrieve the CAD model of the workpiece and align the coordinate system of the model with the physical coordinate system of the tooling platform to obtain the aligned coordinate system. Based on the aligned coordinate system and the bottom geometry of the workpiece's CAD model, a scanning path is planned to obtain the scanning motion trajectory. S12. Scan the bottom surface of the workpiece by scanning the motion trajectory to obtain the initial three-dimensional point cloud data of the bottom surface of the workpiece; Set coverage threshold and point cloud density threshold; detect the coverage and point cloud density of the initial workpiece bottom surface three-dimensional point cloud data, and determine the accuracy by combining the coverage threshold and point cloud density threshold; If the coverage and point cloud density of the initial workpiece bottom surface 3D point cloud data mentioned in the accuracy determination are both greater than the coverage threshold and the point cloud density threshold, then the accuracy is determined to be up to standard, and the initial workpiece bottom surface 3D point cloud data is used as the workpiece bottom surface 3D point cloud data; otherwise, the causes of insufficient coverage or point cloud density in the initial workpiece bottom surface 3D point cloud data are collected and detected to obtain data hole data; a second scan is performed based on the data-controlled data until the accuracy is up to standard.

3. The adaptive hydraulic servo leveling method based on point cloud data according to claim 2, characterized in that, S2 includes the following steps: S21. Perform outlier removal and smoothing filtering on the three-dimensional point cloud data of the bottom surface of the workpiece to obtain the processed three-dimensional point cloud data of the bottom surface of the workpiece. S22. Find the nearest point to each point in the distance-processed 3D point cloud data of the bottom surface of the workpiece on the surface of the workpiece CAD model, establish point pairs, and obtain the point pair set of the bottom surface of the workpiece. S23. By continuously rotating and translating the point pairs in the workpiece bottom surface point pair set, and using the iterative nearest point algorithm, calculate the overall distance error between the point cloud and the model to obtain the distance error set; select the rigid body transformation matrix corresponding to the smallest distance error in the distance error set to obtain the optimal rigid body transformation matrix; move the point cloud in the processed workpiece bottom surface three-dimensional point cloud data according to the optimal rigid body transformation matrix to obtain the initially registered workpiece bottom surface three-dimensional point cloud data. S24. Set the error threshold; repeat S22 and S23. When there is a distance error in the distance error set that is less than the error threshold, stop the iteration and obtain the registered three-dimensional point cloud data of the bottom surface of the workpiece.

4. The adaptive hydraulic servo leveling method based on point cloud data according to claim 3, characterized in that, S3 includes the following steps: S31. For each data point in the 3D point cloud data of the bottom surface of the registered workpiece, calculate the shortest distance from the data point to the theoretical surface of the CAD model along the Z-axis of the coordinate system to obtain the set of height deviations. S32. Take the height deviation of each data point in the height deviation set as the Z-axis coordinate, and keep the X and Y coordinates of each data point unchanged to obtain the three-dimensional map of the warping deformation of the bottom surface of the workpiece. S33. Based on the three-dimensional map of the warping deformation of the workpiece bottom surface and according to the position distribution of the hydraulic cylinder, a multi-objective function is constructed and the optimal collaborative lifting path is obtained through the path planning algorithm.

5. The adaptive hydraulic servo leveling method based on point cloud data according to claim 4, characterized in that, S33 includes the following steps: S331. Construct a multi-objective optimization problem; the multi-objective optimization problem is to find a target lifting displacement for each hydraulic cylinder so that the leveling accuracy is the highest and the efficiency is the greatest, that is, the sum of the absolute values ​​of the residual deviations at all support points after leveling is the smallest and the total leveling time is the shortest. Define weighting coefficients for accuracy and efficiency in a multi-objective optimization problem; based on these weighting coefficients, transform the multi-objective optimization problem into a multi-objective function. S332. Set the maximum number of optimization iterations; construct a chromosome population; each chromosome in the chromosome population is a leveling scheme, including displacement gene segments and sequence gene segments, the displacement gene segments include the target lifting displacement of each hydraulic cylinder, and the sequence gene segments include the lifting sequence of each hydraulic cylinder; S333. Start iteration. In each iteration, calculate the fitness value of each chromosome in the chromosome population according to the multi-objective function to obtain the set of chromosome fitness values. Select chromosomes in the chromosome population according to the fitness values ​​in the set of chromosome fitness values ​​to obtain the selected chromosome population. Crossover and mutation operations are performed on the chromosomes in the selected chromosome population to obtain the manipulated chromosome population; in each iteration, the current best leveling scheme is selected based on the chromosome fitness values ​​in the set of chromosome fitness values; S334, repeat S332 and S333, and when the maximum number of iterations is reached, the optimal leveling scheme is obtained; the optimal leveling scheme is used as the optimal collaborative lifting path.

6. The adaptive hydraulic servo leveling method based on point cloud data according to claim 4, characterized in that, S4 includes the following steps: S41. The optimal collaborative lifting path is converted into control commands; all hydraulic cylinders execute the control commands, and after execution, the real-time position, movement speed, and drive current of each hydraulic cylinder are collected to obtain the real-time data of the hydraulic cylinders. S42. Input the real-time data of the hydraulic cylinder into the model predictive control algorithm to obtain real-time dynamic hydraulic cylinder state prediction data; The hydraulic cylinder is adjusted by feedforward compensation based on real-time dynamic hydraulic cylinder status prediction data and PID control system. S43. Set the hydraulic cylinder position accuracy threshold; repeat S42 to calculate the distance between all hydraulic cylinders and the target position in the optimal collaborative lifting path, and obtain the hydraulic cylinder position accuracy value. When the position accuracy value of the hydraulic cylinder is less than the position accuracy threshold of the hydraulic cylinder, the iteration stops and the lifted workpiece is obtained.

7. The adaptive hydraulic servo leveling method based on point cloud data according to claim 6, characterized in that, The model predictive control algorithm in S42 is obtained through the following steps: S421. Collect historical hydraulic cylinder data and label the historical hydraulic cylinder data to obtain labeled historical hydraulic cylinder data; set the training data ratio and the test data ratio; divide the labeled historical hydraulic cylinder data into training set and test set according to the training data ratio and the test data ratio. S422. Construct an initial BP neural network model; train and test the initial BP neural network model using the training set and test set to obtain the model predictive control algorithm; During training and testing, the model parameters are adjusted based on the training results of each round. When the testing criteria are met, the model prediction and control algorithm is obtained. The testing criteria are set according to specific requirements.

8. The adaptive hydraulic servo leveling method based on point cloud data according to claim 6, characterized in that, S5 includes the following steps: S51. Scan the bottom surface of the workpiece after it has been lifted to obtain the three-dimensional point cloud data of the bottom surface of the second workpiece. S52. Calculate the shortest distance from the data points in the three-dimensional point cloud data of the bottom surface of the second workpiece to the theoretical surface of the CAD model, and obtain the set of residual height deviations. S53. Calculate the actual flatness of the current workpiece based on the residual height deviation value in the residual height deviation set, and obtain the real-time workpiece flatness. Set a flatness threshold; based on the flatness threshold and the real-time workpiece flatness, determine the accuracy; if the real-time workpiece flatness is less than the flatness threshold, the accuracy is determined to be up to standard, and S6 is executed; otherwise, based on the three-dimensional point cloud data of the bottom surface of the second workpiece, return to execute steps S3 and S4 until the accuracy is determined to be up to standard.

9. The adaptive hydraulic servo leveling method based on point cloud data according to claim 8, characterized in that, S6 includes the following steps: S61. Once the accuracy is determined to be up to standard, switch all hydraulic cylinders from position control mode to pressure holding mode to hydraulically lock the workpiece. S62. After hydraulic locking, the safety pre-lock of the auxiliary clamping device is released and a signal indicating leveling completion and start of machining is issued, allowing the machine tool to execute the subsequent finishing program.

10. An adaptive hydraulic servo leveling system based on point cloud data, characterized in that, The system is used to implement an adaptive hydraulic servo leveling method based on point cloud data as described in any one of claims 1-9. The system includes a data acquisition and initialization module, a point cloud processing and registration module, a deviation analysis and path planning module, a lifting control and execution module, a precision verification and iteration module, and a locking and delivery module.

Images