The Experimental Process Design of Artificial Lightweight Aggregates Using an Orthogonal Array Table and Analysis by...

Area of Science:

• Materials Science
• Chemical Engineering
• Data Science

Background:

• Artificial lightweight aggregates (ALAs) are crucial in construction for reducing structural weight.
• Optimizing ALA production processes is essential for cost-effectiveness and performance.
• Current manufacturing methods often lack precise control over critical parameters.

Purpose of the Study:

• To experimentally design and optimize the drying, calcination, and sintering of ALAs.
• To develop a machine learning model for predicting ALA manufacturing outcomes.
• To enhance the data-driven understanding of ALA production.

Main Methods:

• Utilized an L18 orthogonal array for experimental design of ALA processing parameters.
• Expanded experimental data to 486 instances for robust model training.
• Applied machine learning techniques including linear regression, random forest, and Support Vector Regression (SVR).

Main Results:

• Support Vector Regression (SVR) demonstrated superior predictive performance for ALA properties.
• The developed SVR model accurately predicted measured values.
• The model showed effectiveness in forecasting outcomes for untested process conditions.

Conclusions:

• Machine learning, particularly SVR, offers a powerful tool for optimizing ALA manufacturing.
• Experimental design combined with data analytics can significantly improve process efficiency and product quality.
• This approach provides a scalable method for predicting and controlling ALA production.