Technical Journal, Vol. 19 No. 4, 2025.
Original scientific paper
https://doi.org/10.31803/tg-20250604003051
Predictive Modeling with Artificial Neural Networks to Optimize Dosing Accuracy of Galenical Powder Dosing Systems
Jaime Cancho
orcid.org/0000-0002-7476-6979
; Universidad Nacional Federico Villarreal, Graduate School, Jr. Prolongations Camaná N°1014, Lima 01, Perú
*
Ciro Rodriguez
; Universidad Nacional Mayor de San Marcos, Lima, Perú
Ivan Petrlik
; Universidad Nacional Federico Villarreal, Lima, Perú
Milner Liendo
; Newman Escuela de Posgrado, Tacna, Perú
* Corresponding author.
Abstract
A predictive model is proposed based on artificial neural networks (RNA) to optimize the quality of dosing of galenical powders in bottles, where it is necessary to maintain accuracy and stability, as the current electromechanical control methods have these shortcomings. The experimental development of research fosters new skills, which are key to innovating and facing the challenges of today's knowledge society. The RNA model was applied to the control group, resulting in an experimental group with improved. Six neural network models were trained, achieving the best results with the Recurrent Neural Network (RNN) model. Tests were conducted to optimize process capability indicators, improve process accuracy, and effectively predict the accuracy of dosed weight, considering the system's operating parameters. The RNN model was trained and validated with real data. The findings demonstrate that the application of the proposal will optimize accuracy and weight stability, meeting the quality standards in the industry.
Keywords
dosing; galenic powder; prediction; precision; rigid packaging; RNN
Hrčak ID:
335274
URI
Publication date:
15.12.2025.
Visits: 378 *