Thermal Displacement Model Hyperparameter Optimization for Machine Tools

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for optimizing thermal displacement models in machine tools are inefficient due to the vast number of hyperparameter combinations, making it impractical to manually determine optimal values, and requiring an enormous amount of time for trial and error.

Innovation Solution

An information processing device that automatically optimizes hyperparameters in thermal displacement models by selecting and optimizing one hyperparameter at a time, while fixing the value of another hyperparameter, using machine learning to generate and evaluate thermal displacement prediction formulas.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all combinations of hyperparameters are tested to find the optimal thermal displacement model, then the accuracy of thermal displacement compensation is improved, but the time required for optimization becomes excessively long

Engineering Contradiction:

Improveaccuracy of thermal displacement compensationVSAvoidtime required for optimization

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the hyperparameter optimization process into multiple stages: first optimizing time-delay related hyperparameters, then sampling interval hyperparameters, and finally other hyperparameters. This segmentation divides the overwhelming task of optimizing all hyperparameter combinations into manageable stages, reducing the total optimization time while maintaining accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary optimization of certain hyperparameters (time-delay and sampling interval) before optimizing other hyperparameters. This preliminary action establishes a foundation that guides subsequent optimization steps, preventing wasted computational effort on suboptimal configurations and significantly reducing total optimization time.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If manual trial and error is used to determine hyperparameter values, then flexibility in adjusting parameters is maintained, but the process becomes inefficient and cannot uniquely determine optimal values

Engineering Contradiction:

Improveflexibility in parameter adjustmentVSAvoidefficiency of optimization process

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent implements automated feedback mechanisms where the system evaluates the performance of thermal displacement models based on measured data, automatically adjusts hyperparameters according to evaluation results, and iteratively improves the model. This feedback-driven approach replaces manual trial and error with an efficient automated process that can uniquely determine optimal values while maintaining adaptability.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs self-optimization by automatically selecting and adjusting hyperparameter values based on performance evaluation without requiring manual intervention. The automated hyperparameter optimization unit independently carries out the optimization process, significantly improving productivity while maintaining the flexibility to adapt to different machining conditions.

Inventive Principle:

Principle #25Self-service

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach allows for the automatic optimization of hyperparameters in thermal displacement models, significantly reducing the time and effort required to achieve high-accuracy thermal displacement compensation in machine tools.

Implementation Method 1

performing machine learning for, based on a measurement data group including temperature data regarding a mechanical element that undergoes thermal expansion and is in a machine tool and regarding a periphery of the mechanical element and/or operating state data regarding the mechanical element, estimating an amount of thermal displacement of the mechanical element

Methodology Applied

Scientific EffectMachine learning:

Implementation Method 2

due to heat generation in accordance with rotation by the main shaft, heat generation by a main-shaft drive motor, the abovementioned components—in particular mainly the main shaft—thermally deform

Methodology Applied

Scientific EffectThermal expansion: Thermal Expansion

Data Source

PatentUS12314026B2Information processing device, control device, and optimization method

Publication Date: 2025.05.27 FANUC LTD

AI summary

An information processing device includes a parameter selection method determination unit that selects a hyperparameter of a thermal displacement amount prediction formula as a first hyperparameter, a parameter selection unit that sets/changes the value of the first hyperparameter and fixes remaining hyperparameter values as second hyperparameter values, a machine learning unit that generates a thermal displacement amount prediction formula for each value of the first hyperparameter based on thermal displacement teacher data, and a model evaluation unit that stores the first hyperparameter values together with evaluation values which are the difference between the thermal displacement amounts estimated by each thermal displacement amount prediction formula and measured thermal displacement amounts. The parameter selection method determination unit uses a history of the values of the first hyperparameter and the evaluation values to set the value of the first hyperparameter when the evaluation value is smallest as an optimal value.