Seed Crystal Edge Detection for Automated Seeding Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional monocrystal growth control systems rely on manual determination of the seeding process timing, which is subject to human skill variations, affecting yield and efficiency, and lack automated solutions for determining the optimal timing.
Innovation Solution
An automated method using a deep learning model trained on angular images of seed crystal monocrystal lines to identify edges and corners, employing a pre-training model, iteration training, and real-time verification to determine the seeding condition accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual determination is used to decide seeding process timing, then operational flexibility is maintained, but determination consistency and productivity deteriorate due to human skill variations
Solution Approach 1:
The patent replaces the manual mechanical observation and determination process with an automated image processing system. The system captures images of the monocrystal line, processes them through algorithms to detect edges and corners, and automatically determines seeding timing, eliminating human skill variations and improving both consistency and efficiency
Solution Approach 2:
The system enables the monocrystal growth control system to automatically determine seeding timing without human intervention. The automated image processing and analysis system serves itself by continuously monitoring the monocrystal line state and triggering seeding operations based on pre-set criteria, improving productivity while maintaining determination consistency
2Extent of automation
If manual monitoring and operation are performed, then operational control is maintained, but labor costs and device complexity increase
Solution Approach 1:
The patent integrates multiple functions into a single automated system: image capture, image processing, edge detection, corner detection, and seeding control all occur within one system. This multi-functional integration increases automation while managing complexity by consolidating operations rather than adding separate systems
Solution Approach 2:
The patent introduces an image processing system as an intermediary between the monocrystal growth process and the seeding control. This intermediary automatically analyzes the monocrystal line state through image processing and translates it into control signals, enabling automation without requiring direct complex interaction between all system components
Data Source
AI summary
The present application relates to the technical field of identification during crystal growth, and specifically relates to a method for identifying and detecting the edges and corners of a seed crystal single crystal wire in a crystal growth process, comprising four steps, i.e., constructing a high-quality data set, constructing an artificial intelligence algorithm model, performing model training, and performing model evaluation and verification. The present invention overcomes the defects of manual determination of the starting time of a seeding procedure, and provides a method for automatically determining the seeding time, so that a control system automatically controls the operation of starting the seeding procedure after temperature adjustment is ended.