CN-122021272-A - Pipe bending and forming rebound prediction method, device, equipment and medium

CN122021272ACN 122021272 ACN122021272 ACN 122021272ACN-122021272-A

Abstract

The application provides a pipe bending rebound prediction method, a device, equipment and a medium, relates to the technical field of data processing. The method comprises the steps of firstly obtaining first-number-group sample tube bending forming data, secondly performing redundancy screening on tube characteristic quantities on the first-number-group sample tube bending forming data to form first-number-group screened tube bending forming data, then constructing a rebound angle prediction model based on each group of screened tube bending forming data and corresponding rebound angle labels, constructing a rebound radius prediction model based on each group of screened tube bending forming data and corresponding rebound radius labels, and finally determining target process parameter data that rebound angles meet rebound angle threshold conditions and rebound radii meet rebound radius threshold conditions by using the rebound angle prediction model and the rebound radius prediction model. Based on the above, the problems of low prediction accuracy or high prediction cost in the prior art can be improved.

Inventors

DU JUAN
TAN YAO
ZHANG HAOBO
CHEN XIANNENG
CHEN LI
SUN TAO
GAO SHENYUAN
YANG YUANXI

Assignees

成都飞机工业（集团）有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The pipe bending forming rebound prediction method is characterized by comprising the following steps of: Obtaining a first number of sets of sample tube bending forming data, wherein each set of sample tube bending forming data comprises values of at least one technological parameter and values of various tube characteristic quantities, and each set of sample tube bending forming data is correspondingly provided with a rebound angle label and a rebound radius label; redundant screening of pipe characteristic quantities is carried out on the first number of groups of sample pipe bending forming data to form a first number of groups of screened pipe bending forming data, wherein the number of pipe characteristic quantities included in each group of screened pipe bending forming data is smaller than or equal to the number of pipe characteristic quantities included in each group of sample pipe bending forming data; Constructing a rebound angle prediction model at least based on each set of screening pipe bending forming data and a corresponding rebound angle label, and constructing a rebound radius prediction model at least based on each set of screening pipe bending forming data and a corresponding rebound radius label; And searching a front edge surface under the pipe characteristic quantity of the target pipe by utilizing the rebound angle prediction model and the rebound radius prediction model, and determining target process parameter data of which the rebound angle meets the rebound angle threshold condition and the rebound radius meets the rebound radius threshold condition.
2. The method of predicting pipe bending springback of claim 1, wherein the step of performing redundant screening of the pipe characteristic values on the first number of sets of sample pipe bending data to form a first number of sets of screened pipe bending data comprises: performing first redundancy screening of pipe characteristic quantities on the first number group of sample pipe bending forming data to form a first number group of intermediate pipe bending forming data, wherein the first redundancy screening refers to screening based on correlation among the pipe characteristic quantities; And performing second redundant screening of the pipe characteristic quantity on the first number group of intermediate pipe bending forming data to form the first number group of screening pipe bending forming data, wherein the second redundant screening refers to screening out based on the correlation importance between the pipe characteristic quantity and the label.
3. The method of predicting pipe bending springback of claim 2, wherein the step of performing a first redundancy screening of pipe characteristics on the first number of sets of sample pipe bending data to form a first number of sets of intermediate pipe bending data comprises: For each two kinds of pipe characteristic quantities, determining a correlation coefficient between the two kinds of pipe characteristic quantities based on values of the two kinds of pipe characteristic quantities in the first number group of sample pipe bending forming data; And performing first redundancy screening on the pipe characteristic quantities corresponding to the plurality of pipe characteristic quantities of the first number group of sample pipe bending forming data based on correlation coefficients between every two pipe characteristic quantities to form first number group of middle pipe bending forming data.
4. The method of predicting pipe bending springback according to claim 3, wherein the step of performing a first redundancy screening of pipe characteristic values for a plurality of pipe characteristic values corresponding to the first number of sets of sample pipe bending data based on a correlation coefficient between every two kinds of pipe characteristic values to form a first number of sets of intermediate pipe bending data comprises: determining, for each two pipe feature quantities, whether an absolute value of a correlation coefficient between the two pipe feature quantities is greater than a predetermined correlation coefficient threshold; For every two pipe characteristic quantities with the absolute value of the correlation coefficient larger than the correlation coefficient threshold, taking a first pipe characteristic quantity of the two pipe characteristic quantities as input, a rebound angle and a rebound radius as output, respectively, establishing a first machine learning model and a second machine learning model, and taking a second pipe characteristic quantity of the two pipe characteristic quantities as input, a rebound angle and a rebound radius as output, respectively, and establishing a third machine learning model and a fourth machine learning model; Training the first machine learning model by utilizing the value of the first pipe characteristic quantity and the corresponding rebound angle label in the first number group of sample pipe bending forming data, determining a corresponding first training error, training the second machine learning model by utilizing the value of the first pipe characteristic quantity and the corresponding rebound radius label in the first number group of sample pipe bending forming data, and determining a corresponding second training error; training the third machine learning model by using the value of the second type of pipe characteristic quantity in the first number group of sample pipe bending forming data and the corresponding rebound angle label, determining a corresponding third training error, training the fourth machine learning model by using the value of the second type of pipe characteristic quantity in the first number group of sample pipe bending forming data and the corresponding rebound radius label, and determining a corresponding fourth training error; And screening out the corresponding two pipe characteristic quantities based on the magnitude relation between the first training error and the third training error to form a first number group of middle pipe bending forming data corresponding to the rebound angle label, and screening out the corresponding two pipe characteristic quantities based on the magnitude relation between the second training error and the fourth training error to form a first number group of middle pipe bending forming data corresponding to the rebound radius label.
5. The method of predicting pipe bending springback according to claim 2, wherein said step of performing a second redundant screening of pipe characteristics for said first number of sets of intermediate pipe bending data to form a first number of sets of screened pipe bending data comprises: For each pipe characteristic quantity included in the first quantity group of intermediate pipe bending forming data, taking each pipe characteristic quantity except the pipe characteristic quantity as input, respectively taking a rebound angle and a rebound radius as output, establishing a fifth machine learning model and a sixth machine learning model, training the fifth machine learning model by utilizing the value of each pipe characteristic quantity and a corresponding rebound angle label which are taken as input in the first quantity group of intermediate pipe bending forming data, determining a corresponding fifth training error, training the sixth machine learning model by utilizing the value of each pipe characteristic quantity and a corresponding rebound radius label which are taken as input in the first quantity group of intermediate pipe bending forming data, and determining a corresponding sixth training error; Screening out the tube characteristic quantity with the minimum value of the corresponding fifth training error in each tube characteristic quantity included in the first quantity group of intermediate tube bending forming data to form a first quantity group of screening tube bending forming data corresponding to the rebound angle label; and screening out the pipe characteristic quantity with the minimum value of the corresponding sixth training error in each pipe characteristic quantity included in the first quantity group of intermediate pipe bending forming data to form a first quantity group of screening pipe bending forming data corresponding to the rebound radius label.
6. The pipe bending and springback prediction method according to any one of claims 1 to 5, wherein said step of constructing a springback angle prediction model based at least on each set of said screened pipe bending and shaping data and corresponding springback angle labels, and constructing a springback radius prediction model based at least on each set of said screened pipe bending and shaping data and corresponding springback radius labels, comprises: Expanding first number of groups of screening pipe bending forming data corresponding to the rebound angle labels based on a process parameter expansion mode to form second number of groups of screening pipe bending forming data corresponding to the rebound angle labels, and expanding first number of groups of screening pipe bending forming data corresponding to the rebound radius labels based on a process parameter expansion mode to form second number of groups of screening pipe bending forming data corresponding to the rebound radius labels, wherein the values of the same process parameter in the first number of groups of sample pipe bending forming data are the same; And screening the pipe bending forming data by using a second number of groups corresponding to the rebound angle labels, training to form a rebound angle prediction model, and training to form a rebound radius prediction model by using a second number of groups corresponding to the rebound radius labels.
7. The method of predicting pipe bending and springback of claim 6, wherein the step of training to form the springback angle prediction model using a second number of sets of screening pipe bending and shaping data corresponding to springback angle labels and training to form the springback radius prediction model using a second number of sets of screening pipe bending and shaping data corresponding to springback radius labels comprises: Taking process parameters and pipe characteristic quantities as input and rebound angles as output, constructing a plurality of initial rebound angle models, screening pipe bending forming data by using a second number group corresponding to rebound angle labels, respectively training each initial rebound angle model to form a plurality of training rebound angle models, and determining one training rebound angle model with the minimum error as a rebound angle prediction model; And respectively training each initial rebound radius model by using a second number of groups corresponding to rebound radius labels to form a plurality of training rebound radius models, and determining one training rebound radius model with the minimum error as a rebound radius prediction model.
8. A pipe bending rebound prediction device is characterized by comprising: The data acquisition module is used for acquiring a first number of groups of sample pipe bending forming data, wherein each group of sample pipe bending forming data comprises at least one technological parameter value and a plurality of pipe characteristic values, and each group of sample pipe bending forming data is correspondingly provided with a rebound angle label and a rebound radius label; The characteristic quantity screening module is used for conducting redundant screening of pipe characteristic quantities on the first number of groups of sample pipe bending forming data to form a first number of groups of screened pipe bending forming data, wherein the number of pipe characteristic quantities included in each group of screened pipe bending forming data is smaller than or equal to the number of pipe characteristic quantities included in each group of sample pipe bending forming data; The prediction model construction module is used for constructing a rebound angle prediction model at least based on each group of the screened pipe bending forming data and the corresponding rebound angle label, and constructing a rebound radius prediction model at least based on each group of the screened pipe bending forming data and the corresponding rebound radius label; And the parameter determining module is used for searching a front surface under the pipe characteristic quantity of the target pipe by utilizing the rebound angle prediction model and the rebound radius prediction model to determine target process parameter data that the rebound angle meets the rebound angle threshold condition and the rebound radius meets the rebound radius threshold condition.
9. An electronic device, comprising: A memory for storing a computer program; a processor connected to the memory for executing a computer program stored in the memory to implement the pipe bending springback prediction method of any one of claims 1-7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when run, performs the pipe bending springback prediction method according to any one of claims 1-7.

Description

Pipe bending and forming rebound prediction method, device, equipment and medium Technical Field The application relates to the technical field of data processing, in particular to a pipe bending forming rebound prediction method, a pipe bending forming rebound prediction device, pipe bending forming rebound prediction equipment and a pipe bending forming rebound prediction medium. Background The thin-wall pipe has the characteristics of light weight, high strength and toughness, high precision and the like, and is widely applied to the fields of aerospace, ships, automobiles and the like. In order to realize medium and energy transmission in a limited space, the pipe needs to be subjected to three-dimensional bending forming. In the bending and forming process of the thin-wall pipe, the defects of wrinkling, cracking, rebound and the like are particularly easy to generate, wherein the rebound is one of the main defects, particularly the rebound of the bending radius and the bending angle is particularly obvious in the cold bending process (room temperature bending), the processing precision and the efficiency of the pipe fitting can be influenced, the forced assembly can be realized in the subsequent assembly process, and the service performance is further influenced. Along with the increasing requirements on the quality and precision of the thin-wall bent pipe fitting, the bending rebound of the pipe needs to be accurately regulated and controlled. Because the bending forming of the pipe is a complex elastoplastic forming process, material nonlinearity, geometric nonlinearity and boundary condition nonlinearity are integrated, and therefore bending resilience is influenced by the intrinsic properties of the pipe, the geometric dimension of the pipe, the technological parameters and the condition parameters. The traditional trial-and-error method obtains the technological parameters of bending rebound control through a trial-and-error combined experiment, but on one hand, a great deal of time and economic cost are wasted, each material change and each specification change need to be subjected to the technological experiment, and on the other hand, even if a control method is obtained for a certain type of pipe, rebound fluctuation can be generated along with fluctuation of material performance and slight change of the geometric dimension of the pipe, so that the quality of the bent pipe is affected. In addition, the bending rebound control of the thin-wall pipe is carried out by means of an analysis method, finite element simulation and the like, so that the cost can be effectively reduced, but the conventional techniques have the defects that 1) the coupling effect of material nonlinearity, geometric nonlinearity and boundary condition nonlinearity is difficult to process by the conventional analysis method, so that the problem of relatively low prediction precision exists, and 2) the cost of single time of finite element simulation calculation is relatively high, so that the problem of relatively high prediction cost exists. Disclosure of Invention In view of the above, the present application aims to provide a method, a device, an apparatus and a medium for predicting springback of pipe bending, so as to solve the problems of low prediction accuracy or high prediction cost in the prior art. In order to achieve the above purpose, the application adopts the following technical scheme: A pipe bending forming rebound prediction method comprises the following steps: Obtaining a first number of sets of sample tube bending forming data, wherein each set of sample tube bending forming data comprises values of at least one technological parameter and values of various tube characteristic quantities, and each set of sample tube bending forming data is correspondingly provided with a rebound angle label and a rebound radius label; redundant screening of pipe characteristic quantities is carried out on the first number of groups of sample pipe bending forming data to form a first number of groups of screened pipe bending forming data, wherein the number of pipe characteristic quantities included in each group of screened pipe bending forming data is smaller than or equal to the number of pipe characteristic quantities included in each group of sample pipe bending forming data; Constructing a rebound angle prediction model at least based on each set of screening pipe bending forming data and a corresponding rebound angle label, and constructing a rebound radius prediction model at least based on each set of screening pipe bending forming data and a corresponding rebound radius label; And searching a front edge surface under the pipe characteristic quantity of the target pipe by utilizing the rebound angle prediction model and the rebound radius prediction model, and determining target process parameter data of which the rebound angle meets the rebound angle threshold condition and the rebound radius meets the reboun