Self-adaptive regulation and control method and system for powerful spinning forming of metal component

Abstract

The invention relates to the field of intelligent manufacturing, and discloses a metal component power spinning forming self-adaptive regulation and control method which comprises the following steps: establishing a virtual machining environment based on a constitutive model of a metal material of a metal component, a dynamic characteristic model of a machine tool and a springback model; in the virtual machining environment, a corresponding reward value is obtained by simulating the powerful spinning process and the springback process of the metal component; the feeding rate, the machining state and the reward value serve as input, a mapping relation is established through a deep learning algorithm, a trained control model is obtained, and the optimized feeding rate is output; and deploying the trained control model to a spinning machine motion controller to realize real-time control of the spinning machine. By means of the spinning wheel, the metal forming acting force of the spinning wheel can be changed, the stress-strain state of a plastic deformation area is improved, and the workpiece quality is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing, and in particular to a method and system for adaptively controlling the high-force spinning of metal components. Background Art

[0002] Metal spinning is a chipless processing technology that is easy to produce seamless annular parts, especially suitable for forming components with large diameter-to-thickness ratio. The metal spinning process has the advantages of saving material consumption, low cost, simple equipment and high product quality. It is currently widely used in key national industrial fields such as aerospace and machinery. With the rapid development of advanced manufacturing technology, the industrial field has an increasing demand for the forming quality of spinning parts, and metal spinning forming quality control is imminent. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for adaptively controlling the high-force spinning of a metal component, comprising the following steps:

[0004] Step 1: Establish a virtual machining environment based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool, and the springback model;

[0005] Step 2: In a virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback processes of metal components. The processing state, processing quality and stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained through the preset evaluation function combined with the processing quality index.

[0006] Step 3: Taking feed rate, processing status and reward value as input, a mapping relationship is established through a deep learning algorithm, and then the optimization is performed through a reinforcement learning algorithm to update the deep learning strategy. After multiple trainings under different working conditions, a trained control model is obtained, and the optimized feed rate is output;

[0007] Step 4: Deploy the trained control model to the motion controller of the spinning machine. Through real-time data acquisition, the real-time processing status data of the spinning machine is obtained. Based on the real-time processing status data, the real-time control of the spinning machine is achieved through online updating and optimization of the feed rate.

[0008] Furthermore, the virtual machining environment is established based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool and the springback model, including:

[0009] According to the stress-strain relationship of metal materials during the spinning process, the constitutive model of metal materials is constructed; according to the vibration and stiffness of the machine tool, the dynamic characteristic model of the machine tool is established; a rebound model is established to predict the springback behavior of the workpiece after spinning; through the constitutive model of metal materials, the dynamic characteristic model of the machine tool and the springback model, a complete virtual processing environment is formed.

[0010] Furthermore, in the virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback process of metal components; the processing state, processing quality and stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained by combining the preset evaluation function with the processing quality index, including:

[0011] Through sensors in the virtual environment, various state parameters of metal components during the spinning process are monitored in real time to obtain processing simulation results; based on the simulation results, the processing quality of the metal components is evaluated to obtain processing quality evaluation results; based on the stress state of the metal components during the processing, the monitoring results of the low stress state are obtained; based on the processing state, the processing quality evaluation results and the monitoring results of the low stress state, the corresponding reward value is calculated.

[0012] Furthermore, the feed rate, processing state, and reward value are input, a mapping relationship is established through a deep learning algorithm, and a reinforcement learning algorithm is used to find the best, update the deep learning strategy, and after multiple trainings under different working conditions, a trained control model is obtained, and the optimized feed rate is output, including:

[0013] A deep learning algorithm is selected, and a deep learning model is constructed with feed rate, processing status, and reward value as input. The mapping relationship between input and output is learned through training data. The deep learning model is optimized using a reinforcement learning algorithm, and multiple trainings are performed under different working conditions to complete the training of the control model. The deep learning algorithm is one of a convolutional neural network and a recurrent neural network.

[0014] Furthermore, the trained control model is deployed to the motion controller of the spinning machine, and the real-time processing status data of the spinning machine is obtained through real-time data acquisition. According to the real-time processing status data, the real-time control of the spinning machine is realized by online updating and optimizing the feed rate, including:

[0015] The trained control model is deployed to the spinning machine motion controller. During the actual processing, the motion state of the spinning machine and the processing state of the metal components are collected in real time. The collected real-time data is input into the control model, and the feed rate is updated and optimized online according to the current processing state. The optimized feed rate is output to the spinning machine motion controller to realize real-time spinning machine motion control.

[0016] A metal component strong spinning forming adaptive control system, using the metal component strong spinning forming adaptive control method, characterized in that it includes a spinning machine motion controller, a metal component strong spinning simulation module, a data acquisition module, a data processing module and a communication module;

[0017] The spinning machine motion controller, the metal component strong spinning simulation module, the data acquisition module and the communication module are respectively connected to the data processing module.

[0018] The beneficial effects of the present invention are: by simulating the processing of metal components, an optimized feed rate is obtained, the feed rate of the main process parameter of spin forming is automatically adjusted, the force of the spinning wheel on the metal forming is changed, the stress-strain state of the plastic deformation zone is improved, and thus the quality of the workpiece is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of a process flow of an adaptive control method for high-force spinning of metal components;

[0020] Figure 2 A schematic diagram of the principle of an adaptive control system for high-force spinning of metal components;

[0021] Figure 3 It is a schematic diagram of the implementation of an adaptive control method for high-force spinning forming of metal components. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0023] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0024] like Figure 1 As shown, a self-adaptive control method for high-force spinning forming of metal components comprises the following steps:

[0025] Step 1: Establish a virtual machining environment based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool, and the springback model;

[0026] Step 2: In a virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback processes of metal components. The processing state, processing quality and stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained through the preset evaluation function combined with the processing quality index.

[0027] Step 3: Taking feed rate, processing status, and reward value as input, a mapping relationship is established through a deep learning algorithm, and then the deep learning strategy is updated through a reinforcement learning algorithm. After multiple trainings under different working conditions, a trained control model is obtained, and the optimized feed rate is verified and output;

[0028] Step 4: Deploy the trained control model to the motion controller of the spinning machine. Through real-time data acquisition, the real-time processing status data of the spinning machine is obtained. Based on the real-time processing status data, the real-time control of the spinning machine is achieved through online updating and optimization of the feed rate.

[0029] The virtual machining environment is established based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool and the springback model, including:

[0030] According to the stress-strain relationship of metal materials during the spinning process, the constitutive model of metal materials is constructed; according to the vibration and stiffness of the machine tool, the dynamic characteristic model of the machine tool is established; a rebound model is established to predict the springback behavior of the workpiece after spinning; through the constitutive model of metal materials, the dynamic characteristic model of the machine tool and the springback model, a complete virtual processing environment is formed.

[0031] In the virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback process of metal components; the processing state, processing quality and stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained by combining the preset evaluation function with the processing quality index, including:

[0032] Through sensors in the virtual environment, various state parameters of metal components during the spinning process are monitored in real time to obtain processing simulation results; based on the simulation results, the processing quality of the metal components is evaluated to obtain processing quality evaluation results; based on the stress state of the metal components during the processing, the monitoring results of the low stress state are obtained; based on the processing state, the processing quality evaluation results and the monitoring results of the low stress state, the corresponding reward value is calculated.

[0033] The feed rate, processing state, and reward value are input, a mapping relationship is established through a deep learning algorithm, and the optimization is performed through a reinforcement learning algorithm, the deep learning strategy is updated, and a trained control model is obtained after multiple trainings under different working conditions, and the optimized feed rate is verified and output, including:

[0034] A deep learning algorithm is selected, and a deep learning model is constructed with feed rate, processing status, and reward value as input. The mapping relationship between input and output is learned through training data. The deep learning model is optimized using a reinforcement learning algorithm, and multiple trainings are performed under different working conditions to complete the training of the control model. The deep learning algorithm is one of a convolutional neural network and a recurrent neural network.

[0035] The trained control model is deployed to the motion controller of the spinning machine, and the real-time processing status data of the spinning machine is obtained through real-time data acquisition. According to the real-time processing status data, the real-time control of the spinning machine is realized by online updating and optimizing the feed rate, including:

[0036] The trained control model is deployed to the spinning machine motion controller. During the actual processing, the motion state of the spinning machine and the processing state of the metal components are collected in real time. The collected real-time data is input into the control model, and the feed rate is updated and optimized online according to the current processing state. The optimized feed rate is output to the spinning machine motion controller to realize real-time spinning machine motion control.

[0037] like Figure 2 As shown, a metal component strong spinning forming adaptive control system, using the metal component strong spinning forming adaptive control method, characterized in that it includes a spinning machine motion controller, a metal component strong spinning simulation module, a data acquisition module, a data processing module and a communication module;

[0038] The spinning machine motion controller, the metal component strong spinning simulation module, the data acquisition module and the communication module are respectively connected to the data processing module.

[0039] Specifically, the adaptive control method for high-force spinning of metal components provided by the present invention is aimed at the following two situations:

[0040] 1) The random vibration of the machine tool during the high-force spinning process causes the stress-strain state of the metal in the plastic deformation zone to change, resulting in the instability of the plastic flow of the metal during the forming process, which leads to forming defects such as non-adherence to the mold and local bulges, blank swing, and flange wrinkling that are easy to occur during the spinning process of metal components.

[0041] 2) The springback of the spinning workpiece after spinning causes the size of the workpiece to change. The control principle: Based on the precursor symptoms of instability defects in the strong spinning metal forming process, the main process parameter feed rate of spinning is automatically adjusted to change the force of the spinning wheel on the metal forming, improve the stress-strain state of the plastic deformation zone, and thus improve the quality of the workpiece.

[0042] The optimization control of the present invention is mainly composed of an offline control system and an online control system, and its process is as follows: Figure 3 Its main process is as follows:

[0043] (1) Offline control system design: First, in the offline learning stage, a virtual processing environment is established. In the virtual environment, the strong spinning, shearing and springback processes of metal components are simulated, and the actual processing conditions are integrated to reproduce the actual processing process of metal component spinning. The forming defects that are prone to occur during the metal component spinning process, such as non-adherence to the mold and local bulges, blank swing, and flange wrinkling, are presented. The processing state of the metal component corresponds to the processing quality, processing accuracy and low stress, and a reward value is obtained. The mapping relationship between the feed rate, processing state and reward value is established through deep learning, and the reinforcement learning algorithm is used to find the optimal solution and update the deep learning strategy. After multiple trainings under different working conditions, a well-learned control model is obtained to obtain an optimized feed rate.

[0044] (2) Online control system: The trained control model is deployed to the motion controller of the spinning machine. With the real-time processing status of the spinning machine, it is updated online and the feed rate is optimized in real time, thus achieving real-time control.

Claims

1. A self-adaptive control method for high-force spinning of metal components, characterized in that: The following steps are involved: Step 1: Establish a virtual machining environment based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool, and the springback model; Step 2: In a virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback processes of metal components. The processing state, processing quality and stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained through the preset evaluation function combined with the processing quality index. Step 3: Taking feed rate, processing status, and reward value as input, a mapping relationship is established through a deep learning algorithm, and then the deep learning strategy is updated through a reinforcement learning algorithm. After multiple trainings under different working conditions, a trained control model is obtained, and the optimized feed rate is verified and output; Step 4: Deploy the trained control model to the motion controller of the spinning machine. Through real-time data acquisition, the real-time processing status data of the spinning machine is obtained. Based on the real-time processing status data, the real-time control of the spinning machine is achieved through online updating and optimization of the feed rate.

2. The method for adaptively controlling the high-force spinning of metal components according to claim 1, characterized in that: The virtual machining environment is established based on the constitutive model of the metal material of the metal component, the dynamic characteristic model of the machine tool and the springback model, including: According to the stress-strain relationship of metal materials during the spinning process, the constitutive model of metal materials is constructed; according to the vibration and stiffness of the machine tool, the dynamic characteristic model of the machine tool is established; a rebound model is established to predict the springback behavior of the workpiece after spinning; through the constitutive model of metal materials, the dynamic characteristic model of the machine tool and the springback model, a complete virtual processing environment is formed.

3. The method for adaptively controlling the high-force spinning of metal components according to claim 2, characterized in that: In the virtual processing environment, the spinning process and forming defects of metal components are reproduced by simulating the strong spinning, shearing and springback process of metal components; the processing state, processing quality, processing accuracy and low stress state of metal components are monitored in real time by deploying sensor simulation modules, and the corresponding reward value is obtained by combining the preset evaluation function with the processing quality index, including: Through sensors in the virtual environment, various state parameters of metal components during the spinning process are monitored in real time to obtain processing simulation results; based on the simulation results, the processing quality of the metal components is evaluated to obtain processing quality evaluation results; based on the stress state of the metal components during the processing, the monitoring results of the low stress state are obtained; based on the processing state, the processing quality evaluation results and the monitoring results of the low stress state, the corresponding reward value is calculated.

4. The self-adaptive control method for high-force spinning of metal components according to claim 3, characterized in that: The feed rate, processing state, and reward value are input, a mapping relationship is established through a deep learning algorithm, and the optimization is performed through a reinforcement learning algorithm, the deep learning strategy is updated, and after multiple trainings under different working conditions, a trained control model is obtained, and the optimized feed rate is output, including: A deep learning algorithm is selected, and a deep learning model is constructed with feed rate, processing status, and reward value as input. The mapping relationship between input and output is learned through training data. The deep learning model is optimized using a reinforcement learning algorithm, and multiple trainings are performed under different working conditions to complete the training of the control model. The deep learning algorithm is one of a convolutional neural network and a recurrent neural network.

5. The self-adaptive control method for high-force spinning of metal components according to claim 4, characterized in that: The trained control model is deployed to the motion controller of the spinning machine, and the real-time processing status data of the spinning machine is obtained through real-time data acquisition. According to the real-time processing status data, the real-time control of the spinning machine is realized by online updating and optimizing the feed rate, including: The trained control model is deployed to the spinning machine motion controller. During the actual processing, the motion state of the spinning machine and the processing state of the metal components are collected in real time. The collected real-time data is input into the control model, and the feed rate is updated and optimized online according to the current processing state. The optimized feed rate is output to the spinning machine motion controller to realize real-time spinning machine motion control.

6. A self-adaptive control system for high-force spinning of metal components, using a self-adaptive control method for high-force spinning of metal components as claimed in any one of claims 1 to 5, characterized in that: It includes a spinning machine motion controller, a metal component high-force spinning simulation module, a data acquisition module, a data processing module and a communication module; The spinning machine motion controller, the metal component strong spinning simulation module, the data acquisition module and the communication module are respectively connected to the data processing module.

Images