Hierarchical Model Parameter Management for Substrate Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the substrate manufacturing process, the large number of inference models required for device fault detection, component deterioration diagnosis, and process control leads to increased management costs due to the complexity of managing multiple device models across various chambers.
Innovation Solution
A model management system that separates models into multiple layers, with a first management unit calculating new model parameters and controlling their distribution to lower-level management units, allowing for efficient parameter sharing and reduced management costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If each inference model is optimized separately for multiple device models and chambers, then the accuracy and performance of each model is improved, but the number of models to be managed becomes enormous and management cost increases
Solution Approach 1:
The patent segments the model management system into multiple layers: a base layer containing common models shared across device models and chambers, and upper layers containing device model-specific and chamber-specific models. This hierarchical segmentation reduces the total number of models by identifying and extracting commonalities into the base layer, while maintaining specialized models only where necessary.
Solution Approach 2:
The patent merges common inference models that are shared across multiple device models and chambers into a single base layer. By combining these common models and managing them centrally, the system reduces redundancy and simplifies model management while maintaining the ability to apply device-specific and chamber-specific modifications through upper layers.
2Adaptability or versatility
If multiple device models with various chambers are managed separately, then each model can be optimized for its specific function, but the management cost increases due to the large number of models
Solution Approach 1:
The patent introduces a new dimensional organization by arranging models in a hierarchical structure with multiple layers (base layer, device model layer, chamber layer) instead of managing all models at a single level. This dimensional change allows the system to maintain specialized models for different devices and chambers while reducing management costs through shared common models in the base layer.
Solution Approach 2:
The base layer contains universal models that serve multiple device models and chambers simultaneously. These common models provide multi-functional support, reducing the need to create separate specialized models for each device and chamber combination, thereby reducing management costs while maintaining adaptability.
3Reliability
If inference models are constructed for every device model and chamber combination, then comprehensive coverage and accuracy are achieved, but the complexity of model management becomes unmanageable
Solution Approach 1:
The patent implements a nested hierarchical structure where the base layer contains common models, the device model layer contains device-specific models that inherit or extend from the base layer, and the chamber layer contains chamber-specific models. This nesting allows comprehensive coverage at each level while managing complexity through the organized hierarchy, where lower layers provide foundational functionality that upper layers build upon.
Data Source
AI summary
A model management system, a model management method, and a model management program that efficiently manage models applied to a substrate manufacturing process is provided. The model management system separately manages the models applied to the substrate manufacturing process in three or more layers, and includes a first management unit configured to manage a model at a predetermined layer, and one or more second management units configured to manage one or more models at a layer one level lower than the predetermined layer. The first management unit includes a calculating unit configured to, when one or more model parameters of the one or more models managed by the one or more second management units are updated, calculate a new model parameter based on each of the updated one or more model parameters, and a control unit configured to perform control so that the new model parameter is set in a plurality of models respectively managed by a plurality of management units at a lowest layer, belonging to the first management unit.