Adjustment Device for Multi-Axis Motor Control Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning devices currently require extensive time for training when used with machining programs, making them unsuitable for processing small quantities of varied items, and there is a need for a more efficient method to control learning processes.
Innovation Solution
An adjustment device that includes a start command output unit, feedback information acquisition and transmission units, parameter setting information acquisition and transmission units, and a machine learning device configured to perform reinforcement learning using an evaluation program, allowing for controlled learning and parameter adjustment in a control device that drives multiple axes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is performed using a machining program, then the learning accuracy is improved, but the learning time increases significantly
Solution Approach 1:
The patent segments the learning program into two distinct types: evaluation programs for rapid parameter tuning and machining programs for final precision learning. This segmentation allows the system to perform quick iterations with evaluation programs and then finalize with a machining program, resolving the contradiction between learning speed and accuracy.
Solution Approach 2:
The patent implements preliminary action by using evaluation programs to perform preliminary learning and parameter optimization before executing the actual machining program. This preliminary learning phase prepares the control parameters in advance, so that when the machining program runs, the system is already optimized and requires less learning time.
2Adaptability or versatility
If machine learning is performed every time a workpiece varies, then the adaptability is improved, but the productivity decreases
Solution Approach 1:
The patent implements dynamics by creating a hybrid learning system that adapts its behavior based on the situation: it performs rapid evaluation program learning when workpieces change to maintain adaptability, and switches to standard machining programs for production to maintain productivity. The system dynamically adjusts the learning frequency and program type.
Solution Approach 2:
The patent uses parameter changes by modifying control parameters through evaluation programs when workpiece variations are detected. Instead of performing full machine learning every time, the system changes specific control parameters based on evaluation results, maintaining adaptability while minimizing the time spent on parameter adjustment.
3Loss of time
If an evaluation program is used instead of a machining program, then the learning time is reduced, but the measurement precision for machining tool evaluation decreases
Solution Approach 1:
The patent introduces an intermediary approach by using evaluation programs that are specifically designed to bridge the gap between rapid learning and accurate evaluation. These evaluation programs contain optimized motion commands and evaluation points that provide sufficient precision for parameter tuning without requiring the full complexity and time of actual machining programs.
Data Source
AI summary
An adjustment device that controls a control device controlling motors driving at least two axes, and a machine learning device performing machine learning with respect to the control device. The adjustment device includes: a start command output unit configured to output a start command for starting the machine learning device; a feedback information. acquisition unit configured to acquire feedback information acquired on the basis of an evaluation program executed by the control device, from the control device; a feedback information transmission unit configured to transmit the feedback information acquired, to the machine learning device; a parameter setting information acquisition unit configured to acquire control parameter setting information acquired by machine learning using the feedback information, from the machine learning device; and a parameter setting information transmission unit configured to transmit the control parameter setting information acquired, to the control device.