Classification of Sand-Binder Mixtures from the Foundry Industry Using Electrical Impedance Spectroscopy and Support...

Area of Science:

• Materials Science
• Chemical Engineering
• Industrial Chemistry

Background:

• Foundry operations rely on molding sand mixtures, whose optimal control and regeneration are hindered by a lack of efficient in-line monitoring.
• Current monitoring methods are insufficient for real-time process adjustments and quality assurance in sand mixture utilization and regeneration.

Purpose of the Study:

• To investigate the feasibility of electrical impedance spectroscopy (EIS) as an in-line monitoring method for foundry molding sand mixtures.
• To develop and evaluate a machine learning approach for classifying different sand mixtures using EIS data.

Main Methods:

• Characterization of various molding sand mixtures using EIS across a frequency range of 0.5 kHz to 1 MHz under laboratory conditions.
• Development of a database of EIS responses for different sand compositions.
• Application of support vector machines (SVM) with sequential feature selection for classifying sand mixtures based on EIS data.

Main Results:

• EIS successfully characterized diverse sand mixtures.
• Support vector machines achieved high classification accuracies (above 90%) for molding sand mixtures.
• Feature selection further improved classification performance and demonstrated low standard uncertainty in SVM predictions.

Conclusions:

• Machine-learning-enhanced EIS (MLEIS) is a viable technique for in-line monitoring of bulk materials in the foundry industry.
• The developed MLEIS method offers a promising solution for optimizing foundry processes through accurate and efficient sand mixture monitoring.