Evaluación de modelos predictivos para la optimización del costo de las operaciones unitarias de la UM Frances, Pataz – La libertad.
No Thumbnail Available
Date
2024-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Nacional de Trujillo
Abstract
El principal objetivo de este trabajo de investigación fue Evaluar modelos predictivos para la optimización del costo de las operaciones unitarias de la UM. Frances, Pataz-La Libertad.
Se empleó un enfoque de investigación cuantitativo–experimental, consistente en el entrenamiento de 4 modelos predictivos de aprendizaje automático, y posteriormente se evaluó el mejor desempeño, los modelos entrenados fueron Regresión lineal múltiple, Bosque aleatorio, Árbol de decisión y redes neuronales artificiales. Las herramientas utilizadas fueron guías de recolección de datos. La población estaba constituida por los tajos de explotación y la muestra estuvo constituida por los Tajos Hquina, Luz, Maurita N, Maurita S y Jugadora, en total fueron 287 registros que se tomaron desde enero a marzo del año 2024 en la UM. Frances.
Los resultados obtenidos fueron al evaluar la rentabilidad en los meses enero, febrero y marzo fue de 8.58 kUS$, 7.42 kUS$ y 7.40 kUS$ respectivamente. Al entrenar los diferentes modelos predictivos de machine learning en bosques aleatorios se logró un R2 de 89%, en regresión lineal múltiple se logró un R2 de 91%, en árbol de decisión se logró un R2 de 85%, y utilizando redes neuronales artificiales se logró un R2 de 0.92%.
En conclusión, las Redes Neuronales Artificiales demostró un mayor R2 y por tanto se encontró los parámetros óptimos de los costos de las operaciones unitarias (US$/tn); Se obtuvo en perforación y voladura de 41.86, limpieza a pala a 19.78, acarreo 11.36, carguío 15.89, ventilación
7.81, y sostenimiento un 2.07 donde sobrepasando estos costos genera un sobrecosto en las operaciones unitarias como consecuencia afecta la rentabilidad de la empresa.
The main objective of this research was to evaluate predictive models for the optimization of the cost of unit operations at UM. Frances, Pataz La Libertad. A quantitative experimental research approach was used, consisting of training 4 predictive machine learning models, and subsequently the best performance was evaluated; the models trained were Multiple Linear Regression, Random Forest, Decision Tree and Artificial Neural Networks. The tools used were data collection guides. The population was constituted by the exploitation pits and the sample was constituted by the Hquina, Luz, Maurita N, Maurita S and Jugadora pits, in total there were 287 records that were taken from January to March 2024 in the UM. France s The results obtained when evaluating the profitability in January, February and March were 8.58 kUS$, 7.42 kUS$ and 7.40 kUS$ respectively. When training the different predictive models of machine learning in random forests an R2 of 89% was achieved, in mu ltiple linear regression an R2 of 91% was achieved, in decision tree an R2 of 85% was achieved, and using artificial neural networks an R2 of 0.92% was achieved In conclusion, the Artificial Neural Networks showed a higher R2 and therefore the optimum parameters of the unit operations costs (US$/tn) were found; it was obtained in drilling and blasting of 41.86, shovel cleaning at 19.78, hauling 11.36, loading 15.8 9, ventilation 7.81, and support 2.07 where exceeding these costs generates an over cost in the unit operations as a consequence affects the profitability of the company.
The main objective of this research was to evaluate predictive models for the optimization of the cost of unit operations at UM. Frances, Pataz La Libertad. A quantitative experimental research approach was used, consisting of training 4 predictive machine learning models, and subsequently the best performance was evaluated; the models trained were Multiple Linear Regression, Random Forest, Decision Tree and Artificial Neural Networks. The tools used were data collection guides. The population was constituted by the exploitation pits and the sample was constituted by the Hquina, Luz, Maurita N, Maurita S and Jugadora pits, in total there were 287 records that were taken from January to March 2024 in the UM. France s The results obtained when evaluating the profitability in January, February and March were 8.58 kUS$, 7.42 kUS$ and 7.40 kUS$ respectively. When training the different predictive models of machine learning in random forests an R2 of 89% was achieved, in mu ltiple linear regression an R2 of 91% was achieved, in decision tree an R2 of 85% was achieved, and using artificial neural networks an R2 of 0.92% was achieved In conclusion, the Artificial Neural Networks showed a higher R2 and therefore the optimum parameters of the unit operations costs (US$/tn) were found; it was obtained in drilling and blasting of 41.86, shovel cleaning at 19.78, hauling 11.36, loading 15.8 9, ventilation 7.81, and support 2.07 where exceeding these costs generates an over cost in the unit operations as a consequence affects the profitability of the company.
Description
Keywords
TECHNOLOGY::Civil engineering and architecture::Geoengineering and mining engineering::Mining engineering