Servo Feedforward IIR Filter Tuning for Faster Settling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing servo control devices using feedforward control with IIR filters do not effectively optimize the coefficients of the transfer function, leading to prolonged settling times in machine learning processes for servo motors in machine tools and industrial machines.
Innovation Solution
A machine learning device that represents the zero-point and pole of the IIR filter's transfer function in polar coordinates, searches within a predetermined range to optimize the coefficients, utilizing reinforcement learning and Q-learning to adjust the coefficients of the IIR filter, thereby shortening the settling time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional coefficient optimization methods are used for IIR filter in servo control, then the control stability is maintained, but the settling time of machine learning is prolonged
Solution Approach 1:
The patent transforms the coefficient optimization problem from Cartesian coordinates to polar coordinates, representing the transfer function parameters as radius r and angle θ. This parameter transformation enables more efficient search and learning within a predetermined range, significantly reducing the settling time while maintaining control stability through the structured optimization approach.
2Measurement precision
If the search range for coefficient optimization is expanded to improve accuracy, then the optimization precision is improved, but the computing time and complexity increase
Solution Approach 1:
The patent divides the coefficient optimization into two distinct phases: a search phase that explores a predetermined range to identify promising regions, and a learning phase that refines the coefficients within those regions. This segmentation allows the system to achieve high optimization precision without requiring exhaustive search across the entire parameter space, thereby reducing computing complexity.
Solution Approach 2:
The patent performs preliminary search within a predetermined range before executing the main learning process. This preliminary action identifies optimal search regions and narrows down the parameter space, allowing subsequent learning to focus computational resources on the most promising areas, thus achieving high precision with reduced overall computing complexity.
Data Source
AI summary
The settling time of machine learning is shortened. A machine learning device is configured to perform machine learning related to optimization of coefficients of a transfer function of an IIR filter of a feedforward calculation unit with respect to a servo control device configured to control a servo motor configured to drive an axis of a machine tool, a robot, or an industrial machine using feedforward control by a feedforward calculation unit having the IIR filter. The machine learning device represents a zero-point at which the transfer function of the IIR filter is zero and a pole at which the transfer function diverges infinitely in polar coordinates using a radius r and an angle θ, respectively, and searches for and learns, within a predetermined search range, the radius r and the angle θ to thereby perform the optimization of the coefficients of the transfer function of the IIR filter.