Manufacturing Inference Model Tuning Across Similar Processes

Solution Overview

Problem

Existing inference techniques in manufacturing processes require costly and time-consuming model optimization for each process, limiting the applicability and precision of high-precision inference across different manufacturing processes.

Innovation Solution

An inference device and method that utilizes a plurality of machine-learned network sections to process time series data, incorporating a fine-tuning function with correction parameters to adjust for errors caused by individual process differences, allowing for high-precision inference across similar manufacturing processes without re-optimization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If a model is optimized on a process-by-process basis to realize more precise inference, then manufacturing precision is improved, but loss of time and loss of energy increase

Engineering Contradiction:

Improveinference precisionVSAvoidmodel optimization time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent creates a universal inference model that can be applied across multiple manufacturing processes of the same type. The model is trained on aggregated time series data from multiple processes and uses a standardized architecture that can infer states and abnormalities across different processes without requiring separate optimization for each process, thereby reducing time and energy costs while maintaining precision.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent performs preliminary training of the inference model using aggregated data from multiple manufacturing processes before deployment. This preliminary action establishes a robust base model that captures common patterns across processes, eliminating the need for time-consuming process-specific optimization later while maintaining high inference precision.

Inventive Principle:

Principle #10Preliminary action

2Manufacturing precision

If a model is optimized on a process-by-process basis to realize more precise inference, then manufacturing precision is improved, but loss of energy increases

Engineering Contradiction:

Improveinference precisionVSAvoidmodel optimization energy

Core Design Contradiction:

Manufacturing precisionVSLoss of energy

Solution Approach 1:

The patent creates a universal inference model that can be applied across multiple manufacturing processes of the same type. The model is trained on aggregated time series data from multiple processes and uses a standardized architecture that can infer states and abnormalities across different processes without requiring separate optimization for each process, thereby reducing time and energy costs while maintaining precision.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges data from multiple manufacturing processes into a unified training dataset and combines the inference capabilities into a single model architecture. This merging approach consolidates the computational resources needed for model optimization, reducing the total energy consumption compared to training separate models for each process while maintaining comprehensive coverage of process variations.

Inventive Principle:

Principle #5Merging (Combining)

3Device complexity

If a model that achieves high-precision inference for a specific process is applied to other processes of the same type, then device complexity is reduced, but manufacturing precision deteriorates

Engineering Contradiction:

Improvemodel complexityVSAvoidinference precision

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent creates a universal inference model that can be applied across multiple manufacturing processes of the same type. The model is trained on aggregated time series data from multiple processes and uses a standardized architecture that can infer states and abnormalities across different processes without requiring separate optimization for each process, thereby reducing time and energy costs while maintaining precision.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent employs parameter adjustment mechanisms that allow the universal model to adapt to specific process characteristics through correction parameters. These parameters are learned during training on aggregated data and enable the model to maintain high precision across different processes by dynamically adjusting its behavior based on process-specific patterns while maintaining a single unified model structure.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20230004837A1Inference device, inference method and inference program

Publication Date: 2023.01.05 TOKYO ELECTRON LTD

AI summary

An inference device, an inference method, and an inference program that can realize high precision inference regardless of an application target are provided. An inference device includes: an acquisition section configured to acquire a time series data group measured in accordance with processing of a target object in a predetermined processing unit of a manufacturing process; and an inference section configured to tune respective output data that is output by processing the acquired time series data group using a plurality of network sections that have been machine-learned in advance and to output an inference result by combining the respective tuned output data; wherein the inference section is configured to tune the respective output data using a correction parameter corresponding to an error included in the inference result.