Press line tooling track automatic generation method and system based on deep learning

Abstract

The invention discloses a deep learning-based press line tooling track automatic generation method and system, and mainly relates to the field of press line automatic processing. Comprising a data acquisition module, a model training module, a real-time track generation module and a track optimization module. The tooling track generation method has the beneficial effects that the technical problems of low efficiency and insufficient accuracy of a method for manually building and adjusting the tooling track can be solved, the tooling track generation efficiency and accuracy are improved, and the requirement for automatic and efficient production of a stamping production line is met.

Description

Technical Field

[0001] The present invention relates to the field of press line production automation technology, and in particular to a method and system for automatically generating a press line end picker trajectory based on deep learning. Background Art

[0002] In stamping production lines, end pickers serve as important tools for grasping and handling workpieces. The rationality of their trajectory planning directly affects the efficiency and quality of stamping production. At present, setting up end pickers and adjusting the end picker trajectory at offline workstations usually rely on manual operation. There are many problems with manual setting up and adjusting the end picker trajectory: on the one hand, manual operation is inefficient and requires a lot of time and manpower costs; on the other hand, the accuracy and consistency of manual adjustment are difficult to guarantee and are easily affected by the operator's experience and skill level, resulting in unreasonable end picker trajectory planning, which in turn affects the stability of stamping production and product quality. With the continuous improvement of stamping production's requirements for automation and intelligence, the traditional method of manually setting up and adjusting the end picker trajectory can no longer meet the needs of actual production. Therefore, there is an urgent need for a method that can automatically generate end picker trajectories to meet the needs of efficient production. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for automatically generating the end picker trajectory of a stamping line based on deep learning. It can solve the technical problems of low efficiency and insufficient accuracy in the manual method of building and adjusting the end picker trajectory, improve the efficiency and accuracy of end picker trajectory generation, and meet the needs of automated and efficient production of stamping production lines.

[0004] To achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions: The method for automatically generating the trajectory of the end picker of a stamping line based on deep learning includes the following steps: S1. Data acquisition: collect data related to the end picker trajectory during the stamping production process, and perform cleaning and normalization preprocessing on the collected data; S2. Model training: Build a deep learning model. Divide the preprocessed data into a training set and a test set. Use the training set to train the deep learning model. Use the end picker trajectory data as input and the corresponding production result data as a supervisory signal. Train the deep learning model using the backpropagation algorithm. Evaluate the model on the test set and use cross-validation to optimize the model hyperparameters. S3, real-time trajectory generation, obtains the information of the current workpiece to be processed and the real-time parameters of the stamping equipment, inputs them into the trained deep learning model, and generates the end picker trajectory; S4, trajectory optimization, uses the optimization algorithm to optimize the generated end picker trajectory, and then adds the motion state of the end picker and the constraints to avoid collision to obtain the final end picker trajectory.

[0005] Furthermore, the data related to the end picker trajectory includes workpiece information, parameters of the stamping equipment, historical end picker trajectory data and corresponding production result data.

[0006] Furthermore, the workpiece information includes the shape, size, and material of the workpiece.

[0007] Furthermore, the production result data includes forming quality detection data of whether the workpiece has cracks or deformation.

[0008] Furthermore, the deep learning model includes convolutional neural networks, recurrent neural networks, long short-term memory networks, and gated recurrent units.

[0009] Furthermore, the data ratio of the training set and the test set is 7:3.

[0010] Furthermore, the optimization algorithm is a genetic algorithm or a particle swarm optimization algorithm.

[0011] Furthermore, the motion state of the end effector includes the motion speed and acceleration limits of the end effector.

[0012] A deep learning-based automatic trajectory generation system for end pickers on stamping lines, including a data acquisition module, a model training module, a real-time trajectory generation module, and a trajectory optimization module; The data acquisition module is used to collect data related to the end picker trajectory during the stamping production process and pre-process the data; Model training module, used to build deep learning models and train and optimize the models using preprocessed data; The real-time trajectory generation module is used to obtain information about the workpiece to be processed and the stamping equipment, and input it into the trained deep learning model to generate the end effector trajectory; The trajectory optimization module is used to optimize the end picker trajectory generated by the optimization algorithm to obtain the final end picker trajectory.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention automatically generates the end-picker trajectory through deep learning technology, replacing the traditional manual construction and adjustment method, greatly improving the efficiency of end-picker trajectory generation and reducing labor and time costs; 2. The deep learning model can learn the complex mapping relationship between the end-picker trajectory and production results from a large amount of historical data. Compared with manual adjustment, the generated end-picker trajectory is more reasonable and accurate, which can effectively improve the stability of stamping production and product quality; 3. The system can automatically generate and optimize trajectories based on the real-time information of the workpiece to be processed and the stamping equipment. It has strong adaptability and flexibility, can meet the stamping production needs under different working conditions, and improve the intelligence level of the stamping production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Attachment Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0015] Below in conjunction with specific embodiment, further set forth the present invention.Should be understood that these embodiments are only used to illustrate the present invention and are not used in limiting the scope of the present invention.In addition, should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall equally within the scope limited by the application.

[0016] Reference Figure 1 The present invention provides a method and system for automatically generating a stamping line end picker trajectory based on deep learning, wherein the method for automatically generating a stamping line end picker trajectory based on deep learning includes the following steps: S1. Data acquisition: Collect data related to the end-picker trajectory during the stamping production process, and perform cleaning and normalization preprocessing on the collected data. The data related to the end-picker trajectory includes parameter information of the workpiece itself, equipment data related to the stamping process, and historical data. The collected data is preprocessed, including data cleaning and normalization operations, to remove abnormal data and map the data to an appropriate range, thereby improving data quality and usability; S2. Model training: Build a deep learning model. Divide the preprocessed data into a training set and a test set. Use the training set to train the deep learning model. Use data related to the end picker trajectory as input and the corresponding production result data as a supervisory signal. Train the deep learning model through the backpropagation algorithm, and continuously adjust the model parameters so that the model can learn the mapping relationship between the end picker trajectory and the production results. Evaluate the model on the test set and use cross-validation to optimize the model hyperparameters until the model can accurately predict the relationship between the end picker trajectory and the production results, so that the trained deep learning model has good generalization ability. S3, real-time trajectory generation, obtains information about the workpiece to be processed and the real-time parameters of the stamping equipment, and inputs them into the trained deep learning model. The deep learning model automatically generates the end picker trajectory suitable for the current working conditions based on the learned mapping relationship; S4. Trajectory optimization: Use the optimization algorithm to optimize the generated end picker trajectory, then add the motion state of the end picker and the constraints to avoid collisions, fully adapt to the actual motion state of the end picker, avoid collisions between the end picker and the stamping equipment or workpiece, and use the optimization algorithm to continuously adjust the trajectory to obtain the final end picker trajectory.

[0017] Preferably, the data related to the end picker trajectory includes the information of the workpiece, the parameters of the stamping equipment, the historical end picker trajectory data and the corresponding production result data. Changes in the workpiece information and the parameter data of the stamping equipment can cause significant changes in the trajectory of the end picker, thereby affecting the results of the stamping process. Therefore, after the above data is collected, it can accurately reflect the state of the workpiece, thereby forming a mapping relationship that is more matched with the production results, ensuring the accuracy of subsequent deep learning model training. The historical end picker trajectory data and the corresponding production result data can be used as a reference for the mapping relationship to improve the efficiency and accuracy of subsequent training and learning.

[0018] Preferably, the workpiece information includes the shape, size, and material of the workpiece. The above data can fully reflect the three-dimensional information and type of the workpiece, and provide more accurate and efficient data support for the processing of different models of automobile covers.

[0019] Preferably, the production result data includes forming quality detection data of whether the workpiece has cracks or deformations. Cracks or deformations can fully reflect the quality of the stamping process of the workpiece and provide accurate data support for the results of the stamping process.

[0020] Preferably, the deep learning model includes a convolutional neural network, a recurrent neural network, a long short-term memory network, and a gated recurrent unit. The above deep learning models are relatively mature, have good training and reasoning methods, have fewer parameters and stronger expressiveness when processing high-dimensional data, and are widely used in image recognition, target detection and other fields.

[0021] Preferably, the data ratio of the training set and the test set is 7:3, which not only ensures that there is sufficient data for training, but also that there is sufficient data for testing and evaluation after multiple training and optimization, so that the trained model can accurately predict the relationship between the end picker trajectory and the production results, and has good generalization ability.

[0022] Preferably, the optimization algorithm is a genetic algorithm or a particle swarm optimization algorithm. When solving more complex combinatorial optimization problems, this algorithm can usually obtain better optimization results faster than some conventional optimization algorithms. It has been widely used in combinatorial optimization, machine learning, signal processing, adaptive control, artificial life and other fields.

[0023] Preferably, the motion state of the end picker includes the motion speed and acceleration limit of the end picker. The motion speed and acceleration limit can provide comprehensive feedback on the motion state of the end picker, so that the movement of the end picker is not easily interfered with the surrounding equipment and workpiece, thereby ensuring the accuracy of the motion trajectory of the end picker.

[0024] A deep learning-based automatic trajectory generation system for end pickers on stamping lines, including a data acquisition module, a model training module, a real-time trajectory generation module, and a trajectory optimization module; The data acquisition module is used to collect data related to the end picker trajectory during the stamping production process and pre-process the data; Model training module, used to build deep learning models and train and optimize the models using preprocessed data; The real-time trajectory generation module is used to obtain information about the workpiece to be processed and the stamping equipment, and input it into the trained deep learning model to generate the end effector trajectory; The trajectory optimization module is used to optimize the end picker trajectory generated by the optimization algorithm to obtain the final end picker trajectory.

[0025] Example: First, the data acquisition module collected nearly a year's worth of data related to the tooling trajectory of the production line. This included 3D models of various automotive panels (workpieces) (for obtaining shape and size information), material parameters, press parameters such as pressure, stroke, and speed, historical tooling trajectory coordinate data, and corresponding workpiece forming quality inspection data (e.g., presence of cracks, and whether deformation met requirements). This data was cleaned to remove abnormal data, then normalized and mapped to an appropriate range. Next, the model training module constructed a deep learning model based on a long short-term memory (LSTM) network. The preprocessed data was divided into a training set and a test set in a 7:3 ratio. The model was trained using the training set, with appropriate hyperparameters such as the learning rate and number of iterations set during the training process. After multiple training and optimization steps, the model was evaluated on the test set. The model was able to accurately predict the relationship between the end picker trajectory and production results, demonstrating good generalization capabilities. In actual production, when a new automotive panel needs to be stamped, the real-time trajectory generation module obtains the panel's 3D model and material information, as well as the current operating parameters of the stamping equipment. This information is input into the trained LSTM model, which quickly generates the initial end effector trajectory. Finally, the trajectory optimization module uses a genetic algorithm to optimize the initial trajectory, taking into account the speed and acceleration limits of the end effector during movement, as well as the risk of collision with the stamping die, workpiece, and other factors. After multiple iterations, the final end effector trajectory is obtained. Applying this trajectory to stamping production has increased production efficiency by 30% compared to traditional manual trajectory adjustment, and the workpiece forming qualification rate has increased from 85% to 95%, achieving significant economic and technical benefits.

Claims

1. A method for automatically generating the trajectory of the end picker of a stamping line based on deep learning, characterized by , including the following steps: S1. Data acquisition: collect data related to the end picker trajectory during the stamping production process, and perform cleaning and normalization preprocessing on the collected data; S2. Model training: Build a deep learning model. Divide the preprocessed data into a training set and a test set. Use the training set to train the deep learning model. Use the end picker trajectory data as input and the corresponding production result data as a supervisory signal. Train the deep learning model using the backpropagation algorithm. Evaluate the model on the test set and use cross-validation to optimize the model hyperparameters. S3, real-time trajectory generation, obtains the information of the current workpiece to be processed and the real-time parameters of the stamping equipment, inputs them into the trained deep learning model, and generates the end picker trajectory; S4, trajectory optimization, uses the optimization algorithm to optimize the generated end picker trajectory, and then adds the motion state of the end picker and the constraints to avoid collision to obtain the final end picker trajectory.

2. The method for automatically generating a press line end pick-up trajectory based on deep learning according to claim 1, characterized in that: The data related to the end picker trajectory includes workpiece information, stamping equipment parameters, historical end picker trajectory data and corresponding production result data.

3. The method for automatically generating the trajectory of a press line end pick-up based on deep learning according to claim 2, characterized in that: The workpiece information includes the shape, size, and material of the workpiece.

4. The method for automatically generating a stamping line end pick-up trajectory based on deep learning according to claim 2, characterized in that: The production result data includes forming quality detection data of whether the workpiece has cracks or deformation.

5. The method for automatically generating the trajectory of a press line end pick-up based on deep learning according to claim 1, characterized in that: The deep learning model includes convolutional neural networks, recurrent neural networks, long short-term memory networks, and gated recurrent units.

6. The method for automatically generating the trajectory of a press line end pick-up based on deep learning according to claim 1, characterized in that: The data ratio of the training set and the test set is 7:

3.

7. The method for automatically generating a stamping line end pick-up trajectory based on deep learning according to claim 1, characterized in that: The optimization algorithm is a genetic algorithm or a particle swarm optimization algorithm.

8. The method for automatically generating the trajectory of a stamping line end pick-up based on deep learning according to claim 1, characterized in that: The motion state of the end effector includes the motion speed and acceleration limit of the end effector.

9. A deep learning-based automatic generation system for the trajectory of the end pick-up of a stamping line, characterized by: Including data acquisition module, model training module, real-time trajectory generation module, and trajectory optimization module; The data acquisition module is used to collect data related to the end picker trajectory during the stamping production process and pre-process the data; Model training module, used to build deep learning models and train and optimize the models using preprocessed data; The real-time trajectory generation module is used to obtain information about the workpiece to be processed and the stamping equipment, and input it into the trained deep learning model to generate the end effector trajectory; The trajectory optimization module is used to optimize the end picker trajectory generated by the optimization algorithm to obtain the final end picker trajectory.

Images