Automated Model Building for Semiconductor Process Monitoring

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In the semiconductor manufacturing industry, creating and updating time-trajectory specifications for manufacturing processes is labor-intensive and prone to human error, requiring significant expertise and time, especially after maintenance or tool changes, which affects process consistency and fault detection.

Innovation Solution

An automated system and method using statistical multivariate analysis to generate and update models, reducing manual input by selecting and processing data to specify acceptable intervals for manufacturing process variables, employing techniques like Hotelling T2-type and DModX calculations for outlier exclusion and model generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual specification of time trajectories by process engineer is used, then model accuracy and expertise-based quality are improved, but time consumption and labor intensity increase significantly

Engineering Contradiction:

Improvemodel accuracyVSAvoidtime consumption

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs automated model building and specification generation without requiring manual intervention by process engineers. The automated model builder selectively retrieves historical process data, applies multivariate statistical analysis (PCA, PLS), and generates time trajectory specifications autonomously, eliminating the 20+ hour manual process while maintaining model quality through algorithmic consistency.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The manual mechanical process of specification creation is replaced with an automated computational system. The system uses computer-based multivariate statistical methods to analyze process data and generate specifications, substituting the human engineer's manual analysis and specification-writing process with automated algorithms that perform the same function more efficiently.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual specification creation by process engineer is used, then expertise-based model quality is improved, but consistency between maintenance operations deteriorates due to human error

Engineering Contradiction:

Improvemodel qualityVSAvoidconsistency

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The system eliminates human variability by performing automated model building that consistently applies the same multivariate statistical algorithms across all maintenance operations. The automated process retrieves historical data, applies PCA and PLS analysis, and generates specifications with identical methodology each time, ensuring reliable consistency between maintenance operations while maintaining model quality through proven statistical techniques.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system incorporates automated validation and comparison mechanisms that evaluate generated specifications against historical performance data. The multivariate statistical analysis provides feedback on model fit and predictive capability, allowing the system to refine and validate specifications systematically, ensuring both quality and consistency across different maintenance operations.

Inventive Principle:

Principle #23Feedback

3Quantity of substance

If comprehensive process data analysis is performed manually, then model completeness is improved, but labor requirements and operational complexity increase

Engineering Contradiction:

Improvedata coverageVSAvoidoperational complexity

Core Design Contradiction:

Quantity of substanceVSDevice complexity

Solution Approach 1:

The system replaces manual data analysis operations with automated computational processes. The automated model builder systematically retrieves comprehensive process data from multiple sources, applies multivariate statistical analysis (PCA for dimensionality reduction, PLS for predictive modeling), and generates complete specifications without requiring manual data collection or analysis efforts, thereby maintaining data coverage while reducing operational complexity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The automated model building system performs multiple functions within a single integrated process: data retrieval from various sources, data preprocessing, multivariate statistical analysis (PCA and PLS), model validation, and specification generation. This multi-functional approach comprehensively analyzes process data while simplifying operational complexity by consolidating what would otherwise require multiple separate manual operations.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS8494798B2Automated model building and batch model building for a manufacturing process, process monitoring, and fault detection

Publication Date: 2013.07.23 SARTORIUS STEDIM DATA ANALYTICS AB

AI summary

A method for creating a new model of a manufacturing process according to a multivariate analysis including selecting a set of data representative of multidimensional data measured during a step or phase of a manufacturing process. The method also includes determining a set of model generation conditions based on the set of data and generating the new model specifying intervals for the multidimensional data measured during a future manufacturing process based on the set of model generation conditions.