Servo Control Disturbance Suppression via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing servo control devices require complex adjustments to suppress disturbances, making it difficult for users to implement effective disturbance suppression in machine learning applications.
Innovation Solution
A machine learning device that acquires labels and input data to construct a learning model for estimating currents to drive control targets, using supervised learning to generate functions that compensate for disturbances, thereby simplifying the suppression of disturbances in servo control systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a linear model is used as a reference model for adaptive control, then disturbance suppression capability is improved, but the complexity of user preparation and adjustment increases
Solution Approach 1:
The system performs self-learning by automatically constructing the linear model through supervised learning using training data consisting of input-output pairs collected during normal operation. The learning unit autonomously adjusts the model parameters without requiring user intervention, making the system self-configuring and eliminating the need for complex manual preparation while maintaining disturbance suppression capability
Solution Approach 2:
The system collects training data during normal operation and performs learning in advance to establish the linear model before actual disturbance suppression is needed. This preliminary learning phase allows the system to be fully prepared and configured ahead of time, so that when disturbances occur, the model is already ready to provide accurate compensation without requiring real-time user adjustment
2Measurement precision
If supervised learning is performed with labels and input data to construct a learning model, then disturbance compensation accuracy is improved, but the complexity of the learning process increases
Solution Approach 1:
The learning unit uses a universal supervised learning framework that can handle various types of input data and generate appropriate learning models for different disturbance scenarios. The same learning mechanism works for different control targets and disturbance types, providing accurate compensation across multiple applications without requiring separate complex learning processes for each case
Solution Approach 2:
The system uses the output of the learning model to generate disturbance compensation signals that are fed back to the control target. The model continuously learns from the relationship between input data and actual system responses, using feedback mechanisms to refine its predictions and improve compensation accuracy over time while maintaining a manageable learning process through iterative optimization
Data Source
AI summary
A machine learning device includes a label acquisition unit for acquiring, as a label, a current command that drives a control target of a servo control device in a state in which disturbance is suppressed; an input data acquisition unit for acquiring, as input data, a velocity of the control target driven based on the current command in the state in which disturbance is suppressed; and a learning unit for constructing a learning model for estimating a current to drive the control target from the velocity of the control target, by way of performing supervised leaning with a group of the label and the input data as training data.