Wafer Transfer Behavior Recognition for Cluster Tool Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of multi-chamber type cluster semiconductor manufacturing apparatuses makes it difficult to identify and address productivity losses, as manufacturing operations are intertwined, and idle operations in one unit can affect others, hindering root cause analysis.
Innovation Solution
A behavior recognition device connected to semiconductor manufacturing apparatuses processes log data to build a machine learning model that identifies good and bad behaviors, simulates scenarios to reproduce and adjust control rules, and automatically generates attributions to capture the root cause of productivity losses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multi-chamber type cluster semiconductor manufacturing apparatus is used to increase throughput, then productivity is improved, but device complexity increases making root cause analysis difficult
Solution Approach 1:
The patent segments the complex manufacturing operations into discrete state transitions within a Markov chain model. Each manufacturing unit's operations are divided into distinct states (e.g., wafer loading, processing, unloading) and transitions between these states. This segmentation allows the system to analyze complex multi-chamber operations by breaking them down into manageable state transitions that can be individually monitored and analyzed for root cause identification.
2Productivity
If multiple manufacturing units operate simultaneously to increase throughput, then productivity is improved, but difficulty of detecting and measuring root causes increases
Solution Approach 1:
The patent implements feedback mechanisms by continuously monitoring state transitions of manufacturing units and comparing actual transitions against the Markov chain model's expected transitions. When deviations are detected (indicating potential root causes), the system generates alerts and performs automated analysis. This feedback loop enables real-time detection of anomalies in multi-unit operations and automatically traces them to root causes, making root cause detection feasible despite simultaneous operations across multiple units.
Solution Approach 2:
The Markov chain model serves as an intermediary between the complex multi-unit manufacturing system and the root cause analysis process. The model acts as a mediator that translates raw operational data from multiple manufacturing units into meaningful state transition patterns. This intermediary layer simplifies the detection and measurement of root causes by providing a structured framework that captures the relationships between different units' operations without requiring direct analysis of the entire complex system.
3Measurement precision
If automated behavior recognition is implemented to identify root causes, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent creates a virtual copy of the manufacturing system's operational behavior through the Markov chain model. Instead of directly analyzing the complex physical system, the system uses the mathematical model as a copy that replicates the essential state transition patterns. This copying approach enables automated behavior recognition and precise root cause identification by analyzing the model's transitions rather than the raw complex operational data, thereby improving measurement precision without proportionally increasing the analysis system's complexity.
Data Source
AI summary
A behavior recognition device for recognizing behaviors of a semiconductor manufacturing apparatus includes a storage device and a control unit. The storage device is configured to store log data of the semiconductor manufacturing apparatus. The control unit is cooperatively connected to the storage device, and configured to build a transition state model based on the log data to analyze behaviors related to wafer transfer sequences and manufacturing operations of the semiconductor manufacturing apparatus.