Automatic Virtual Metrology Model Porting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional virtual metrology methods require extensive manual effort and cost to create individual models for each process chamber, leading to inefficiencies in implementing virtual metrology across a plant, and lack real-time data quality evaluation capabilities, affecting accuracy and requiring manual retraining with abnormal data.
Innovation Solution
A server-based system for automatic virtual metrology (AVM) that includes a dual-phase VM algorithm, data pre-processing, and quality evaluation models to automatically evaluate and sift data, allowing for the porting and refreshing of VM models across similar equipment types, reducing implementation time and maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual VM models are constructed for each process chamber using conventional methods, then VM accuracy is maintained, but implementation time and cost become prohibitively high
Solution Approach 1:
The patent creates a master VM model that can be copied and adapted to multiple process chambers. Instead of building individual models for each chamber, the system develops one comprehensive model that captures the essential relationships between process parameters and outcomes, then replicates this model across different chambers with minimal customization, dramatically reducing implementation time while preserving accuracy
Solution Approach 2:
The patent develops a universal VM model framework that can serve multiple process chambers simultaneously. The model is designed to be chamber-agnostic, handling variations through standardized adaptation procedures rather than requiring chamber-specific model development, thus enabling one model to perform the function of many individual models
2Measurement precision
If individual VM models are constructed for each process chamber using conventional methods, then VM accuracy is maintained, but manpower and cost increase significantly
Solution Approach 1:
By creating a replicable master model, the system eliminates the need for extensive manual model development for each chamber. The copying approach requires minimal human intervention after the initial model creation, reducing both manpower requirements and associated costs while maintaining the accuracy needed for effective VM
Solution Approach 2:
The system incorporates automated model adaptation mechanisms that reduce manual intervention. Once the master model is established, it can automatically adapt to different chambers using standardized procedures, reducing the need for expert manual configuration and lowering both labor costs and complexity
3Device complexity
If conventional VM methods are used without real-time data quality evaluation, then system complexity is reduced, but data quality affects model accuracy and requires manual intervention
Solution Approach 1:
The patent implements real-time data quality evaluation that provides feedback to the model training and prediction processes. By continuously monitoring data quality metrics and using this information to adjust model operations, the system maintains high accuracy without requiring overly complex manual intervention mechanisms, achieving a balance between automation and precision
Data Source
AI summary
A server, a system and a method for automatic virtual metrology (AVM) are disclosed. The AVM system comprises a model-creation server and a plurality of AVM servers. The model-creation server is used to construct the first set of virtual metrology (VM) models (of a certain equipment type) including a VM conjecture model, a RI (Reliance Index) model, a GSI (Global Similarity Index) model, a DQIx (Process Data Quality Index) model, and a DQIy (Metrology Data Quality Index) model. In the AVM method, the model-creation server also can fan out or port the first set of VM models generated to other AVM servers of the same process apparatus (equipment) type, and each individual fan-out-acceptor's AVM server can perform automatic model refreshing processes so as to gain and maintain its VM models' accuracy.