Virtual Sensor for Advanced Grinding Control Through Machine Learning

Vilcahuaman P.M.; Perez Paredes M.G.S. · IEEE Andescon, ANDESCON 2024 - Proceedings (2024) · DOI: 10.1109/andescon61840.2024.10755536

Resumen

This paper introduces the development of a virtual sensor utilizing machine learning algorithms in response to challenges observed with physical sensors in grinding control processes, such as calibration issues, communication failures, and frequent disconnections. To enhance automation process reliability, we propose employing estimators based on Random Forest Regression (RFR), Multilayer Perceptron (MLP), and one-dimensional Convolutional Neural Networks (CNN-1D) with kernel filters. The performance of these models is evaluated within an advanced control loop incorporating OPC protocol communication with a PLC. This evaluation identifies the most robust algorithm across diverse operational conditions.

Instituciones: Universidad Nacional de Ingeniería; Universidad Tecnológica del Perú