Convolutional Neural Network-Based Machine Vision for Non-Destructive Detection of Flooding in Packed Columns.

Area of Science:

• Chemical Engineering
• Process Control
• Artificial Intelligence

Background:

• Packed columns are vital in chemical processes but face operational limits due to flooding.
• Current flooding detection methods lack real-time accuracy, relying on manual inspections or indirect data.

Purpose of the Study:

• To develop a non-destructive, real-time flooding detection system for packed columns.
• To enhance the safety and efficiency of packed column operations.

Main Methods:

• A convolutional neural network (CNN)-based machine vision approach was proposed.
• Real-time images of packed columns were captured and analyzed using a trained CNN model.
• The CNN method was compared against deep belief networks and PCA-SVM.

Main Results:

• The CNN-based machine vision approach demonstrated feasibility and advantages in detecting flooding.
• The method provides a real-time pre-alarm system for flooding events.
• Experimental validation on a real packed column confirmed the effectiveness.

Conclusions:

• The proposed CNN machine vision method offers an effective solution for real-time, non-destructive flooding detection.
• This approach enables timely intervention by process engineers, preventing operational disruptions.
• The study highlights the potential of AI in improving chemical process monitoring and control.