TITLE:
Hole Cleaning Prediction in Foam Drilling Using Artificial Neural Network and Multiple Linear Regression
AUTHORS:
Reza Rooki, Faramarz Doulati Ardejani, Ali Moradzadeh
KEYWORDS:
Foam Drilling; Hole Cleaning; Artificial Neural Network; Multiple Linear Regression
JOURNAL NAME:
Geomaterials,
Vol.4 No.1,
January
17,
2014
ABSTRACT:
Foam drilling is increasingly used to develop low pressure reservoirs or
highly depleted mature reservoirs because of minimizing the formation damage
and potential hazardous drilling problems. Prediction of the cuttings
concentration in the wellbore annulus as a function of operational drilling
parameters such as wellbore geometry, pumping rate, drilling fluid rheology and
density and maximum drilling rate is very important for optimizing these
parameters. This paper describes a simple and more reliable artificial neural
network (ANN) method and multiple linear regression (MLR) to predict cuttings
concentration during foam drilling operation. This model is applicable for
various borehole conditions using some critical parameters associated with foam
velocity, foam quality, hole geometry, subsurface condition (pressure and
temperature) and pipe rotation. The average absolute percent relative error (AAPE) between the experimental cuttings
concentration and ANN model is less than 6%, and using MLR, AAPE is less than 9%. A comparison of
the ANN and mechanistic model was done. The AAPE values for all datasets in this study were 3.2%, 8.5% and 10.3% for ANN model,
MLR model and mechanistic model respectively. The results show high ability of
ANN in prediction with respect to statistical methods.