Implementación de modelos predictivos para incrementar las ventas de jugos en LEYDE S.A.
Loading...
Date
2025-06-20
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Tecnológica Centroamericana UNITEC
Abstract
El presente trabajo se enfocó en reducir el consumo de carbonato de calcio en calderas de lecho fluidizado, manteniendo las emisiones de dióxido de azufre (SO₂) dentro de los límites normativos. El objetivo principal fue optimizar las condiciones operativas mediante el uso de herramientas tecnológicas y técnicas de modelado predictivo. La metodología incluyó un análisis cuantitativo que evaluó variables como el exceso de oxígeno, la presión diferencial y la temperatura, combinadas con modelos de regresión polinomial y algoritmos de machine learning como Random Forest. Los resultados destacaron que el modelo polinomial explicó el 58% de la variabilidad en las emisiones, mientras que Random Forest clasificó eficazmente escenarios óptimos y no óptimos. Se concluyó que estas herramientas permitieron reducir costos operativos y cumplir con normativas ambientales, recomendando su implementación en industrias similares.
This study focused on reducing calcium carbonate consumption in fluidized bed boilers while maintaining sulfur dioxide (SO₂) emissions within regulatory limits. The main objective was to optimize operational conditions using technological tools and predictive modeling techniques. The methodology included a quantitative analysis evaluating variables such as excess oxygen, differential pressure, and temperature, combined with polynomial regression models and machine learning algorithms like Random Forest. Results showed that the polynomial model explained 58% of the variability in emissions, while Random Forest effectively classified optimal and non-optimal scenarios. It was concluded that these tools enabled cost reductions and compliance with environmental regulations, recommending their implementation in similar industries.
This study focused on reducing calcium carbonate consumption in fluidized bed boilers while maintaining sulfur dioxide (SO₂) emissions within regulatory limits. The main objective was to optimize operational conditions using technological tools and predictive modeling techniques. The methodology included a quantitative analysis evaluating variables such as excess oxygen, differential pressure, and temperature, combined with polynomial regression models and machine learning algorithms like Random Forest. Results showed that the polynomial model explained 58% of the variability in emissions, while Random Forest effectively classified optimal and non-optimal scenarios. It was concluded that these tools enabled cost reductions and compliance with environmental regulations, recommending their implementation in similar industries.
Keywords
Aprendizaje automático, Clusterización, Patrones de consumo, Random forest y series de tiempo