INFERENCIA DE VELOCIDADES DE ONDAS S MEDIANTE LA TÉCNICA DE REDES NEURO-DIFUSAS
No Thumbnail Available
Date
Advisors
Evaluators
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Industrial de Santander
Abstract
Description
En este trabajo utilizamos la técnica de redes neuronales difusas (RND) con la finalidad de obtener ecuaciones que permitan predecir Velocidades de Cizalla (Vs) a partir de registros de porosidad (φ), saturación de agua (Sw) y volumen de arcilla (Vsh) pertenecientes a un pozo petrolero. Para el entrenamiento de las RND se utilizaron como datos de entrada valores de φ, Vsh y Sw y, como salida el registro de Vs. Los registros corresponden a un pozo perteneciente del Campo Guafita, ubicado en la subcuenca de Apure. El entrenamiento se realizó con múltiples combinaciones de los parámetros independientes φ, Vsh y Sw. Los resultados obtenidos sugieren que el uso de los tres tipos de registros simultáneamente mejora de forma notoria la inferencia de Vs, comparado con el uso de sólo uno de ellos o combinaciones de dos. El número de reglas difusas se varió para todas las combinaciones de parámetros. Se observó que un aumento del número de reglas no produce una mejoría notoria en los resultados.
In this work we use neuro fuzzy system (NFS) in order to obtain inference equations of shear velocity (Vs) of a well in terms of the logarithmic of its porosity (φ), its water saturation (Sw) and its clay volume (Vsh) data. The data belong to a well of the Guafita Field located in the Sub-basin of Apure, Venezuela. For training of NFS were used as input values: φ, Sw, Vsh; and Vs a s output. The training was made with multiple combinations of the independent parameters φ, Vsh and Sw. The results suggest that the use of the three types of registers simultaneously improves notoriously Vs inference, compared with the use of only one of them or combinations of two parameters. The number of fuzzy rules was changed for all combinations of parameters. It was observed that increasing the number of rules does not produce a marked improvement in the results.
In this work we use neuro fuzzy system (NFS) in order to obtain inference equations of shear velocity (Vs) of a well in terms of the logarithmic of its porosity (φ), its water saturation (Sw) and its clay volume (Vsh) data. The data belong to a well of the Guafita Field located in the Sub-basin of Apure, Venezuela. For training of NFS were used as input values: φ, Sw, Vsh; and Vs a s output. The training was made with multiple combinations of the independent parameters φ, Vsh and Sw. The results suggest that the use of the three types of registers simultaneously improves notoriously Vs inference, compared with the use of only one of them or combinations of two parameters. The number of fuzzy rules was changed for all combinations of parameters. It was observed that increasing the number of rules does not produce a marked improvement in the results.
Keywords
neuro fuzzy system, Porosity, shear velocity, water saturation, clay volume, ANFIS, redes neuronales difusas, velocidad de cizalla, porosidad, saturación de agua, volumen de arcilla, petróleo