Análisis de las propiedades mecánicas de la guadua angustifolia mediante técnicas estadísticas y redes neuronales
Cargando...
Archivos
Autores
Rojas Alvarado, Javier Mauricio
Director
Tipo de contenido
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
En este trabajo se investigó la dependencia y comportamiento de las propiedades físicas y mecánicas de la guadua angustifolia mediante análisis estadístico y técnicas de inteligencia artificial. Específicamente se relacionaron el contenido de humedad, la densidad, módulo de elasticidad, resistencia a la compresión paralela y perpendicular a la fibra, resistencia al corte paralelo a la fibra, a la tensión y a la flexión. Lo anterior con el fin de obtener modelos de predicción para las variables en estudio, con los cuales se pretende predecir los valores medios de resistencia del material ante diferentes solicitaciones. El estudio se desarrolló sobre la base experimental de un trabajo previo, elaborado empleando probetas de guadua obtenidas de tres ubicaciones geográficas diferentes en Colombia y de tres posiciones dentro del culmo, con el fin de contar con el número suficiente de datos para generar modelos de predicción confiables y de aplicación general en el país. Se desarrollaron modelos a partir de las siguientes técnicas: redes neuronales artificiales, lógica difusa y Sistema de inferencia neuronal difuso adaptativo, ANFIS, siendo esta última la única mediante la cual se obtuvieron resultados satisfactorios, permitiendo desarrollar modelos que predicen con suficiente aproximación algunas de las propiedades mecánicas de la guadua Angustifolia.
Abstract. In this research the physical and mechanical properties of angustifolia guadua were investigated by statistical analysis and artificial intelligence techniques. Specifically related moisture content, density, modulus of elasticity, compressive strength parallel and perpendicular to the fiber, shear parallel to the grain, tension, bending and twisting. In order to obtain models for the prediction of guadua mechanical properties, several nonconventional modeling techniques were used. The study was conducted on the experimental basis of a previous work, made using guadua specimens obtained from three different Colombian geographical locations and three positions within the culm. This was done in order to have a sufficient number of data to generate reliable prediction models suitable in the country. Models were developed from the following techniques: Artificial Neural Networks, Fuzzy Logic and Adaptive Neuro Fuzzy Inference System, the latter being the only one by which satisfactory results were obtained. The final ANFIS model was suitable to predict with enough approximation the mechanical properties of angustifolia guadua..
Abstract. In this research the physical and mechanical properties of angustifolia guadua were investigated by statistical analysis and artificial intelligence techniques. Specifically related moisture content, density, modulus of elasticity, compressive strength parallel and perpendicular to the fiber, shear parallel to the grain, tension, bending and twisting. In order to obtain models for the prediction of guadua mechanical properties, several nonconventional modeling techniques were used. The study was conducted on the experimental basis of a previous work, made using guadua specimens obtained from three different Colombian geographical locations and three positions within the culm. This was done in order to have a sufficient number of data to generate reliable prediction models suitable in the country. Models were developed from the following techniques: Artificial Neural Networks, Fuzzy Logic and Adaptive Neuro Fuzzy Inference System, the latter being the only one by which satisfactory results were obtained. The final ANFIS model was suitable to predict with enough approximation the mechanical properties of angustifolia guadua..
Abstract
Palabras clave propuestas
Guadua angustifolia; Propiedades físicas; Propiedades mecánicas; Inteligencia artificial; Redes neuronales artificiales; Lógica difusa; Adaptive neuro fuzzy inference system; Modelos predictivos; Angustifolia guadua; Physical properties; Mechanical properties; Artificial intelligence; Artificial neural networks; Fuzzy logic; Adaptive neuro fuzzy inference system; Predictive models