Servo Feedforward Control Using Machine Learning Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing servo control systems require users to prepare multiple linear models for feedforward control, making adjustments cumbersome and time-consuming.
Innovation Solution
A machine learning device that acquires labels and input data from a servo control apparatus to build a learning model through supervised learning, enabling the generation of correction values for feedforward control without the need for manual linear model creation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If linear models are prepared manually for feedforward control, then control accuracy can be improved, but the complexity of operation and time required for adjustment increase significantly
Solution Approach 1:
The system performs self-learning by automatically acquiring input data and teacher data, then uses machine learning to autonomously generate the feedforward control model without requiring manual preparation of linear models by users
Solution Approach 2:
The manual mechanical process of preparing and adjusting linear models is replaced by an automated machine learning system that uses algorithms to automatically generate the feedforward control model from acquired data
2Adaptability or versatility
If multiple linear models are prepared for different control scenarios, then adaptability improves, but device complexity and preparation time increase
Solution Approach 1:
A single machine learning-based feedforward control model generation system can handle multiple control scenarios by acquiring and learning from diverse input data and teacher data pairs, eliminating the need for separate linear models for each scenario
Solution Approach 2:
The system automatically adapts to different control scenarios through self-learning from acquired data without requiring manual intervention to prepare separate models for each scenario
Data Source
AI summary
A machine learning device acquires, as a label, a command output by a servo control apparatus to a control target device so as to drive and control the control target device. The machine learning device acquires, as input data, an output of the control target device driven based on the command, and constructs a learning model relating to feedforward control for correcting the command, by performing supervised learning by use of a set of the label and the input data serving as teaching data.