Real-Time Automated Solubility Screening Method Using Deep Neural Networks with Handcrafted Features.

Area of Science:

• Materials Science
• Chemical Engineering
• Computer Vision

Background:

• Automated and real-time solubility measurements are increasingly vital in research and industry.
• Traditional methods often lack real-time capabilities, hindering process automation.
• Limited labeled image datasets in industrial settings necessitate alternative approaches to end-to-end learning.

Purpose of the Study:

• To develop a computer vision-based method for automatic classification of solution dissolution states.
• To leverage handcrafted features for improved accuracy with limited labeled data.
• To enable real-time, automated solubility screening.

Main Methods:

• Utilized computer vision algorithms to extract nine handcrafted features from solution images.
• Trained a deep neural network (DNN) classifier using these handcrafted features.
• Constructed a dataset of diverse solution images, from partially to fully dissolved states.

Main Results:

• The proposed method successfully classifies solutions based on their dissolution states.
• Achieved automatic, real-time screening of solubility status using common devices like tablets or mobile phones.
• Demonstrated the effectiveness of handcrafted features in conjunction with DNN for this specific task.

Conclusions:

• The developed method offers an efficient and automated approach to solubility measurement.
• Integration with automatic solubility changing systems can lead to fully automated processes.
• This technique has significant potential for industrial applications requiring real-time monitoring.