Method and device for correcting first bend position deviation of pipe bending machine

Abstract

The present application provides a method and device for correcting the first bend position deviation of a pipe bending machine, which relates to the technical field of pipe bending machines, including: setting bending control parameters according to bending parameters and pipe fitting data; generating springback prediction results; performing springback compensation optimization, updating the bending control parameters with the springback compensation optimization results, and establishing a time series mapping fitting result; configuring monitoring key points, configuring a camera array with the monitoring key points, and establishing a time series image data set; establishing error recognition results; and correcting the first bend position deviation according to the error recognition results and the updated bending control parameters. The present application can solve the technical problem in the prior art that the recognition accuracy of the first bend position deviation is insufficient, which leads to poor correction effect of the first bend deviation, so as to achieve the technical effect of correcting the deviation of the first bend position, ensuring the accuracy of the bending operation, and thus improving the quality and consistency of the product.

Description

Technical Field

[0001] The present application relates to the technical field of pipe bending machines, and in particular to a method and device for correcting first bend position deviation of a pipe bending machine. Background Art

[0002] Pipe bending machines are widely used in the fields of automobile manufacturing, aerospace, petrochemicals, etc. Their main function is to bend straight pipes into the required shape through mechanical force to meet specific engineering requirements. However, in the actual pipe bending process, due to the influence of factors such as material properties, bending angle, bending radius, etc., there will often be problems such as bending angle deviation, uneven deformation, and springback, resulting in the final product quality not meeting the design requirements. Especially in the first pipe bending operation, the accuracy of the first bend position is particularly important because it directly affects the quality of subsequent pipe bending and the accuracy of the overall structure.

[0003] With the development of intelligent manufacturing technology, how to effectively predict and correct the rebound phenomenon in pipe bending operations, and how to improve the accuracy of the first bend position, have become technical problems that need to be solved urgently. At present, some pipe bending machines have begun to introduce machine learning models to assist pipe bending operations in order to improve pipe bending accuracy and production efficiency. However, most of these methods only optimize a single link, lack systematic consideration of the entire pipe bending process, lack accuracy in error identification and compensation, and are difficult to meet the needs of high-precision pipe bending operations.

[0004] In summary, the prior art has a technical problem in that the recognition accuracy of the first turn position deviation is insufficient, which leads to poor correction effect of the first turn deviation. Summary of the invention

[0005] The purpose of the present application is to provide a method and device for correcting the first bend position deviation of a pipe bending machine, so as to solve the technical problem in the prior art that the recognition accuracy of the first bend position deviation is insufficient, resulting in poor correction effect of the first bend deviation.

[0006] In view of the above problems, the present application provides a method and device for correcting the first bend position deviation of a pipe bending machine.

[0007] In the first aspect, the present application provides a method for correcting the first bend position deviation of a pipe bending machine, and the method for correcting the first bend position deviation of a pipe bending machine is implemented by a first bend position deviation correction device for a pipe bending machine, wherein the method for correcting the first bend position deviation of a pipe bending machine comprises: after the initialization of the pipe bending machine system is completed, the pipe fitting data is synchronized to the pipe bending machine system, and the three-dimensional model data is called through the pipe bending machine system, and the bending control parameters are set according to the bending parameters and pipe fitting data in the three-dimensional model data; a historical bending database is established, and a rebound prediction network is configured with the historical bending database, and the rebound prediction network is constructed by performing supervised learning by dividing the historical bending database into a training set and a test set. The rebound prediction network is a machine learning model network, and after feature extraction of the pipe fitting data, the feature extraction is performed. The result is input into the springback prediction network, and the springback prediction result is generated after being processed by the springback prediction network; the springback compensation is optimized according to the springback prediction result and the bending control parameters, the bending control parameters are updated with the springback compensation optimization result, and the timing mapping fitting result is established according to the updated bending control parameters; the monitoring key points are configured based on the bending parameters, the camera array is configured with the monitoring key points, and a time series image data set is established based on the camera array, and the time series image data set is constructed with the monitoring key points as the acquisition center; the time series image data set is dynamically spliced ​​in real time, and error identification is performed based on the dynamic real-time splicing result, the timing mapping fitting result, and the monitoring key points to establish an error identification result; the first bend position deviation is corrected according to the error identification result and the updated bending control parameters.

[0008] In the second aspect, the present application also provides a first bend position deviation correction device for a pipe bending machine, which is used to execute the first bend position deviation correction method for a pipe bending machine as described in the first aspect, wherein the first bend position deviation correction device for a pipe bending machine includes: a pipe bending control parameter setting module, which is used to synchronize the pipe fitting data to the pipe bending machine system after the pipe bending machine system is initialized, and call the three-dimensional model data through the pipe bending machine system, and set the pipe bending control parameters according to the bending parameters and pipe fitting data in the three-dimensional model data; a rebound prediction module, which is used to establish a historical bending database, and configure a rebound prediction network with the historical bending database. The rebound prediction network is constructed by dividing the historical bending database into a training set and a test set for supervised learning. The rebound prediction network is a machine learning model network. After feature extraction of the pipe fitting data, the feature extraction result is input into the rebound prediction network. , generating a springback prediction result after being processed by the springback prediction network; a springback compensation optimization module, used to perform springback compensation optimization according to the springback prediction result and the bending control parameter, update the bending control parameter with the springback compensation optimization result, and establish a time series mapping fitting result according to the updated bending control parameter; an image data acquisition module, used to configure the monitoring key points based on the bending parameters, configure the camera array with the monitoring key points, establish a time series image data set based on the camera array, and the time series image data set is constructed with the monitoring key points as the acquisition center; an error recognition module, used to perform real-time dynamic splicing of the time series image data set, and perform error recognition based on the dynamic real-time splicing result, the time series mapping fitting result, and the monitoring key points to establish an error recognition result; a deviation correction module, used to perform first bend position deviation correction according to the error recognition result and the updated bending control parameter.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] After the initialization of the pipe bending machine system is completed, the pipe fitting data is synchronized to the pipe bending machine system, and the three-dimensional model data is called through the pipe bending machine system, and the bending control parameters are set according to the bending parameters in the three-dimensional model data and the pipe fitting data; a historical bending database is established, and a springback prediction network is configured with the historical bending database. The springback prediction network is constructed by supervised learning by dividing the historical bending database into a training set and a test set. The springback prediction network is a machine learning model network. After feature extraction of the pipe fitting data, the feature extraction result is input into the springback prediction network, and the springback prediction result is generated after processing by the springback prediction network; according to the springback prediction result The springback compensation optimization is performed based on the results and the bending control parameters, the bending control parameters are updated with the springback compensation optimization results, and the time series mapping fitting results are established according to the updated bending control parameters; the monitoring key points are configured based on the bending parameters, the camera array is configured with the monitoring key points, and a time series image data set is established based on the camera array, and the time series image data set is constructed with the monitoring key points as the collection center; the time series image data set is spliced ​​in real time, and error identification is performed based on the dynamic real-time splicing results, the time series mapping fitting results, and the monitoring key points to establish error identification results; the first bend position deviation is corrected according to the error identification results and the updated bending control parameters. Springback prediction is performed by combining the three-dimensional model data and the historical bending database through the springback prediction network, and then springback compensation optimization is performed according to the springback prediction results, and a time series mapping fitting result is established. The monitoring key points are further configured for monitoring image splicing, and error identification is performed in combination with the time series mapping fitting results, so that the first bend position deviation in the bending process is corrected in real time, so as to correct the deviation of the first bend position, ensure the accuracy of the bending operation, and thus improve the quality and consistency of the product.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0013] Figure 1 This is a flow chart of a method for correcting the first bend position deviation of a pipe bending machine used in this application;

[0014] Figure 2 This is a schematic diagram of the structure of the first bend position deviation correction device used for a pipe bending machine in this application.

[0015] Description of reference numerals:

[0016] The pipe bending control parameter setting module 11, the springback prediction module 12, the springback compensation optimization module 13, the image data acquisition module 14, the error identification module 15, and the deviation correction module 16. DETAILED DESCRIPTION

[0017] The present application solves the technical problem in the prior art that the accuracy of the recognition of the first bend position deviation is insufficient, which leads to poor correction effect of the first bend deviation, by providing a method and device for correcting the first bend position deviation of a pipe bending machine. The springback prediction network is combined with the three-dimensional model data and the historical bending database to perform springback prediction, and then the springback compensation optimization is performed according to the springback prediction result, and the time series mapping fitting result is established. The monitoring key points are further configured to perform monitoring image splicing, and the error identification is performed in combination with the time series mapping fitting result, so as to perform real-time correction of the first bend position deviation in the pipe bending process, so as to correct the deviation of the first bend position, ensure the accuracy of the pipe bending operation, and thus improve the quality and consistency of the product.

[0018] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0019] For example, please refer to the attached Figure 1 The present application provides a method for correcting the first bend position deviation of a pipe bending machine, and the method for correcting the first bend position deviation of a pipe bending machine specifically comprises the following steps:

[0020] Step 1: After the pipe bending machine system is initialized, the pipe fitting data is synchronized to the pipe bending machine system, the three-dimensional model data is called through the pipe bending machine system, and the bending control parameters are set according to the bending parameters in the three-dimensional model data and the pipe fitting data.

[0021] Specifically, the pipe bending machine system refers to the mechanical equipment and its control system used to bend straight pipes or other shapes of pipes, which usually include mechanical parts and electronic control parts. The mechanical parts include hydraulic cylinders, molds, clamps and other actuators, which are responsible for bending the pipes into specified shapes; the electronic control part controls the movement of mechanical parts through computers or microcontrollers to ensure that the pipe bending process is carried out according to the predetermined procedures and parameters. The pipe bending machine system can automatically calculate and execute the pipe bending operation according to the input bending parameters and pipe data to ensure the production of bent pipes that meet the design requirements.

[0022] The initialization of the pipe bending machine system refers to a series of preparation and configuration work before the pipe bending machine starts to work normally. This process includes self-checking the system hardware and software, testing the various functions of the pipe bending machine, loading the necessary operating system or control software, configuring the initial parameters, and ensuring that all actuators and sensors can work properly. Through system initialization, it can ensure that the pipe bending machine is in the best working state, and all operating parameters and control logic have been correctly configured, thus laying the foundation for subsequent pipe bending operations. After the initialization is completed, the pipe data can be accurately read and the corresponding pipe bending tasks can be performed.

[0023] After the pipe bending machine system is initialized, the pipe fitting data is synchronized to the pipe bending machine system. Pipe fitting data usually includes parameters such as pipe diameter, wall thickness, material properties, length, and bending radius, which are critical input information in the pipe bending process.

[0024] Next, the tube bending machine system will automatically call the pre-stored 3D model data. The 3D model data is usually a virtual model established based on the geometric shape and structural characteristics of the pipe fitting, which can accurately reflect the actual shape of the pipe fitting in space. These data can not only provide visualization support, but also provide the necessary geometric parameters and reference information for the tube bending machine. For example, if the pipe fitting is a complex-shaped pipe, these model data will include all the bending points, connection angles and lengths of the pipe.

[0025] After calling the 3D model data, the bending control parameters will be set based on the bending parameters in the 3D model data and the previously synchronized pipe data. Bending parameters include bending angle, bending radius, bending speed, etc.; bending control parameters are operating instructions automatically calculated and set by the bending machine system according to the bending parameters, which control how the actuators of the bending machine actually operate. For example, the bending machine system will set the pressure and displacement speed of the hydraulic cylinder according to the bending angle and bending radius as bending control parameters to ensure that excessive stress concentration will not occur during the bending process, and to avoid rupture or uneven deformation of the pipe. Therefore, by setting the bending control parameters, a basis is provided for the subsequent correction of the first bend position deviation.

[0026] Step 2: Establish a historical bending database, and configure a springback prediction network with the historical bending database. The springback prediction network is constructed by performing supervised learning by dividing the historical bending database into a training set and a test set. The springback prediction network is a machine learning model network. After feature extraction of the pipe data, the feature extraction result is input into the springback prediction network, and the springback prediction result is generated after processing by the springback prediction network.

[0027] Specifically, configuring a springback prediction network is to use the existing historical bending data, through machine learning or neural network technology, to establish a model that can predict the springback phenomenon in the pipe bending process. This process can be divided into the following steps: First, a historical bending database needs to be established. The historical bending database contains records of all pipe bending operations that have been performed, including the basic properties of pipe fittings (such as material type, diameter, wall thickness, etc.), bending process parameters (such as bending angle, bending radius, bending speed, etc.), mechanical data recorded during the pipe bending process, and the actual springback amount after bending. Next, select a suitable machine learning model to build a springback prediction network. Common choices include regression models, neural network models (such as multi-layer perceptrons, convolutional neural networks, etc.) and other deep learning-based methods. For example, the basic unit of a neural network model is a neuron, which is connected layer by layer to form a complex network, including an input layer, a hidden layer, and an output layer. During the model training process, the historical bending data will be divided into a training set and a test set. The training set is used for model learning, that is, the network adjusts the internal weight parameters to make the prediction result as close to the actual springback as possible. The test set is used to verify the effect of the model to ensure that the model can accurately predict the springback amount on unseen data. After training, the springback prediction network is strictly verified and optimized, such as using the test set to verify the prediction accuracy of the model and analyzing the overfitting or underfitting of the model. If the prediction accuracy is less than the preset accuracy requirement, the model structure can be adjusted to improve the performance of the model. The verified and optimized springback prediction network can be integrated into the pipe bending machine system.

[0028] For example, there are 1,000 historical bending data in the historical bending database, which describes the springback of stainless steel pipes under different bending conditions. The historical bending data is divided into two parts: the training set is used to train the springback prediction network, which generally accounts for 70%-80% of the total data; the test set is used to verify the prediction ability of the network, which generally accounts for 20%-30%. 800 of the 1,000 historical bending data are selected as the training set, and the remaining 200 are used as the test set. In order for the network to learn useful information from the data, key features need to be extracted from the original data. For example: the elastic modulus, bending angle, bending radius, bending speed of the material, these features as input will help the network understand the bending behavior under different conditions. For each data, these features are extracted and standardized (converting features of different orders of magnitude to the same scale). The extracted features are input into the machine learning network for supervised learning. Taking the recurrent neural network as an example, it can learn the relationship between the feature and the target springback angle. Each layer of the network adjusts the output result by calculating the weight and bias value, so that the output (predicted springback angle) gradually approaches the real historical springback value. During the training process, a loss function (such as mean square error) is used to measure the difference between the predicted value and the actual value. The network continuously adjusts parameters to reduce the training error. For example, the network may predict that the springback angle of a 90-degree bend is 3 degrees at the beginning, while the actual springback value is 2 degrees. By adjusting the network parameters, the network gradually learns and improves the prediction, and may eventually get a prediction result closer to the actual value. After the training is completed, the test set is input into the trained springback prediction network to see if it can accurately predict the springback angle on unseen data. At this time, the network no longer adjusts the parameters, but gives a prediction value based on the learning results in training. For example, a piece of data in the test set represents a pipe with a 90-degree bend. The springback angle in the historical record is 2.1 degrees, and the prediction result given by the network is 2.05 degrees. By comparing the predicted value with the actual value, the prediction accuracy of the network can be evaluated, so as to obtain a springback prediction network that meets the requirements.

[0029] Then, feature extraction is performed on the pipe data, that is, the most useful information for predicting springback is extracted from the original data, such as the material properties of the pipe (such as elastic modulus), bending angle, bending radius, etc. Through feature extraction, the input data can be simplified and the calculation efficiency and accuracy of the prediction network can be improved. The feature extraction results are input into the springback prediction network, and the springback prediction network is used for nonlinear calculation and analysis to predict the springback angle that may occur under these conditions (for example, the predicted springback angle is 2 degrees). This is used as the springback prediction result, which is convenient for the subsequent adjustment of the bending control parameters, so that the final bending result is more accurate, the bending accuracy is improved, and the error caused by the springback phenomenon is reduced.

[0030] Step three: performing springback compensation optimization according to the springback prediction result and the pipe bending control parameter, updating the pipe bending control parameter with the springback compensation optimization result, and establishing a time series mapping fitting result according to the updated pipe bending control parameter.

[0031] Specifically, springback compensation optimization refers to finding the best combination of bending control parameters through an optimization algorithm to offset the springback amount in the springback prediction results. Springback is the deformation recovery phenomenon caused by the elastic restoring force of the material after bending. If no compensation is performed, the shape of the pipe after bending will deviate from the design target. Therefore, it is necessary to adjust the bending control parameters such as bending angle, applied force, bending speed, etc. according to the springback prediction results to achieve springback compensation. Specifically, the springback prediction results are used as input parameters, and springback compensation optimization is performed through optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.). The optimization algorithm will find a parameter combination that can minimize the impact of springback among multiple possible combinations of bending control parameters. This process may require multiple iterations and calculations, and the final result is an updated set of bending control parameters.

[0032] For example, suppose in a pipe bending operation, the springback prediction result shows that the pipe will rebound 2 degrees after completing a 90-degree bend. It is necessary to compensate for this 2-degree springback by adjusting the control parameters of the pipe bending machine (such as increasing the bending angle to 92 degrees). The goal of springback compensation optimization is to find an optimal combination of pipe bending control parameters so that the pipe can accurately reach 90 degrees after the bending is completed.

[0033] After the optimization of springback compensation is completed, the bending control parameters are updated according to the optimized results. Next, the timing mapping fitting results are established according to the updated bending control parameters. Timing mapping refers to the relationship between the control parameters at different time points and the actual bending results during the bending operation. Specifically, the structure and historical bending data of the bending machine are used to establish a virtual simulation platform using existing digital simulation technology. The updated bending control parameters are input into the virtual simulation platform for simulation, and the actual bending results at different time nodes can be output. In this way, the timing mapping fitting results are established to monitor and evaluate the accuracy and stability of the bending process.

[0034] In summary, through springback compensation optimization and timing mapping fitting, the pipe bending machine system can dynamically adjust the pipe bending control parameters, so as to complete the pipe bending task more accurately, reduce the error caused by springback, and optimize the control effect of the entire pipe bending process.

[0035] Step 4: configuring monitoring key points based on the bending parameters, configuring a camera array based on the monitoring key points, and establishing a time series image data set based on the camera array, wherein the time series image data set is constructed with the monitoring key points as the collection center.

[0036] Specifically, the areas that need to be monitored during the pipe bending process are determined according to the bending parameters, that is, the key monitoring points. The key monitoring points refer to the parts that are most likely to deform or error during the pipe bending process, and these parts often have a decisive influence on the bending results. Bending parameters refer to various process parameters that affect the shape and quality of pipe fittings during pipe bending operations, including bending angle, bending radius, bending speed, etc. Each bending parameter will affect the deformation and final shape of the pipe fittings. For example, a smaller bending radius may cause greater stress inside the material, while too fast a bending speed may cause material breakage or uneven deformation. Based on the above bending parameters, analyze which areas may have greater stress concentration, deformation or other quality problems during the pipe bending process. The key monitoring points are usually the areas with the greatest stress, the most significant deformation or the most prone to quality defects during the bending process. The determination of these key points requires comprehensive consideration of the bending parameters, material properties, and previous experience and data. For example, when performing large-angle bending (such as 90-degree bending), the inner and outer sides of the bending point may have different degrees of stress concentration and deformation, so these points are usually selected as key monitoring points. Specifically, areas where deformation frequently occurs in different historical bending parameters can be collected, a database can be established, and areas corresponding to current bending parameters can be matched in the database as key monitoring points.

[0037] Next, in order to fully monitor the key monitoring points, a camera array is configured around the key monitoring points. The camera array is a monitoring system composed of multiple cameras arranged in a specific geometric layout, which can simultaneously collect image data of the key monitoring points from different angles and distances. This multi-angle monitoring method ensures that every subtle deformation in the pipe bending process can be accurately captured and recorded.

[0038] In actual operation, when the pipe bending machine starts working, the camera array will capture the image data of the pipe fittings in the bending process in real time. It should be noted that the images collected here are obtained by the pipe bending machine for proofing test, that is, they are not used for actual production, but only for bending tests on some samples, and then collecting images for error analysis. These image data will be organized into a time series image data set in chronological order. The time series image data set is composed of images taken by the camera array at different times, which can reflect the dynamic changes in the shape of the pipe fittings during the bending process. Each frame of the image records the shape of the pipe fittings at a specific time point, and by analyzing these time series images, the complete deformation process during the bending process can be reconstructed.

[0039] The time-series image data set is constructed with the monitoring key points as the acquisition center, that is, in order to ensure the accuracy of the time-series image data set, these image data are constructed with the monitoring key points as the acquisition center. In other words, the focus of the camera array is always kept on the monitoring key points to ensure that every deformation of the monitoring key points is accurately captured. For example, when bending a pipe with a complex shape, if a certain bending point is determined as a monitoring key point, the camera array will focus on monitoring the deformation process of this point. As the bending process proceeds, the camera array will continue to collect images of this point and generate a complete deformation data set based on the collected time-series images.

[0040] By configuring monitoring key points based on bending parameters, deploying camera arrays, and building a time-series image dataset, the tube bending machine system can accurately monitor key areas and capture every detail of the tube bending process in real time, facilitating subsequent error identification and correction.

[0041] Step 5: Perform real-time dynamic splicing on the time series image data set, and perform error identification based on the dynamic real-time splicing result, the time series mapping fitting result, and the monitoring key points to establish the error identification result.

[0042] Specifically, real-time dynamic stitching refers to integrating image data from different time points and angles in a time-series image data set to generate a continuous and complete visual model that reflects the dynamic changes in the pipe bending process. Specifically, the time-series image data set is processed in real time, and multiple image data are dynamically stitched based on existing 3D modeling technology to form a continuous and dynamic 3D model. This stitching process must not only consider the time series relationship of each frame of the image, but also ensure the matching of the spatial position and angle between the images, so as to generate a dynamic model that can reflect the entire process of pipe bending, as a dynamic real-time stitching result, which can be achieved through existing image stitching technology.

[0043] Then, error identification is performed based on the dynamic real-time splicing results, the timing mapping fitting results and the monitoring key points. The dynamic real-time splicing results provide a complete and continuous visual model that reflects the actual deformation of the pipe during the pipe bending process. The timing mapping fitting results are obtained by fitting and analyzing the bending control parameters at different time points in the pipe bending operation with the actual deformation data, and the relationship between the control parameters at different time points and the actual pipe bending results is obtained to help understand and predict the deformation of the pipe at different time points. Error identification refers to identifying the deviations or anomalies that occur at the monitoring key points in the pipe bending process by comparing the dynamic real-time splicing results and the timing mapping fitting results. These deviations may be caused by rebound, material stress concentration or improper setting of control parameters. The monitoring key points play a key role in this process because these points are often the areas most prone to deformation or errors. By focusing on the dynamic changes of the monitoring key points, potential errors can be identified more accurately.

[0044] For example, when bending a steel pipe, the camera array collects image data from the entire bending process and splices these image data into a continuous 3D model in real time. If the actual deformation of a key point in the bending process is inconsistent with the timing mapping fitting result, the error will be identified. For example, the timing mapping fitting result shows that the point should be bent to 30 degrees, but the splicing result shows that the actual bending angle is only 28 degrees, which indicates that a 2-degree deviation occurred in the process. In this way, errors can be discovered and identified in a timely manner, providing a basis for subsequent adjustments.

[0045] Finally, based on the identified errors, error identification results are established and fed back to the pipe bending control system for further optimization and adjustment of pipe bending parameters to ensure the accuracy and consistency of subsequent operations. Through this series of steps, every detail of the pipe bending process can be effectively monitored, errors can be identified and corrected in a timely manner, and the quality and accuracy of pipe bending can be improved.

[0046] Step 6: Correct the first bend position deviation according to the error identification result and the updated bend control parameters.

[0047] Specifically, in the pipe bending operation, the error identification results include the angle deviation, insufficient or excessive deformation of the pipe fittings at specific positions during the pipe bending process. Through these data, the specific problems and degree of deviation in the current pipe bending operation can be clearly understood. The first bend position deviation correction is performed based on the error identification results and the updated pipe bending control parameters. The goal of the first bend position deviation correction is to adjust the pipe bending machine system so that the pipe fittings can accurately reach the predetermined position and angle during the first bend. The updated pipe bending control parameters are updated again based on the error identification results, that is, if the error identification results show that there is a deviation in the first bend position, the pipe bending control parameters of the pipe bending machine will be automatically adjusted to correct the deviation. For example, if the error identification results show that the actual displacement of the bending point is 2 mm more than expected, the initial force application position will be slightly adjusted back to correct this deviation.

[0048] For example, when an aluminum alloy tube is bent 90 degrees, the error identification result shows that the first bend position is 3 mm offset from the expected position. The bending control parameters are automatically adjusted to increase the initial bending angle by 1 degree and adjust the force application point slightly toward the target position. After the first bend position deviation correction is performed, the bending operation will be performed again. This time the operation will be closer to the design requirements, reduce the position deviation, and ensure that the final bending result is accurate.

[0049] Through this series of steps, the deviation of the first bend position can be effectively corrected to ensure the accuracy of the pipe bending operation, thereby improving the quality and consistency of the product.

[0050] Further, step three of this application includes:

[0051] Establish an optimization goal for springback compensation optimization, wherein the optimization goal includes a minimum springback error goal, a shortest response time goal, and a minimum material stress concentration; configure a basic judgment goal and a reward response goal for each goal according to the optimization goal; establish an evaluation function through the basic judgment goal and the reward response goal, and use the evaluation function to complete springback compensation optimization.

[0052] Specifically, the optimization goal of springback compensation optimization refers to the specific performance indicators to be achieved during the optimization process. These indicators ensure that the final bending results meet the expected quality standards. The above optimization goals include the minimum springback error target, the shortest response time target, and the minimum material stress concentration. The minimum springback error means that after bending the pipe, the springback error between the actual bending angle and the target angle should be as small as possible; the shortest response time target means that when performing springback compensation, the response time should be as short as possible. A shorter response time means that adjustments can be made faster, thereby improving production efficiency; the minimum material stress concentration means that during the bending process, the stress distribution inside the material should be as uniform as possible to avoid damage or deformation of the material due to excessive local stress. This goal is to ensure that the material maintains structural integrity and strength after bending.

[0053] In order to achieve the above optimization goals, it will be further refined into basic judgment goals and reward response goals. The basic judgment goals refer to the basic requirements that must be met in each optimization calculation. For example, the rebound error must be controlled within an acceptable range (such as not more than 1 degree), and the material stress concentration must be less than the yield strength of the material. The reward response goal is an additional goal set to encourage higher performance. For example, if the compensation can be completed in a shorter time, a higher evaluation will be obtained; or, if the stress distribution is more uniform, additional points will be obtained in the evaluation.

[0054] In order to quantify the achievement of the above goals, an evaluation function is constructed. The evaluation function is a mathematical formula. The evaluation function is usually composed of the weighted combination of each basic judgment target and the reward response target. Exemplarily, the basic judgment targets include actual rebound error, target rebound error, actual response time, target response time, actual material stress concentration, and target material stress concentration. The reward response target includes the reward response of the partial rebound error, the reward response of the response time, and the reward response of the material stress concentration. The evaluation function is: ,in, represents the evaluation function, Used to measure the actual springback error Relative to target error The closer the deviation is to 0, the closer the rebound error is to the target value. Used to measure actual response time Relative to target response time Deviation; Used to measure actual material stress concentration Stress concentration relative to the target material Deviation; , , They are the bonus coefficients for rebound error, response time and stress concentration, which are set by professionals in this field; , , It is the importance coefficient for different objectives, which can adjust the relative weight of each objective in the overall optimization. Using the above evaluation function, multiple springback compensation schemes are optimized. Through repeated calculations and evaluations, a parameter combination that can minimize the evaluation function value is selected as the springback compensation optimization result. This means that the selected scheme has achieved the best balance in terms of springback error, response time, material stress concentration, etc., and meets the set optimization goals. In the complex pipe bending process, through scientific optimization goals and evaluation functions, springback compensation optimization can be effectively performed, thereby improving the accuracy, efficiency and product quality of the pipe bending.

[0055] Taking a set of experimental data as an example, the importance coefficient , , Set to 0.5, 0.3, and 0.2 respectively; reward coefficient , , They are set to 0.1, 0.2, and 0.05 respectively; the target springback error is 2.0; the target response time is 1.3; and the target material stress concentration is 80.

[0056] The first group of experimental data: the actual springback value is 2.0, the actual response time is 1.2, the actual material stress concentration is 80, and the evaluation function value is -0.143; the second group of experimental data: the actual springback value is 1.5, the actual response time is 1.4, the actual material stress concentration is 75, and the evaluation function value is -0.234; the third group of experimental data: the actual springback value is 1.8, the actual response time is 1.3, the actual material stress concentration is 85, and the evaluation function value is -0.158; the fourth group of experimental data: the actual springback value is 2.2, the actual response time is 1.1, the actual material stress concentration is 78, and the evaluation function value is -0.121; the fifth group of experimental data: the actual springback value is 1.9, the actual response time is 1.5, the actual material stress concentration is 82, and the evaluation function value is -0.094. The smaller the evaluation function value, the better. From the experimental data, it can be seen that the springback compensation scheme corresponding to the second group of experimental data is the best among the five groups of schemes, and the scheme can be iteratively optimized according to this scheme.

[0057] Furthermore, the present application also includes the following steps:

[0058] Initialize the population, and initialize the population including randomly generating a combination of bending pipe control parameters; evaluate the fitness of each solution in the population according to the evaluation function, if any solution cannot meet all the basic judgment targets, the corresponding solution is eliminated, if any solution meets all the basic judgment targets, the reward response target trigger judgment of the corresponding solution is executed, and the fitness evaluation is completed according to the trigger result; iterate the population according to the fitness evaluation result to complete the rebound compensation optimization.

[0059] Specifically, the population refers to possible combinations of bending control parameters. The bending control parameter combinations can be randomly generated based on the historical bending control parameters. The purpose of random generation is to ensure that the population is diverse so that more possible solutions can be covered in the subsequent optimization process to complete the population initialization. Each solution in the population, that is, a set of bending control parameter combinations, is called an individual. After the population is generated, the fitness of each individual is evaluated according to the evaluation function constructed previously, that is, the individuals in the population are input into the evaluation function and the corresponding fitness is output. Fitness refers to the degree of superiority or inferiority of an individual in the current environment, and reflects the performance of the bending control parameter combination in meeting the rebound compensation optimization goal. The evaluation function will comprehensively score the rebound error, response time and stress concentration of each individual to generate a fitness value.

[0060] If the fitness of an individual does not meet the set basic judgment target, such as the rebound error exceeds the allowable range or the stress concentration is too high, then the individual will be eliminated, that is, removed from the population. If an individual meets all the basic judgment targets, it will also be checked whether it meets the reward response target. If it does, it will be given extra points in the fitness evaluation. For example, after the evaluation function of an individual is calculated, the rebound error is 0.4 degrees (lower than the basic judgment target of 0.5 degrees), the response time is 1.5 seconds (lower than the basic judgment target of 2 seconds), and the stress concentration is 75% (lower than the basic judgment target of 80%). This individual not only meets all the basic judgment targets, but also obtains additional fitness points because its response time is less than 1 second. The reward target, thus obtaining a fitness score. After completing the fitness evaluation, the population is iterated based on the fitness evaluation results. That is, the genetic algorithm is used to simulate the natural selection process, retain individuals with higher fitness, and generate new individuals through crossover and mutation operations to replace individuals with lower fitness. Crossover refers to combining some parameters of two high-fitness individuals to generate a new individual, while mutation refers to randomly changing some parameters in a certain individual to increase the diversity of the population. For example, in a first-generation population, two high-fitness individuals have the following parameter combinations: the bending angle of individual A is 93 degrees and the applied force is 520N; the bending angle of individual B is 91 degrees and the bending speed is 9mm / s. Through the crossover operation, a new individual C is generated, with a bending angle of 93 degrees and a bending speed of 9mm / s, combining the advantages of A and B. At the same time, through the mutation operation, the bending angle of individual C may be fine-tuned to 92.5 degrees to further optimize its fitness.

[0061] With the continuous iteration of the population, it will gradually approach the optimal combination of bending control parameters. Each new generation of population will have better fitness than the previous generation, and eventually an individual with the highest fitness will be found after multiple iterations as the result of springback compensation optimization. Therefore, by continuously optimizing the combination of bending control parameters, the best solution is finally found in the springback compensation optimization, which can not only effectively reduce the springback error, but also ensure that the response time and material stress distribution in the bending process are in the best state, thereby improving the overall quality and efficiency of the bending operation.

[0062] Further, step five of this application includes:

[0063] The bending parameters are segmented into stages, and the distance depth change characteristics of the monitoring key points are established based on the stage segmentation results; the distance radiation weights of the monitoring key points are configured according to the distance depth change characteristics; additional verification of the monitoring key point area of ​​the dynamic real-time splicing results and the timing mapping fitting results is performed according to the distance radiation weights, and the additional verification is a weighted consistency verification of the dynamic real-time splicing results and the timing mapping fitting results through the distance radiation weights; and the error identification results are established according to the additional verification results and the monitoring key point verification results.

[0064] Specifically, the stage segmentation of bending parameters refers to dividing the entire pipe bending process into several stages, each stage corresponding to different bending conditions and process requirements, that is, according to the changing characteristics of the pipe bending process, the bending parameters are segmented and analyzed by time or operation steps. Through segmentation, the specific impact of the pipe bending operation on the deformation of the pipe fittings in different stages can be more clearly identified. For example, when bending a steel pipe, the pipe bending process can be divided into the initial bending stage (small bending angle, gradually increasing stress), the intermediate bending stage (stress reaches the maximum, deformation is most obvious), and the final bending stage (springback begins, stress release). Each stage corresponds to a different combination of bending parameters.

[0065] After the stage segmentation is completed, the distance depth change characteristics of the key monitoring points are established based on the stage segmentation results. The distance depth change characteristics refer to the distance change of the key monitoring points on the pipe surface relative to the initial position during the pipe bending process. These changes reflect the deformation and stress distribution of the pipe at different stages. For example, in the middle bending stage, the distance depth change characteristics of a key monitoring point show that its vertical displacement relative to the initial position has increased by 5 mm, indicating that the point has been subjected to a large bending stress at this stage.

[0066] Based on the distance depth variation characteristics, a corresponding distance radiation weight is configured for each key monitoring point. The distance radiation weight refers to the weight value assigned to different key monitoring points, which is used to measure the importance of these key points to deformation monitoring in the overall pipe bending process. The distance radiation weight refers to the weight coefficient of the influence of the key monitoring point on the surrounding area in space, that is, a weight field is applied around a key point. The closer the area is to the point, the greater its influence and the higher the weight. Therefore, more attention needs to be paid in subsequent error identification. The weight value is proportional to the distance depth variation characteristics. For example, in the pipe bending process, there is a key monitoring point located at the center of the bending radius, while other monitoring points are distributed at different positions of the pipe. Since the deformation at the center of the bending radius is larger and the distance depth variation characteristics are more significant, the distance radiation weight of this monitoring point will be larger than other positions and is set to 0.8, while the weight of a monitoring point far away from the bending radius is only 0.3. In this way, when analyzing the data, the influence of the monitoring point at the center of the bending radius will be given priority, because it has a greater impact on the final bending quality. Assuming that the deformation detected by this monitoring point is 10 mm, and the deformation of the distant monitoring point is 2 mm, according to the weight ratio of 0.8 and 0.3, the focus is on the data of the former, thereby ensuring that the analysis of key positions is more accurate.

[0067] After configuring the distance radiation weights, use these weights to perform additional verification on the dynamic real-time stitching results and the timing mapping fitting results. Additional verification means further confirming the accuracy and reliability of these key points in actual operation by comparing the performance of the key points in the stitching results and the fitting results. If the additional verification shows that the actual position of some key points is significantly different from the fitted position, record these deviations as additional verification results.

[0068] Finally, based on the additional verification results and the verification results of the key monitoring points, the error identification results are comprehensively established, where the verification results of the key monitoring points refer to the errors of the key monitoring points. The error identification results show the comprehensive analysis results of the deviations shown by each key monitoring point in the entire pipe bending process, which are used to guide the subsequent springback compensation and operation adjustments, so as to facilitate the subsequent finding of the best compensation strategy under complex pipe bending conditions and ensure the high quality and high precision of the final product.

[0069] Furthermore, the present application also includes the following steps:

[0070] Based on the monitoring key points, key point positioning of the dynamic real-time splicing results and the timing mapping fitting results is performed, and the key point positioning includes evaluating and screening the position criticality and deformation amount of each monitoring key point in the dynamic real-time splicing results and the timing mapping fitting results; performing independent error analysis of the key points according to the key point positioning results to generate a first error verification result; locating the maximum deformation point based on the independent error analysis, taking the maximum deformation point as a reference point, performing error fitting, establishing an error fitting result, and generating a second error verification result based on the determination coefficient of the error fitting result; generating a monitoring key point verification result based on the first error verification result and the second error verification result, and establishing an error identification result based on the generated monitoring key point verification result and the additional verification result.

[0071] Specifically, key point positioning is to accurately align and evaluate the dynamic real-time splicing results and the time series mapping fitting results through monitoring key points. First, based on the monitoring key points, the real-time dynamic real-time splicing results and the time series mapping fitting results are analyzed to determine the relative position and importance of each monitoring key point in the two data sets. Position criticality refers to the degree of influence of each monitoring key point on the bending shape during the entire pipe bending process. After collecting historical data, the degree of influence of different positions on the bending shape can be analyzed by existing association analysis methods. Usually, the position criticality is higher at the bending radius or the part where the pipe is subjected to the greatest force. For example, the position criticality of the key point located in the middle of the pipe bending is relatively large when bending, because its deformation has the most significant impact on the final bending quality. Then, the deformation evaluation and screening of each monitoring key point is performed, that is, the monitoring points with greater impact on the overall accuracy are screened out by analyzing the deformation amplitude of each monitoring point during the pipe bending process. For example, if the deformation of a certain monitoring point reaches 10mm during the entire pipe bending process, while the deformation of other monitoring points is only 2 to 3mm, the point with larger deformation will be given priority. A monitoring point with a large deformation means that the pipe at that point is subjected to greater stress or deformation and is more prone to errors, so it needs to be monitored and analyzed in detail.

[0072] After completing the positioning of the key points, an independent error analysis is performed. Independent error analysis refers to evaluating the deviation of each key point separately and analyzing the error size of each point independently, rather than evaluating all monitoring points as a whole. For example, for key points with large deformation variables, focus on analyzing the difference between the dynamic splicing results and the timing fitting results of the point to see if there are significant deviations in position or deformation. Suppose the dynamic splicing results of a key point show that it has offset by 5mm, and the fitting result predicts an offset value of 3mm, then the independent error of the key point is 2mm. Through independent error analysis, the first error verification result can be generated, which is a preliminary error assessment used to identify which monitoring points have large errors during the pipe bending process, thereby providing a basis for subsequent pipe bending accuracy corrections.

[0073] Based on the results of independent error analysis, the maximum deformation point is further located, that is, the key point where the deformation is most significant during the pipe bending process. The maximum deformation point is usually the area with the largest error. Taking this point as the reference point, error fitting is performed, that is, the error of this reference point is fitted and calculated to more accurately understand and describe its deformation.

[0074] For example, if in the independent error analysis, a key point is found to have the maximum deformation, and its deviation value is 10 mm, this point is used as the reference point, and its deformation trend and error distribution are calculated through the fitting model. The fitting model is used to simulate deformation and can be built based on historical deformation data and existing machine learning model training. After the error fitting is completed, the determination coefficient is calculated according to the fitting results to generate the second error verification result. The determination coefficient is a statistical indicator used to measure the ability of the fitting model to explain the data, that is, the accuracy and reliability of the error fitting results evaluated by the fitting model. If the determination coefficient is high, it means that the error fitting result can better describe the actual error situation. It can be determined by the convergence accuracy when training the fitting model. That is, the fitting result is weighted by the determination coefficient to generate the second error verification result.

[0075] Finally, the first error verification result is combined with the second error verification result to generate the monitoring key point verification result, which is then integrated with the additional verification result to finally establish the error identification result. The error identification result is a comprehensive evaluation of the error situation in the entire pipe bending process, identifies the key areas that need to be corrected, and provides data support for springback compensation optimization.

[0076] Through the above steps, the errors generated in the pipe bending process can be accurately located and analyzed, which is convenient for subsequent optimization of springback compensation, improves the accuracy of the pipe bending operation, and ensures the high quality and consistency of the final product.

[0077] Furthermore, the present application also includes the following steps:

[0078] The additional verification results are subjected to fitting consistency evaluation of the error fitting results to generate a fitting consistency evaluation result; the monitoring key point verification results are compensated according to the fitting consistency evaluation results to construct an error identification result, and the error compensation specifically includes establishing an error coefficient according to the fitting consistency evaluation results, performing coefficient-weighted calculation on the monitoring key point verification results through the error coefficient to generate an error identification result.

[0079] Specifically, the fitting consistency evaluation refers to judging the accuracy and reliability of the error fitting by evaluating the degree of match between the additional verification results and the error fitting results, or in layman's terms, the degree of similarity, and obtaining a fitting consistency evaluation result including the degree of match. The fitting consistency evaluation result reflects the degree of match between the additional verification results and the error fitting results. A high consistency evaluation result means that the error fitting process is reliable. According to the generated fitting consistency evaluation results, the verification results of the key monitoring points are compensated. Compensation means adjusting or correcting the verification results of the key monitoring points according to the fitting consistency evaluation results to better reflect the actual error situation and improve the final correction accuracy. Specifically, an error coefficient is generated for each key monitoring point according to the fitting consistency evaluation results, and the error coefficient is used to adjust the key monitoring point verification results of the monitoring point. For example, the initial error of a certain monitoring point is 3mm, and the fitting consistency evaluation result is 0.9, indicating that there is an error of 0.1, and 0.1 is the error coefficient. The verification results of the key points are weighted according to the error of 0.1, and the correction error of the point after weighted calculation will become 3.3mm. In this way, the deviation value is corrected to improve the accuracy of error identification and ensure the accuracy of pipe bending operation. Exemplarily, the coefficient weighted calculation is performed using the following formula: fitting error = error coefficient × monitoring key point verification result.

[0080] Finally, based on the verification results of the monitored key points after compensation and combined with the additional verification results, the error identification results are finally constructed to accurately identify and compensate for the errors generated in the pipe bending process, providing an important reference for subsequent springback compensation and ensuring the quality and consistency of the pipe bending operation.

[0081] Furthermore, the present application also includes the following steps:

[0082] Configure the stage deviation threshold; after the error identification result is generated, call the actual error data to perform trigger verification of the stage deviation threshold and generate a trigger warning; use the trigger warning to manage the abnormality of the bent pipe.

[0083] Specifically, the stage deviation threshold refers to the allowable error range set by professional and technical personnel in this field based on practical experience for different stages of the pipe bending operation. The pipe bending process is usually divided into multiple stages, such as initial bending, transition bending and final shaping. At each stage, an allowable error range is set according to the bending parameters and material properties, namely the stage deviation threshold. This threshold is used to evaluate whether the bending results of each stage are within an acceptable range to ensure the accuracy of the pipe bending operation.

[0084] When the pipe bending operation is completed, the error identification results will be generated, including the actual error and deformation of each key monitoring point. The error identification results will be called and compared with the pre-set stage deviation threshold for trigger verification. Trigger verification refers to checking whether the actual error exceeds the set deviation threshold. If the error of a key point exceeds the threshold range, an early warning will be triggered immediately to remind the operator or the automatic control system that the current pipe bending operation may be abnormal and needs to be adjusted in time. After the trigger warning is generated, the pipe bending abnormality management mechanism is started. Pipe bending abnormality management refers to automatically or manually adjusting the pipe bending operation parameters according to the early warning information to correct possible errors and prevent further deviations. Management measures may include recalibrating the pipe bending machine, adjusting the bending angle or speed, or even suspending the operation for detailed inspection. The specific settings are set by professional and technical personnel in this field, so as to effectively detect and prevent abnormal conditions that may affect the quality of the final product, improve the accuracy and stability of the pipe bending, and ensure the high quality and consistency of the final product.

[0085] In summary, the method for correcting the first bend position deviation of a pipe bending machine provided in this application has the following technical effects:

[0086] After the initialization of the pipe bending machine system is completed, the pipe fitting data is synchronized to the pipe bending machine system, and the three-dimensional model data is called through the pipe bending machine system, and the bending control parameters are set according to the bending parameters in the three-dimensional model data and the pipe fitting data; a historical bending database is established, and a springback prediction network is configured with the historical bending database. The springback prediction network is constructed by supervised learning by dividing the historical bending database into a training set and a test set. The springback prediction network is a machine learning model network. After feature extraction of the pipe fitting data, the feature extraction result is input into the springback prediction network, and the springback prediction result is generated after processing by the springback prediction network; according to the springback prediction result The springback compensation optimization is performed based on the results and the bending control parameters, the bending control parameters are updated with the springback compensation optimization results, and the time series mapping fitting results are established according to the updated bending control parameters; the monitoring key points are configured based on the bending parameters, the camera array is configured with the monitoring key points, and a time series image data set is established based on the camera array, and the time series image data set is constructed with the monitoring key points as the collection center; the time series image data set is spliced ​​in real time, and error identification is performed based on the dynamic real-time splicing results, the time series mapping fitting results, and the monitoring key points to establish error identification results; the first bend position deviation is corrected according to the error identification results and the updated bending control parameters. Springback prediction is performed by combining the three-dimensional model data and the historical bending database through the springback prediction network, and then springback compensation optimization is performed according to the springback prediction results, and a time series mapping fitting result is established. The monitoring key points are further configured for monitoring image splicing, and error identification is performed in combination with the time series mapping fitting results, so that the first bend position deviation in the bending process is corrected in real time, so as to correct the deviation of the first bend position, ensure the accuracy of the bending operation, and thus improve the quality and consistency of the product.

[0087] Embodiment 2: Based on the method for correcting the first bend position deviation of a pipe bender in the previous embodiment, the present application also provides a device for correcting the first bend position deviation of a pipe bender, as shown in the attached Figure 2 , the first bend position deviation correction device for a pipe bending machine comprises:

[0088] The pipe bending control parameter setting module 11 is used to synchronize the pipe fitting data to the pipe bending machine system after the pipe bending machine system is initialized, call the three-dimensional model data through the pipe bending machine system, and set the pipe bending control parameters according to the bending parameters in the three-dimensional model data and the pipe fitting data.

[0089] The springback prediction module 12 is used to establish a historical bending database and configure a springback prediction network with the historical bending database. The springback prediction network is constructed by supervised learning by dividing the historical bending database into a training set and a test set. The springback prediction network is a machine learning model network. After feature extraction of the pipe data, the feature extraction result is input into the springback prediction network, and the springback prediction result is generated after processing by the springback prediction network.

[0090] The springback compensation optimization module 13 is used to perform springback compensation optimization according to the springback prediction result and the pipe bending control parameter, update the pipe bending control parameter with the springback compensation optimization result, and establish a time series mapping fitting result according to the updated pipe bending control parameter.

[0091] The image data acquisition module 14 is used to configure monitoring key points based on the bending parameters, configure a camera array based on the monitoring key points, and establish a time-series image data set based on the camera array, wherein the time-series image data set is constructed with the monitoring key points as the acquisition center.

[0092] The error identification module 15 is used to dynamically splice the time series image data set in real time, and to identify errors based on the dynamic real-time splicing results, the time series mapping fitting results, and the monitoring key points to establish error identification results.

[0093] The deviation correction module 16 is used to correct the first bend position deviation according to the error identification result and the updated pipe bending control parameters.

[0094] Furthermore, the error identification module 15 in the first bend position deviation correction device for a pipe bending machine is also used for:

[0095] The bending parameters are segmented into stages, and the distance depth variation characteristics of the monitoring key points are established based on the stage segmentation results; the distance radiation weights of the monitoring key points are configured according to the distance depth variation characteristics; additional verification of the monitoring key point area of ​​the dynamic real-time splicing results and the timing mapping fitting results is performed according to the distance radiation weights, and the additional verification is weighted consistency verification of the dynamic real-time splicing results and the timing mapping fitting results through the distance radiation weights; and the error identification results are established according to the additional verification results and the monitoring key point verification results.

[0096] Furthermore, the error identification module 15 in the first bend position deviation correction device for a pipe bending machine is also used for:

[0097] Based on the monitoring key points, key point positioning of the dynamic real-time splicing results and the timing mapping fitting results is performed, and the key point positioning includes evaluating and screening the position criticality and deformation amount of each monitoring key point in the dynamic real-time splicing results and the timing mapping fitting results; performing independent error analysis of the key points according to the key point positioning results to generate a first error verification result; locating the maximum deformation point based on the independent error analysis, taking the maximum deformation point as a reference point, performing error fitting, establishing an error fitting result, and generating a second error verification result based on the determination coefficient of the error fitting result; generating a monitoring key point verification result based on the first error verification result and the second error verification result, and establishing an error identification result based on the generated monitoring key point verification result and the additional verification result.

[0098] Furthermore, the error identification module 15 in the first bend position deviation correction device for a pipe bending machine is also used for:

[0099] The additional verification results are subjected to fitting consistency evaluation of the error fitting results to generate a fitting consistency evaluation result; the monitoring key point verification results are compensated according to the fitting consistency evaluation results to construct an error identification result, and the error compensation specifically includes establishing an error coefficient according to the fitting consistency evaluation results, performing coefficient-weighted calculation on the monitoring key point verification results through the error coefficient to generate an error identification result.

[0100] Furthermore, the springback compensation optimization module 13 in the first bend position deviation correction device for a pipe bending machine is also used for:

[0101] Establish an optimization goal for springback compensation optimization, wherein the optimization goal includes a minimum springback error goal, a shortest response time goal, and a minimum material stress concentration; configure a basic judgment goal and a reward response goal for each goal according to the optimization goal; establish an evaluation function through the basic judgment goal and the reward response goal, and use the evaluation function to complete springback compensation optimization.

[0102] Furthermore, the springback compensation optimization module 13 in the first bend position deviation correction device for a pipe bending machine is also used for:

[0103] Initialize the population, and initialize the population including randomly generating a combination of bending pipe control parameters; evaluate the fitness of each solution in the population according to the evaluation function, if any solution cannot meet all the basic judgment targets, the corresponding solution is eliminated, if any solution meets all the basic judgment targets, the reward response target trigger judgment of the corresponding solution is executed, and the fitness evaluation is completed according to the trigger result; iterate the population according to the fitness evaluation result to complete the rebound compensation optimization.

[0104] Furthermore, the first bend position deviation correction device for a pipe bending machine further includes a pipe bending abnormality management module, and the pipe bending abnormality management module is used to:

[0105] Configure the stage deviation threshold; after the error identification result is generated, call the actual error data to perform trigger verification of the stage deviation threshold and generate a trigger warning; use the trigger warning to manage the abnormality of the bent pipe.

[0106] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The method and specific examples for correcting the first bend position deviation of a pipe bender in the first embodiment are also applicable to the device for correcting the first bend position deviation of a pipe bender in this embodiment. Through the detailed description of the method for correcting the first bend position deviation of a pipe bender, those skilled in the art can clearly know the device for correcting the first bend position deviation of a pipe bender in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method description.

[0107] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0108] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for correcting the first bend position deviation of a pipe bending machine, characterized in that: The first bend position deviation correction method for a pipe bender comprises: After the pipe bending machine system is initialized, the pipe fitting data is synchronized to the pipe bending machine system, the three-dimensional model data is called by the pipe bending machine system, and the pipe bending control parameters are set according to the bending parameters in the three-dimensional model data and the pipe fitting data; Establishing a historical bending database, configuring a springback prediction network with the historical bending database, wherein the springback prediction network is constructed by performing supervised learning by dividing the historical bending database into a training set and a test set, wherein the springback prediction network is a machine learning network, and after extracting features from the pipe fitting data, the feature extraction results are input into the springback prediction network, and the springback prediction results are generated after being processed by the springback prediction network; Performing springback compensation optimization according to the springback prediction result and the pipe bending control parameter, updating the pipe bending control parameter with the springback compensation optimization result, and establishing a time series mapping fitting result according to the updated pipe bending control parameter; Configure monitoring key points based on the bending parameters, configure a camera array based on the monitoring key points, establish a time series image data set based on the camera array, and construct the time series image data set with the monitoring key points as the collection center; The time series image data set is dynamically spliced ​​in real time, and error identification is performed based on the dynamic real-time splicing result, the time series mapping fitting result, and the monitoring key points to establish the error identification result; The first bend position deviation is corrected according to the error identification result and the updated bend control parameters.

2. The method for correcting the first bend position deviation of a pipe bending machine according to claim 1, characterized in that: The error identification is performed based on the dynamic real-time splicing result, the time series mapping fitting result, and the monitoring key points, and the error identification result is established, and further includes: Performing stage segmentation on the bending parameters, and establishing distance depth variation characteristics of key monitoring points based on the stage segmentation results; Configure the distance radiation weight of the monitoring key point according to the distance depth variation characteristics; Additional verification of the monitoring key point area of ​​the dynamic real-time splicing result and the time series mapping fitting result is performed according to the distance radiation weight, and the additional verification is weighted consistency verification of the dynamic real-time splicing result and the time series mapping fitting result is performed according to the distance radiation weight; The error identification results are established based on the additional verification results and the verification results of the key monitoring points.

3. The method for correcting the first bend position deviation of a pipe bending machine according to claim 2, characterized in that: The establishing of the error identification result according to the additional verification result and the monitoring key point verification result also includes: Based on the monitoring key points, key point positioning of the dynamic real-time splicing result and the time series mapping fitting result is performed, wherein the key point positioning includes evaluating and screening the position criticality and deformation amount of each monitoring key point in the dynamic real-time splicing result and the time series mapping fitting result; Perform independent error analysis of key points based on key point positioning results to generate a first error verification result; Locating the maximum deformation point based on the independent error analysis, taking the maximum deformation point as a reference point, performing error fitting, establishing an error fitting result, and generating a second error verification result based on a determination coefficient of the error fitting result; A monitoring key point verification result is generated according to the first error verification result and the second error verification result, and an error identification result is established according to the generated monitoring key point verification result and the additional verification result.

4. The method for correcting the first bend position deviation of a pipe bending machine according to claim 3, characterized in that: The step of establishing the error identification result according to the generated monitoring key point verification result and the additional verification result also includes: Performing a fitting consistency evaluation of the error fitting result on the additional verification result to generate a fitting consistency evaluation result; The monitoring key point verification results are compensated according to the fitting consistency evaluation results to construct an error identification result. The error compensation specifically includes establishing an error coefficient according to the fitting consistency evaluation results, performing coefficient-weighted calculation on the monitoring key point verification results through the error coefficient, and generating an error identification result.

5. The method for correcting the first bend position deviation of a pipe bending machine according to claim 1, characterized in that: The springback compensation optimization according to the springback prediction result and the bending pipe control parameter also includes: Establishing optimization objectives for springback compensation optimization, wherein the optimization objectives include a minimum springback error objective, a shortest response time objective, and a minimum material stress concentration; According to the optimization goal, a basic judgment goal and a reward response goal are configured for each goal; An evaluation function is established through the basic judgment target and the reward response target, and the rebound compensation optimization is completed using the evaluation function.

6. The method for correcting the first bend position deviation of a pipe bending machine according to claim 5, characterized in that: The method of using the evaluation function to complete the springback compensation optimization also includes: Initializing the population, the initialization of the population includes randomly generating a combination of bending pipe control parameters; The fitness of each solution in the population is evaluated according to the evaluation function. If any solution cannot meet all the basic judgment targets, the corresponding solution is eliminated. If any solution meets all the basic judgment targets, the reward response target trigger judgment of the corresponding solution is executed, and the fitness evaluation is completed according to the trigger result; The population is iterated according to the fitness evaluation results to complete the rebound compensation optimization.

7. The method for correcting the first bend position deviation of a pipe bender according to claim 1, characterized in that: Also includes: Configure stage deviation threshold; After the error identification result is generated, the actual error data is called to perform trigger verification of the stage deviation threshold and generate a trigger warning; The trigger warning is used to manage abnormal bending of pipes.

8. A device for correcting the first bend position deviation of a pipe bending machine, characterized in that: The steps for implementing the method for correcting the first bend position deviation of a pipe bending machine according to any one of claims 1 to 7, wherein the device for correcting the first bend position deviation of a pipe bending machine comprises: A pipe bending control parameter setting module is used to synchronize the pipe fitting data to the pipe bending machine system after the pipe bending machine system is initialized, call the three-dimensional model data through the pipe bending machine system, and set the pipe bending control parameters according to the bending parameters in the three-dimensional model data and the pipe fitting data; A springback prediction module is used to establish a historical bending database, configure a springback prediction network with the historical bending database, and construct the springback prediction network by performing supervised learning by dividing the historical bending database into a training set and a test set. The springback prediction network is a machine learning model network, and after extracting features from the pipe fitting data, the feature extraction results are input into the springback prediction network, and the springback prediction results are generated after being processed by the springback prediction network; A springback compensation optimization module, used for performing springback compensation optimization according to the springback prediction result and the pipe bending control parameter, updating the pipe bending control parameter with the springback compensation optimization result, and establishing a time series mapping fitting result according to the updated pipe bending control parameter; An image data acquisition module, configured to configure monitoring key points based on the bending parameters, configure a camera array based on the monitoring key points, and establish a time-series image data set based on the camera array, wherein the time-series image data set is constructed with the monitoring key points as the acquisition center; An error identification module is used to dynamically splice the time series image data set in real time, and to identify errors based on the dynamic real-time splicing results, the time series mapping fitting results, and the monitoring key points, and to establish error identification results; The deviation correction module is used to correct the first bend position deviation according to the error identification result and the updated bend control parameters.

Images