Machine Learning Methods Applied for Modeling the Process of Obtaining Bricks Using Silicon-Based Materials.

Area of Science:

• Materials Science
• Environmental Engineering
• Chemical Engineering

Background:

• Industrial brick manufacturing facilities often face challenges in maintaining productivity and product quality.
• Process engineers seek methods to improve efficiency and reduce environmental impact.

Purpose of the Study:

• To evaluate the impact of adding new auxiliary materials on exhaust emissions in brick manufacturing using machine learning.
• To develop predictive models for exhaust emission changes based on raw material composition.

Main Methods:

• A database of 121 brick batches was created from experimental data.
• Artificial neural networks (ANNs) and regression algorithms, including k-Nearest Neighbors (kNN), were employed for predictive modeling.
• A multi-objective optimization procedure using a grid search method was implemented.

Main Results:

• Feed-forward neural networks with two hidden layers achieved Mean Squared Error (MSE) < 0.01 and R-squared (r²) > 0.82.
• The kNN regression model demonstrated an error < 0.6.
• An optimization procedure identified optimal raw material parameters (dry product mass, clay, ash, organic materials) for minimizing gas emissions.

Conclusions:

• Machine learning models effectively predict exhaust emissions in brick production.
• Optimization techniques can determine raw material compositions that minimize gas emissions while considering energy potential.
• The findings support sustainable practices in the industrial brick manufacturing sector.