Deposition Condition Selection Using Confidence-Bounded Regression

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing deposition prediction models, particularly nonlinear regression models, face challenges with global prediction accuracy due to biased samples or a shortage of data, leading to significant deviations between anticipated and actual deposition results.

Innovation Solution

A deposition-condition output device that calculates deposition conditions using both linear and nonlinear regression models, with the latter estimating a confidence interval, and selects the optimal condition based on whether the confidence interval meets a predetermined condition to ensure accurate predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a nonlinear regression model is used to optimize deposition conditions, then local prediction accuracy is improved, but global prediction accuracy deteriorates due to biased samples or shortage of samples

Engineering Contradiction:

Improvelocal prediction accuracyVSAvoidglobal prediction accuracy

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent combines linear regression models and nonlinear regression models into a unified prediction system. The linear model provides stable global prediction performance, while the nonlinear model captures local complex relationships. By merging both models and selecting predictions based on confidence intervals, the system achieves both local accuracy and global reliability that neither model could achieve alone.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the parameter of model selection from static to dynamic by introducing confidence interval evaluation. Instead of always using the nonlinear model for local accuracy, the system dynamically adjusts between linear and nonlinear models based on the calculated confidence interval, thereby adapting to data quality variations and maintaining global prediction reliability.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If a nonlinear regression model is used for deposition optimization, then flexibility in modeling complicated regression curves is improved, but prediction reliability deteriorates when samples are biased or insufficient

Engineering Contradiction:

Improvemodeling flexibilityVSAvoidprediction reliability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent merges the flexibility of nonlinear regression modeling with the reliability of linear regression by implementing a hybrid system. The nonlinear model component provides the needed flexibility for complicated regression curves, while the linear model component and confidence interval mechanism ensure prediction reliability when samples are biased or insufficient.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces feedback through confidence interval evaluation of the nonlinear model's predictions. When the confidence interval indicates low reliability (due to biased or insufficient samples), the system feedbacks to switch to the linear model, thereby maintaining prediction reliability while preserving modeling flexibility when conditions are favorable.

Inventive Principle:

Principle #23Feedback

3Stability of the object's composition

If only a linear regression model is used, then global prediction stability is maintained, but local prediction accuracy for complicated regression curves deteriorates

Engineering Contradiction:

Improveglobal prediction stabilityVSAvoidlocal prediction accuracy

Core Design Contradiction:

Stability of the object's compositionVSMeasurement precision

Solution Approach 1:

The patent merges linear and nonlinear regression models to simultaneously achieve global prediction stability and local prediction accuracy. The linear model ensures stable global predictions, while the nonlinear model captures local complex patterns. The confidence interval-based selection mechanism enables the system to leverage the strengths of both models appropriately.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent transforms the static linear model approach into a dynamic system that adapts between linear and nonlinear modeling based on local conditions. By dynamically selecting the appropriate model type based on confidence interval evaluation, the system maintains global stability while achieving local accuracy where needed.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12148600B2Deposition-condition output device, method for outputting deposition condition, recording medium and depositon apparatus

Publication Date: 2024.11.19 TOKYO ELECTRON LTD

AI summary

A deposition-condition output device includes a first calculating unit that calculates a first deposition condition under which a target result of deposition is obtained, based on a liner regression model applied to a deposition process by a deposition apparatus. The deposition-condition output device includes a second calculating unit that calculates a second deposition condition under which the target result of the deposition is obtained, based on a nonlinear regression model applied to the deposition process by the deposition apparatus, the second deposition condition being calculated by estimating a confidence interval of a predicted result of the deposition. The deposition-condition output device includes a selector that selects either the first deposition condition or the second deposition condition, based on whether the confidence interval of the predicted result estimated under the calculated second deposition condition satisfies a predetermined condition.