Multilabel Image Classification Based Fresh Concrete Mix Proportion Monitoring Using Improved Convolutional Neural...

Area of Science:

• Construction Engineering
• Materials Science
• Artificial Intelligence

Background:

• Accurate concrete mix proportions are vital for structural integrity and performance.
• Current methods lack real-time, full-scale monitoring capabilities for fresh concrete during manufacturing.
• Deficiencies in monitoring hinder green manufacturing and construction safety.

Purpose of the Study:

• To develop a state-of-the-art method for real-time monitoring of fresh concrete mix proportions.
• To address the limitations of existing testing methods in concrete manufacturing.
• To improve green manufacturing and safety in construction through accurate mix proportion control.

Main Methods:

• Utilized improved convolutional neural network (CNN) multilabel image classification.
• Collected a high-quality dataset of concrete mixture images.
• Fine-tuned four CNNs for multilabel classification to determine multiple mix proportion parameters.

Main Results:

• The improved CNN models demonstrated excellent learning and generalization abilities.
• The best-performing model was integrated into a monitoring system with hardware sensors.
• The system achieved real-time, full-scale monitoring of fresh concrete mix proportions.

Conclusions:

• The developed system enables accurate sensing and warning for incorrect concrete mix proportions.
• This image-based approach replaces traditional laboratory experiments for proportion verification.
• The system facilitates intelligent sensing and contributes to safer, more efficient construction practices.