Machine Learning Current Gain Optimization for Motor Control

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for adjusting current gain parameters in motor control systems are inefficient, requiring manual fine-tuning and varying significantly between motors, leading to suboptimal performance and increased labor due to discrepancies in motor physical constants.

Innovation Solution

A machine learning apparatus that observes state variables such as overshoot, undershoot, and rise time in response to torque commands to optimize integral and proportional gain functions through a training data set, allowing for automatic adjustment of current gain parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual fine-tuning of current gain parameters is performed by observing step response or frequency response, then the parameter can be adjusted for each individual motor, but the adjustment takes time and labor and is not realistic for mass production

Engineering Contradiction:

Improvecurrent gain parameter optimizationVSAvoidadjustment time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent pre-calculates and stores optimal current gain parameters for multiple motors during a setup phase. When a motor is used, the system automatically selects the pre-calculated parameters based on motor identification, eliminating the need for time-consuming manual fine-tuning for each motor while maintaining optimization quality

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a mapping relationship between motor identification information and optimal current gain parameters. By copying pre-determined parameter sets associated with each motor type, the system achieves rapid parameter application without repeating the optimization process for each individual motor

Inventive Principle:

Principle #26Copying

2Ease of operation

If current gain parameters are calculated from physical constants, then the parameter setting is simple, but there is discrepancy between the calculated value and the optimum value due to variations in inductance

Engineering Contradiction:

Improveparameter setting simplicityVSAvoidparameter optimization accuracy

Core Design Contradiction:

Ease of operationVSManufacturing precision

Solution Approach 1:

The patent incorporates feedback mechanisms where the system observes the actual motor response (step response or frequency response) and uses this information to select or adjust current gain parameters. This feedback loop ensures that the parameters are both easy to apply and optimized for actual performance, bridging the gap between simplicity and accuracy

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent changes the approach from calculating parameters based solely on physical constants to selecting parameters based on observed motor characteristics. By changing the parameter selection criterion from theoretical calculation to empirical observation, the system achieves both ease of operation and manufacturing precision

Inventive Principle:

Principle #35Parameter changes

3Device complexity

If conventional PI control with fixed gain parameters is used, then the control system is simple, but the motor response is suboptimal and feed unevenness occurs

Engineering Contradiction:

Improvecontrol system complexityVSAvoidmotor response quality

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent transitions from fixed gain parameters to dynamically selected gain parameters. The system automatically adjusts the current gain parameters based on motor identification and observed characteristics, making the control system adaptive rather than static. This maintains relative simplicity while significantly improving motor response quality and eliminating feed unevenness

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS10416618B2Machine learning apparatus for learning gain optimization, motor control apparatus equipped with machine learning apparatus, and machine learning method

Publication Date: 2019.09.17 FANUC LTD

AI summary

A machine learning apparatus according to one embodiment of the present invention is a machine learning apparatus for learning a condition associated with adjustment of a current gain parameter in electrical machine control, and comprises: a state observing unit which acquires actual current as well as an integral gain function and a proportional gain function in a current control loop, and observes state variables which include the integral gain function, the proportional gain function, and at least one of an amount of overshoot, an amount of undershoot, and a rise time of the actual current occurring in response to a step-like torque command; and a learning unit which learns the condition associated with the adjustment of the current gain parameter in accordance with a training data set constructed from the state variables.