A method and system for predicting and controlling straightening stroke of linear guide

Abstract

The present application relates to the technical field of precision machinery manufacturing, and discloses a straight line guide rail straightening stroke prediction and control method and system, which solves the problems of existing straightening process relying on experience, low precision and poor efficiency. The present application establishes the nonlinear bending moment-curvature constitutive relationship of the straight line guide rail under three-point bending straightening, and constructs the implicit function model between the straightening force and the straightening stroke. A two-layer nested iteration algorithm is used, the outer layer iteratively optimizes the straightening force, and the inner layer iteratively solves the curvature distribution based on the Newton-Raphson method, so as to realize the rapid convergence and solution of the straightening stroke with the goal of residual deflection approaching zero. The straightening correction coefficient based on the machine learning GBDT algorithm is introduced to data fit and correct the overall process comprehensive error in the real straightening scene. The method can significantly improve the precision and efficiency of the straight line guide rail straightening, reduce the over-straightening or under-straightening phenomenon, and is suitable for the automatic straightening production of high-precision straight line guide rails.

Description

Technical Field

[0001] This invention relates to the fields of precision machinery manufacturing and automation technology, specifically to a method and system for stroke prediction and control in a high-precision linear guide straightening process. Background Technology

[0002] Precision linear guides are core functional components in high-end equipment manufacturing industries such as high-end CNC machine tools, semiconductor manufacturing equipment, and aerospace precision machining equipment. Their straightness accuracy directly determines the motion positioning accuracy, operational stability, and service life of the final equipment. The three-point bending straightening process, as a core step in the finished linear guide manufacturing process to correct bending deformation and ensure the straightness accuracy of the finished product, directly determines the product quality and large-scale production capacity of the linear guides through its process accuracy and execution efficiency. It is a key technical link in the entire precision linear guide manufacturing process.

[0003] The most widely used mechanical pressure straightening process in current industrial production is generally based on the three-point bending straightening principle. However, there are still insurmountable pain points in the industry: First, it is highly dependent on the experience of operators. The straightening process usually adopts a cyclical mode of "measurement-trial straightening-remeasurement". There is no precise quantitative theoretical basis for the straightening force and stroke, making it impossible to achieve quantitative and automated straightening control. Second, it is difficult to guarantee the straightening accuracy and batch consistency. The guide rail material has significant elastoplastic deformation characteristics during the straightening process. After unloading, it will produce a complex springback effect. Experienced operation is very likely to lead to "under-straightening" or "over-straightening", requiring repeated adjustments and resulting in low production efficiency. Third, it cannot meet the micron-level precision requirements of high-end equipment. Traditional methods have large prediction deviations in the straightening stroke, poor accuracy stability, and difficulty in guaranteeing product qualification rate, which has become the core bottleneck restricting the large-scale production of high-precision linear guides.

[0004] In the three-point bending straightening process of linear guideways, the constitutive relationship between the bending moment and curvature of the cross-section is the core mechanical foundation for characterizing the elastoplastic deformation behavior throughout the straightening process and constructing a mapping model for straightening process parameters. The essence of three-point bending straightening is to apply a controllable straightening force to induce a preset elastoplastic bending deformation in the guideway. After unloading, the initial bending deformation is offset by elastic rebound, ultimately achieving the required straightness of the guideway. The correspondence between the bending moment and curvature directly determines the amount of deformation, the distribution of the plastic deformation zone, and the amount of rebound after unloading under a given straightening force. It is the sole theoretical basis for predicting the straightening stroke and quantitatively determining process parameters. Existing research often uses pure elastic theory or simplified ideal elastoplastic models to approximate the bending moment-curvature relationship, neglecting the nonlinear strengthening characteristics and stress redistribution laws of the cross-section after the material enters the plastic deformation stage. This fails to accurately describe the true mechanical response of the guideway under large plastic deformation straightening, resulting in inherent biases in straightening stroke prediction from a theoretical perspective. This is the core reason why current processes cannot escape reliance on experience and are difficult to improve straightening accuracy.

[0005] To address the aforementioned theoretical root causes, existing technologies have yet to provide an effective industrial solution. On one hand, the nonlinear moment-curvature constitutive relationship between straightening force and straightening stroke is a strongly nonlinear implicit functional relationship. Conventional numerical methods suffer from slow convergence, poor stability, and susceptibility to oscillations, failing to meet the demands of real-time calculation of straightening parameters for continuous industrial production. On the other hand, even with a precise theoretical constitutive model, it's impossible to avoid the comprehensive errors across the entire process under real-world conditions, such as the simplified assumption of rectangular cross-sections, material property fluctuations, and straightening machine system deformation. Traditional springback correction methods can only adjust for single springback errors, failing to achieve systematic fitting and correction of errors across the entire process, leading to a severe disconnect between theoretical calculations and practical applications. Therefore, developing a straightening stroke calculation method that combines theoretical rigor, computational efficiency, and adaptability to various working conditions has become a pressing technical challenge for the industry.

[0006] The purpose of this invention is to propose a novel method for accurately and efficiently predicting straightening process parameters. This method must be able to deeply characterize the elastoplastic mechanical behavior of materials, accurately correct the comprehensive errors of the entire process between theoretical calculations and actual working conditions, and achieve rapid calculation through efficient algorithms. Ultimately, it realizes the intelligent upgrade of linear guide straightening processes from "experience-dependent" to "model-driven." Driven by a nonlinear moment-curvature constitutive relationship, a two-layer nested iterative numerical solution model, and a GBDT machine learning error correction model, this invention achieves rapid and stable solutions for straightening process parameters and promotes the application of theoretical models, thereby meeting the growing quality and efficiency demands of high-end manufacturing for precision linear guides. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for predicting and controlling the straightening stroke of linear guideways. This method is based on elastoplastic mechanics theory and combines numerical iterative algorithms and machine learning techniques to solve the springback calculation errors caused by stress-strain nonlinearity after the material enters the plastic deformation stage. It also corrects comprehensive error problems under real working conditions, such as the simplified assumption of rectangular cross-sections, material property fluctuations, and equipment deformation. This invention can predict the optimal straightening parameters with high accuracy and efficiency, realizing the intelligent and automated straightening process.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for predicting and controlling the straightening stroke of a linear guide rail includes the following steps:

[0010] S1. Obtain the basic parameters of the linear guide, including geometric parameters, material mechanical parameters, and initial form and position parameters;

[0011] S2. Based on the elastic-plastic constitutive theory of materials, establish the nonlinear constitutive relationship between bending moment and curvature of the guide rail section; based on the three-point bending straightening principle, establish the mapping relationship between the straightening force and the bending moment distribution along the guide rail.

[0012] S3. Construct a plastic correction model for calculating springback, predict the straightening correction coefficient under the current working condition through a pre-trained machine learning model, and calculate the springback based on the straightening correction coefficient.

[0013] S4. With the residual deflection of the straightened guide rail approaching zero as the convergence objective, execute a nested iterative algorithm to solve for the optimal straightening force and the corresponding optimal straightening stroke.

[0014] S5. Output the obtained optimal straightening force and optimal straightening stroke to the straightening machine control system to control the straightening actuator to complete the straightening operation.

[0015] Furthermore, in S1, the geometric parameters include the straightening span. Cross-sectional width Cross-sectional height Material mechanical parameters include elastic modulus. Yield strength Strengthening coefficient Initial form and position parameters include initial deflection. ;

[0016] Intermediate constants, including the moment of inertia of the cross section, are calculated based on fundamental parameters. Elastic limit bending moment Elastic limit curvature The initial estimate of the straightening force was calculated based on the theory of small elastic deformation. .

[0017] Furthermore, in S2, the bending moment-curvature nonlinear constitutive relation established based on the linear strengthening elastoplastic model of the material is as follows:

[0018]

[0019] in, For bending moment, The curvature of the cross section;

[0020] The straightening force is established based on the three-point bending straightening principle. The mapping relationship with the friction moment distribution is as follows: ,in To calculate the axial distance between the cross section and the straightening support point.

[0021] Furthermore, the nested iterative algorithm in S4 is a two-level nested iterative algorithm, specifically including:

[0022] Outer layer iteration: for straightening force Perform iterative optimization to optimize the residual function. The absolute value is less than the preset convergence tolerance ,in For straightening stroke, This refers to the rebound amount;

[0023] Inner iteration: For any straightening force given by the outer iteration Perform forward mechanics calculations and solve for the curvature distribution along the guide rail using numerical methods. and to The straightening stroke corresponding to the current straightening force is obtained by performing a double integral calculation. .

[0024] Furthermore, in the inner iteration, the Newton-Raphson method is used to process discrete location points. Solve the nonlinear equations:

[0025]

[0026] The curvature values ​​at each discrete point are obtained by solving the problem. .

[0027] Furthermore, in S3, the plastic correction model for calculating the springback amount is constructed as follows:

[0028]

[0029] in, The straightening correction coefficient is predicted by a pre-trained machine learning model.

[0030] Furthermore, the machine learning model is constructed using the Gradient Boosting Regression Tree (GBDT) algorithm; the model's input features include span... Loading factor Relative deflection stiffness ratio The training labels for the model are the straightening correction coefficients calibrated in the experiment. The straightening correction coefficient The model is used to fit the comprehensive error of the entire straightening process, including the error in calculating the rebound amount, the error in the simplified assumption of the rectangular cross section, the error in the material property fluctuation, and the deformation error of the straightening machine system. The hyperparameters of the gradient boosting regression tree model are set as follows: the base learner is a regression tree, the maximum depth is 5, the minimum number of split samples is 11, the learning rate is 0.04, the number of base learners is 500, the subsampling ratio is 0.8, the validation set ratio is 0.2, the number of early stopping rounds is 20, and the fixed random seed is 100.

[0031] The machine learning model uses a gradient boosting regression tree (GBDT) model to construct a straightening correction coefficient predictor, and the prediction process is as follows:

[0032] 1) First, construct the input feature vector based on the guide rail's geometric parameters, material parameters, and initial deflection. This characterizes the matching relationship between the bending stiffness of the guide rail section and the straightening span;

[0033] Among them, span The axial distance between the two support points of the straightening machine; load factor Relative deflection characterizes the load level corresponding to the initial deflection per unit span; The dimensionless ratio representing the initial deflection to the height of the guide rail section; stiffness ratio ;

[0034] 2) Input the input feature vector into the trained GBDT model to obtain the predicted straightening correction coefficients. The GBDT model is composed of It is composed of a stack of regression tree-based learners, and its prediction expression is:

[0035]

[0036] in, For the initial regression model, For the first The output of a regression tree-based learner The learning rate;

[0037] 3) Straightening correction coefficient This is used to simultaneously compensate for errors in springback calculation, simplified rectangular cross-sections, material property fluctuations, and deformation errors in the straightening equipment system, and then substitute these errors into the springback correction model. ,in, Theoretical rebound amount, The actual rebound amount is corrected; the output value is physically limited, with a limit range of 0~50, and the straightening correction coefficient under the current working condition is finally obtained. .

[0038] Furthermore, in the outer iteration, the straightening force is updated using a quasi-Newton method, and the update formula is:

[0039]

[0040] Wherein, derivative The step size of the finite difference method is used for approximate calculation. An adaptive strategy is used to dynamically adjust the step size based on the current residual. Tolerance with Target ratio Dynamic reduction; if the residuals of two consecutive iterations have opposite signs, it is determined to be iterative oscillation, the step size coefficient is reduced and the average of the first two straightening forces is taken as the new iteration point.

[0041] Furthermore, the convergence target is the residual deflection after straightening. ,in The preset convergence tolerance has a range of values. .

[0042] The present invention also provides a linear guide straightening control system, comprising:

[0043] The parameter acquisition module is used to perform the parameter acquisition operation in step S1;

[0044] The mechanical modeling module is used to perform the constitutive relation and bending moment distribution modeling operations in step S2;

[0045] The error correction module is used to perform the straightening correction coefficient prediction and error correction calculation operations in step S3.

[0046] The iterative solution module is used to perform the nested iteration and optimal parameter solution operation in step S4;

[0047] The control output module is used to perform the parameter output and straightening control operations in step S5.

[0048] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0049] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0050] The design concept of this invention is as follows:

[0051] To address the problems of existing straightening processes being reliant on experience, having low precision, and poor efficiency, this invention establishes a nonlinear moment-curvature constitutive relationship for linear guides under three-point bending straightening, constructing an implicit function model between straightening force and straightening stroke. A two-layer nested iterative algorithm is employed: the outer layer iterates to optimize the straightening force, while the inner layer iterates based on the Newton-Raphson method to solve for the curvature distribution, aiming to achieve rapid convergence of the straightening stroke with the residual deflection approaching zero. A straightening correction coefficient based on the machine learning-based GBDT algorithm is introduced. This invention uses data fitting to correct the overall error of the entire process in a real straightening scenario. It can significantly improve the accuracy and efficiency of linear guide straightening, reduce over-correction or under-correction, and is suitable for automated straightening production of high-precision linear guides.

[0052] By employing the above-described technology, the present invention has the following beneficial effects compared with the prior art:

[0053] 1) This invention establishes a nonlinear moment-curvature constitutive relationship considering the elastoplastic properties of materials, breaking through the limitations of traditional elasticity theory and enabling a more accurate description of the mechanical behavior during the straightening process. Through a two-layer nested iterative algorithm, the outer layer uses a quasi-Newton method for rapid optimization, while the inner layer uses the Newton-Raphson method for efficient solution of the nonlinear equations. Combined with an adaptive step-size control strategy, this significantly improves computational efficiency while maintaining accuracy, precisely solving the complex implicit relationship between straightening force and straightening stroke.

[0054] 2) This invention uses the Gradient Boosting Regression Tree (GBDT) algorithm to construct a straightening correction coefficient prediction model, which breaks through the limitation of traditional models that only correct the springback amount. It can simultaneously fit the comprehensive errors of the entire process, such as the springback amount calculation error, the simplified assumption error of the rectangular cross section, the material property fluctuation error, and the deformation error of the straightening machine system. This significantly improves the accuracy of the straightening stroke prediction and controls the residual deflection error after straightening to within 0.1mm, achieving a comprehensive surpassing of the traditional expert database method in terms of accuracy and adaptability.

[0055] 3) The predictive model of this invention can adapt to different guide rail specifications and process parameters, reducing reliance on manual experience. It can be integrated into the straightening equipment control system to achieve fully automated processing from parameter input to straightening force output, significantly improving production efficiency and product consistency, and is suitable for industrial mass production. Compared with current inventions that focus on the design of straightening equipment, this invention focuses more on exploring the straightening mechanism of linear guide rails and calculating the key parameters required for the straightening process to achieve precise straightening of linear guide rails. Attached Figure Description

[0056] Figure 1 This is a diagram showing the structure of a linear guide straightening experimental platform.

[0057] Figure 2 This is a schematic diagram of three-point bending straightening of a linear guide rail;

[0058] Figure 3 Flowchart of the linear guide straightening process;

[0059] Figure 4 This is a flowchart illustrating the overall process of the method of the present invention.

[0060] Figure 5 This is a curve comparing the changes in deflection before and after straightening. Detailed Implementation

[0061] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0062] This invention provides a method for predicting and controlling the straightening stroke of a linear guide, comprising the following steps:

[0063] S1. Obtain the basic parameters of the linear guide, including geometric parameters, material parameters, and initial state. The geometric parameters include the straightening span. Cross-sectional width Cross-sectional height The material parameters include elastic modulus. Yield strength Strengthening coefficient The initial state is the initial deflection. ;

[0064] Calculate the moment of inertia of the cross section Elastic limit bending moment Elastic limit curvature Intermediate constants; based on the theory of small deformation elasticity, through formulas Calculate the initial estimate of the straightening force This serves as the starting point for iterative calculations.

[0065] S2. Construct a nonlinear elastoplastic mechanical model, and establish the section bending moment based on the linear hardening elastoplastic model of the material. With curvature Nonlinear constitutive relations:

[0066]

[0067] Among them, the elastic limit bending moment Elastic limit curvature ;

[0068] Based on the three-point bending straightening principle, establish the straightening force. With bending moment distribution Relationship: ,in This is the distance from the support point.

[0069] S3. Machine learning-assisted straightening correction coefficient prediction, constructing a system including span... Loading factor Relative deflection stiffness ratio The feature set, which comprehensively reflects the geometry, material and load state of the guide rail;

[0070] The feature set is input into a pre-trained Gradient Boosting Regression Tree (GBDT) model to predict the straightening correction coefficient under the current operating conditions. The model was trained based on 120 sets of experimental data, covering 5 cross-sectional sizes and 4 span conditions, ensuring prediction accuracy and generalization ability; the straightening correction coefficient The error is used to calculate the rebound amount, simplify the assumption error of the rectangular section, the material property fluctuation error, and the deformation error of the straightening machine system, which are all part of the overall error of the straightening process.

[0071] S4. Construct a two-layer nested iterative calculation to calculate the residual deflection. To achieve convergence, a two-level nested iterative algorithm is executed:

[0072] 4.1 Outer layer iteration: for straightening force Iterative optimization is performed, and the straightening force is updated using a quasi-Newton method:

[0073]

[0074]

[0075] in For the residual function, The finite difference method is used for approximate calculation, with a step size of... The ratio of the current residual to the tolerance is dynamically adjusted to achieve rapid convergence and avoid oscillations. With the goal The ratio is used to dynamically adjust the effective step size for each iteration:

[0076]

[0077]

[0078] in This represents a percentage of the initial step size.

[0079] To prevent iterative oscillations, measures are taken to suppress oscillations and promote convergence. The residuals of each iteration are recorded. If the residuals of two consecutive iterations have opposite signs, then oscillation has occurred, and the step size should be reduced by a percentage. Discard the increment calculated by the current Newton method. Take the average of the first two straightening forces as the new trial point to break out of the cycle of oscillation between the two points;

[0080] 4.2 Inner Layer Iteration: For a given straightening force Complete one forward mechanics calculation:

[0081] 4.2.1 Discretize the guide rail into For each point... Solve Nonlinear equations:

[0082]

[0083] 4.2.2 The Newton-Raphson method is used to iteratively solve the nonlinear equations to obtain the curvature distribution. .

[0084] 4.2.3 Regarding Perform double numerical integration, combined with boundary conditions. , , Calculate the straightening stroke .

[0085] 4.2.4 Calculate the springback amount using the straightening correction formula. :

[0086]

[0087] This formula corrects the classical elastic rebound formula by adjusting the straightening correction coefficient, and at the same time fits the combined error between theoretical calculation and actual straightening conditions, accurately reflecting the deformation and rebound behavior in the actual straightening process.

[0088] S5. Result Output and Straightening Control, when... When the iteration converges, the optimal straightening force is output. and straightening stroke .

[0089] Finally, the optimal parameters are sent to the straightening machine control system, which controls the actuator to complete the straightening operation, realizing fully automatic processing from parameter input to straightening stroke output.

[0090] Example: Figure 1The linear guide rail straightening experimental platform shown consists of a programmable three-point bending straightener, a laser displacement straightening stroke monitoring system, and a deflection measurement system. The straightener is equipped with a servo pressure system, allowing the maximum straightening pressure to be set via an external knob. The stroke control measures the downward pressure, inputting the straightening stroke into the PLC control system, which then converts it into an analog signal via an AD converter and transmits it to the servo pressure system for stroke straightening. The laser displacement straightening stroke monitoring system is installed on the straightener's pressure platform to monitor the downward pressure of the pressure head and coordinates with the straightener's servo pressure system for displacement control. The deflection measurement system consists of a marble testing platform and a dial indicator. The guide rail is fixed to the marble platform with mounting screws, and the dial indicator records the deflection changes before and after straightening.

[0091] In an embodiment, such as Figure 2 , 3 The diagram shows a schematic of three-point bending straightening of a linear guide and a flowchart of the straightening process. The linear guide employs a three-point straightening method, applying a straightening force at the center of the guide to create a three-point bending stress state. This brings the guide into a stress state above the elastic critical value, causing elastoplastic deformation and bringing the residual deformation towards a preset straightness, thus achieving the straightening purpose. First, the guide to be straightened is placed on a marble platform, and the linear guide is limited by mounting screws. A dial indicator is used to record the deflection change of the entire guide, and the straightening points are marked. Then, the guide is placed on the straightening platform of the straightening machine. An appropriate span is selected based on the guide's condition, and the required straightening stroke is calculated. After inputting the straightening stroke data, a laser displacement straightening stroke monitoring system is used to perform displacement straightening of the guide. Finally, after straightening, the guide's deflection is measured to determine if the residual deflection curve meets the straightness requirements; otherwise, the above steps are repeated.

[0092] Specifically, select a cross-sectional dimension of The linear guide rail was used as the straightening object and placed stably on a marble platform. A dial indicator was used to measure the initial axial deflection distribution of the guide rail, and the deflection data at each point was recorded as follows: ,in Number of tests j is the deflection position. According to Draw the deflection curve, select the peak or trough position and mark it as the part that needs to be straightened.

[0093] Adjust the distance between the two support points of the straightening machine appropriately, and set the relevant parameters as follows: , , , , , , Substitute them into the calculation.

[0094] Calculate the moment of inertia of the cross section Elastic limit bending moment Elastic limit curvature Constants. The initial straightening force is estimated based on the elastic solution. .

[0095] In this embodiment, the machine learning model is constructed using the Gradient Boosting Regression Tree (GBDT) algorithm, and the specific implementation process is as follows:

[0096] 1. Dataset Construction

[0097] We collected linear guides with five mainstream cross-sectional specifications (cross-sectional dimensions: 15mm×15mm, 20mm×17.5mm, 23mm×22mm, 28mm×26mm, 34mm×29mm) and four straightening span conditions (200mm, 300mm, 400mm, 500mm) from industrial production, using S55C steel as the commonly used guide material, for a total of 120 sets of orthogonal experimental samples.

[0098] Each sample group contains input features and label values: the input features are span L and loading factor. Relative deflection stiffness ratio The label value is the straightening correction coefficient obtained through experimental calibration. The calibration method is as follows: the actual rebound amount under the corresponding working condition is measured through single-factor experiments. Combining the classic elastic rebound formula Inverse calculation yields In some scenarios, calibration coefficients are determined by manual testing. .

[0099] The dataset is divided into a training set and a test set in an 8:2 ratio. The training set is used for model fitting, and the test set is used to verify the model's generalization ability.

[0100] 2. Model hyperparameter settings

[0101] In this embodiment, the core hyperparameters of the GBDT model are set as follows: the base learner is a regression tree, the maximum depth is 5, the minimum number of split samples is 11; the learning rate is set to 0.04; the number of base learners is 500; the subsampling ratio is 0.8; the number of early stopping rounds is 20; the loss function is the mean squared error (MSE); and the random seed is fixed at 100 to ensure that the model training results are reproducible.

[0102] 3. Model Training and Validation

[0103] A gradient boosting algorithm is used to fit the training set, and hyperparameters are optimized through 5-fold cross-validation. An early stopping mechanism is implemented during training; training stops when the validation set loss shows no decrease for 20 consecutive iterations to prevent overfitting. After training, the model achieves a coefficient of determination R² ≥ 0.94, a mean absolute error (MAE) ≤ 2.2, and a mean squared error (MSE) ≤ 7.94 on the test set, meeting industrial prediction accuracy requirements. Furthermore, feature importance analysis and SHAP value analysis are used to quantify the importance of each input feature. The contribution of the prediction results contributes to the interpretability of the model.

[0104] 4. Model Deployment and Prediction

[0105] The trained model and feature columns are saved sequentially as a callable file. During prediction, the geometric, material, and working condition parameters of the target guide rail are input, and the system automatically calculates the input features and outputs straightening correction coefficients. The output value is limited to a range of 3 to 50 to avoid prediction results outside the physical range and ensure the stability of industrial applications.

[0106] Iterative solution, enter the outer loop, set the current force. .

[0107] Inner loop: Discretize the guide rails into For each point... ,according to The curvature distribution was obtained by iteratively solving using the Newton-Raphson method. .

[0108] right Perform numerical integration according to the formula:

[0109]

[0110] Calculate the straightening stroke .

[0111] According to the formula Calculate rebound amount .

[0112] According to the formula Calculate residuals .

[0113] like The iteration ends when the optimal straightening force is output. and straightening stroke Otherwise, according to the formula:

[0114]

[0115]

[0116]

[0117]

[0118] The new straightening force estimate is calculated using the adaptive finite difference method. Return to step 3 and continue iterating.

[0119] The optimal straightening force was obtained through iterative calculation. Predicting the straightening stroke The calculation results are sent to the PLC control system of the straightening machine, which controls the pressure head to press down, completing the high-precision straightening operation. The experimental guide rail sample is straightened, and its straightening position and optimal straightening force are determined. and predicting the straightening stroke As shown in Table 1, the calculation results are sent sequentially to the PLC control system of the straightening machine, which controls the pressure head to press down to the corresponding preset stroke, thus completing the entire straightening operation.

[0120] ;

[0121] Ultimately, the residual deflection of the guide rail was controlled to within 0.1 mm. The comparison curve of deflection changes before and after guide rail straightening is shown below. Figure 5 As shown, the error is much smaller than that of the traditional method, proving the effectiveness of the present invention. The specific performance differences between the present invention and the traditional method are shown in Table 2. Compared with the traditional method, the straightening accuracy of the present invention is improved by 5 times and the production efficiency is improved by about 70%; compared with the ideal elastic-plastic model method, the straightening accuracy is improved by more than 2 times and the batch consistency is improved by more than 1 time, which completely solves the industry pain points of traditional straightening process, such as reliance on manual experience, low accuracy, poor efficiency and poor batch consistency.

[0122] ;

[0123] In summary, this invention combines linear guide straightening theory with GBDT prediction optimization, which ensures that the calculation results of linear guide straightening conform to physical constraints. The GBDT model is used to predict the straightening correction coefficient. This coefficient is used to correct the formula for calculating the rebound amount, and simultaneously fits the comprehensive errors under real-world scenarios such as the rectangular cross-section assumption, material fluctuations, and equipment deformation, thereby improving the accuracy of the straightening stroke prediction. This invention fully utilizes the learning capabilities of GBDT (Guided Basic Data Theory) machine learning; the optimized straightening stroke calculation results can compensate for errors generated during the calculation process based on experimental data, making the prediction results more accurate. In practical applications, this not only saves time and effort but also reduces the possibility of human error.

Claims

1. A method for predicting and controlling the straightening stroke of a linear guide, characterized in that, Includes the following steps: S1. Obtain the basic parameters of the linear guide, including geometric parameters, material mechanical parameters, and initial form and position parameters; S2. Based on the elastic-plastic constitutive theory of materials, establish the nonlinear constitutive relationship between bending moment and curvature of the guide rail section; based on the three-point bending straightening principle, establish the mapping relationship between the straightening force and the bending moment distribution along the guide rail. S3. Construct a plastic correction model for calculating springback, predict the straightening correction coefficient under the current working condition through a pre-trained machine learning model, and calculate the springback based on the straightening correction coefficient. S4. With the residual deflection of the straightened guide rail approaching zero as the convergence objective, execute a nested iterative algorithm to solve for the optimal straightening force and the corresponding optimal straightening stroke. S5. Output the obtained optimal straightening force and optimal straightening stroke to the straightening machine control system to control the straightening actuator to complete the straightening operation.

2. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 1, characterized in that, In S1, the geometric parameters include the straightening span. Cross-sectional width Cross-sectional height Material mechanical parameters include elastic modulus. Yield strength Strengthening coefficient Initial form and position parameters include initial deflection. ; Intermediate constants, including the moment of inertia of the cross section, are calculated based on fundamental parameters. Elastic limit bending moment Elastic limit curvature The initial estimate of the straightening force was calculated based on the theory of small elastic deformation. .

3. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 2, characterized in that, In S2, the bending moment-curvature nonlinear constitutive relation established based on the linear strengthening elastoplastic model of the material is as follows: ； in, For bending moment, The curvature of the cross section; The straightening force is established based on the three-point bending straightening principle. The mapping relationship with the friction moment distribution is as follows: ,in To calculate the axial distance between the cross section and the straightening support point.

4. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 2, characterized in that, The nested iteration algorithm in S4 is a two-level nested iteration algorithm, specifically including: Outer layer iteration: for straightening force Perform iterative optimization to optimize the residual function. The absolute value is less than the preset convergence tolerance ,in For straightening stroke, This refers to the rebound amount; Inner iteration: For any straightening force given by the outer iteration Perform forward mechanics calculations and solve for the curvature distribution along the guide rail using numerical methods. and to The straightening stroke corresponding to the current straightening force is obtained by performing a double integral calculation. .

5. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 4, characterized in that, In the inner iteration, the Newton-Raphson method is used to process discrete location points. Solve the nonlinear equations: ； The curvature values ​​at each discrete point are obtained by solving. .

6. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 4, characterized in that, In S3, the plastic correction model for calculating the springback is constructed as follows: ； in, The straightening correction coefficient is predicted by a pre-trained machine learning model.

7. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 6, characterized in that, The machine learning model uses a gradient boosting regression tree (GBDT) model, and the input features of the model include span. Loading factor Relative deflection stiffness ratio ; The output is the straightening correction coefficient under the current operating conditions. .

8. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 4, characterized in that, In the outer iteration, the straightening force is updated using a quasi-Newton method, and the update formula is: ； Wherein, derivative The step size of the finite difference method is used for approximate calculation. An adaptive strategy is used to dynamically adjust the step size based on the current residual. Tolerance with Target ratio Dynamic reduction; if the residuals of two consecutive iterations have opposite signs, it is determined to be iterative oscillation, the step size coefficient is reduced and the average of the first two straightening forces is taken as the new iteration point.

9. The method for predicting and controlling the straightening stroke of a linear guide rail according to claim 1, characterized in that, The convergence target is the residual deflection after straightening. ,in The preset convergence tolerance has a range of values. , For straightening stroke, This refers to the rebound amount.

10. A linear guide straightening control system, characterized in that, It includes a parameter acquisition module, a mechanical modeling module, an error correction module, an iterative solution module, and a control output module, among which: The parameter acquisition module is used to perform the parameter acquisition operation in step S1 of claim 1; The mechanical modeling module is used to perform the constitutive relation and bending moment distribution modeling operation in step S2 of claim 1; The error correction module is used to perform the straightening correction coefficient prediction and error correction calculation operation in step S3 of claim 1; The iterative solution module is used to perform the nested iteration and optimal parameter solution operation of step S4 as described in claim 1; The control output module is used to perform the parameter output and straightening control operation of step S5 as described in claim 1.

Images