A Machining Deformation Control Method Based on Meta Reinforcement Learning

Abstract

A processing deformation control method based on meta-reinforcement learning, which is characterized in that the processing deformation control process optimization of each part in different groups of source data is regarded as a task, a reinforcement learning model is established for each task, and the workpiece processing is divided into For several processing steps, the processing state of the workpiece is taken as the state, the process selection of the next processing step is used as the action, and the next processing state and the subsequent processing state are used as the basis for designing the reward function; based on the meta-learning method, each reinforcement learning model is used as The base model, through the collaborative training of the base model and the meta-model through the source data; when faced with a new processing task, fine-tuning the meta-model through a small amount of sample data of the new task to obtain a reinforcement learning model adapted to the optimization of the machining deformation control process for the new task. The invention improves the effect of deformation control, can realize online optimization of processing technology, and reduces the demand for actual processing technology data.

Description

technical field [0001] The invention relates to the field of numerical control machining, in particular to a method for controlling deformation of numerical control machining of parts, in particular to a method for controlling deformation of machining based on meta-intensive learning. Background technique [0002] Machining deformation is one of the main reasons for the quality problems of parts, so the control of machining deformation has also become an important aspect to ensure the quality of parts. Process optimization during the machining process is an effective way to control the machining deformation and plays an important role in the final deformation of the part. Although the deformation after machining can be handled by shape correction, the shape correction process is complicated and the workload is large, which is easy to cause workpiece cracking. The control accuracy is still difficult to meet the requirements of high-precision deformation control. Reducing the ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the problem of deformation control process optimization, and propose a processing deformation control method based on meta-reinforcement learning, establish a reinforcement learning base model for different deformation control process data, and use monitoring data during workpiece processing as labels , the number of sample data is increased, and the deformation control process simulation data can be obtained through the processing simulation environment, and the sample data is further increased. Using the small sample learning method of meta-learning, the base model is trained through the data generated by interacting with the workpiece processing environment and iteratively. Obtain a meta-model, use the monitoring data of the new processing task as sample data, fine-tune the meta-model to adapt to the processing of the new task through the small amount of sample data, and improve the generalization ability and deformation control effect of the model

Images