Substrate Spray Process Monitoring Using Deep Learning Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional substrate treatment processes using vision sensors face difficulties in detecting the normalcy of the process due to obstacles like bowls blocking the vision and irregular aerosol forms, making it hard to diagnose the substrate treatment process effectively.
Innovation Solution
A deep learning model-based method that preprocesses images by recognizing nozzle tips, correcting perspective distortions, and converting coordinate systems to determine the normalcy of substrate treatment processes by analyzing spray states and wetting conditions in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional vision sensor is used to detect substrate treatment process, then the detection system is simple, but the detection accuracy deteriorates due to bowl blocking and irregular aerosol forms
Solution Approach 1:
The patent replaces the conventional vision sensor system with a deep learning-based diagnostic system that uses multiple sensors (including vision sensors, flow rate sensors, and temperature sensors) to capture comprehensive data about the substrate treatment process. The deep learning model processes this multi-source data to accurately detect process normalcy despite obstacles like bowl blocking and irregular aerosol forms, thereby improving detection accuracy while managing system complexity through intelligent data processing.
Solution Approach 2:
The patent changes the detection parameters by transitioning from simple visual detection to multi-parameter analysis including flow rate, temperature, and image data. The deep learning model integrates these multiple parameters to compensate for the limitations of individual sensors, enabling accurate detection even when the bowl blocks the vision sensor or when aerosol forms are irregular.
2Productivity
If multiple nozzles are used for processing, then the processing capability is improved, but the detection difficulty increases due to blocked vision and irregular spray forms
Solution Approach 1:
The patent applies segmentation by dividing the detection task into multiple independent detection zones, one for each nozzle. The deep learning model processes images and sensor data separately for each nozzle region, allowing simultaneous monitoring of multiple nozzles without the vision being blocked by the bowl. This segmentation approach maintains high processing capability while reducing detection difficulty by localizing the analysis for each nozzle.
Solution Approach 2:
The patent implements a universal detection system that can simultaneously monitor multiple nozzles, the bowl, and various process parameters using the same deep learning framework. This multi-functional system handles detection for different nozzles with different spray patterns, flow rates, and positions, thereby maintaining high productivity while managing the complexity of detecting multiple irregular spray forms.
3Manufacturing precision
If aerosol spray forms are monitored, then the process quality is improved, but the measurement precision deteriorates due to transparent and irregular spray forms
Solution Approach 1:
The patent introduces an intermediary deep learning model that acts as a mediator between the transparent, irregular aerosol spray forms and the detection system. The model processes raw image data and sensor readings, extracting meaningful features that represent spray form quality despite the transparency and irregularity. This intermediary processing layer enables accurate manufacturing precision monitoring by transforming difficult-to-detect spray characteristics into quantifiable metrics.
Solution Approach 2:
The patent changes the measurement parameters by transitioning from direct visual detection of transparent aerosols to indirect detection through multiple parameters including flow rate, temperature, and image texture analysis. The deep learning model integrates these parameters to infer spray form quality, thereby maintaining manufacturing precision while overcoming the limitation of detecting transparent and irregular spray forms directly.
Data Source
AI summary
The inventive concept provides a method to determine whether a substrate treatment process is normal using a deep learning model. The method comprising receiving input on a substrate treatment process video, preprocessing the inputted video, using the deep learning model to study a preprocessed video, and determining whether the substrate treatment process is normal by comparing the trained model and a real time substrate treatment process video.