Servo Feedforward Learning Sequence for Stable Position Error
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In servo control devices using feedforward control with multiple loops, the simultaneous learning of position and velocity feedforward control operations leads to increased information processing, causing interference and reduced accuracy due to variations in position errors.
Innovation Solution
A machine learning device optimizes the coefficients of feedforward calculation units sequentially, where one unit's optimized coefficients are used as a basis for the other, reducing information processing and stabilizing position errors by performing reinforcement learning on the inner loop calculations first.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simultaneous learning of position and velocity feedforward control operations is performed, then comprehensive control optimization is achieved, but information processing load increases and learning accuracy deteriorates due to interference
Solution Approach 1:
The patent segments the simultaneous learning process into sequential learning stages: first performing learning on the velocity feedforward calculation unit, then performing learning on the position feedforward calculation unit after the velocity unit is optimized. This segmentation reduces information processing load at each stage while achieving comprehensive control optimization through the combined effect of both learning phases.
Solution Approach 2:
The patent applies preliminary action by completing the velocity feedforward learning first before initiating position feedforward learning. The optimized velocity feedforward coefficients serve as a foundation for the subsequent position learning phase, reducing interference and information processing requirements while maintaining comprehensive control optimization.
2Reliability
If simultaneous learning of position and velocity feedforward control is performed, then both control aspects are optimized, but position error variation increases due to mutual interference
Solution Approach 1:
The patent segments the learning process to eliminate mutual interference between position and velocity learning. By learning velocity feedforward coefficients first and then position feedforward coefficients separately, the system achieves control optimization without the position error variations caused by simultaneous learning interference.
Solution Approach 2:
The patent performs velocity feedforward learning as a preliminary step before position feedforward learning. This preliminary action stabilizes the velocity control parameters first, creating a stable foundation that prevents position error variations during the subsequent position learning phase while maintaining comprehensive control optimization.
3Manufacturing precision
If sequential learning is performed with inner loop first, then information processing is reduced and position error stability is improved
Solution Approach 1:
The patent performs velocity feedforward learning as a preliminary action before position feedforward learning. This approach reduces information processing load and improves position error stability by eliminating mutual interference. Although learning is sequential rather than simultaneous, the structured two-phase approach achieves faster overall settling by preventing error variations during the learning process.
Data Source
AI summary
A machine learning device performs machine learning with respect to a servo control device including at least two feedforward calculation units among a position feedforward calculation unit configured to calculate a position feedforward term on the basis of a position command, a velocity feedforward calculation unit configured to calculate a velocity feedforward term on the basis of a position command, and a current feedforward calculation unit configured to calculate a current feedforward term on the basis of a position command. Machine learning related to the coefficients of a transfer function of one feedforward calculation unit among the at least two feedforward calculation units is performed earlier than machine learning related to the coefficients of a transfer function of the other feedforward calculation unit.