Investigation and prediction of viscosity of spud type drilling muds added barite, calcium carbonate and olivine by artificial neural networks with limiting data

Abstract

Drilling is one of the most common methods for the production of hydrocarbon or geothermal sources. Success of drilling operations depend on the mechanical and thermal properties of drilling mud. Since ensure direct connect between formation and surface, it is main factor which should be control in drilling. It provides some basic properties during the drilling process such as controlling the formation pressure, carrying the cuttings from bit to surface, suspending solids in the mud when circulation is stopped, forming low-permeability filter cake, maintaining the stability of the borehole, reducing friction between the drilling string and the sides of the hole, cooling and lubricating the bit, assisting in the collection and interpretation of information available from cuttings etc. One of the most significant tasks of the drilling mud is controlling the formation pressure. This pressure, also named as hydrostatic pressure, depends on the depth and mud weight. Also, the mud weight depends mainly High Gravity Solids (HGS), such as barite or calcium carbonate, in the drilling mud. Hence, it is necessary to add these materials when the weight of the mud is desired to be increased. Drilling mud is one of the most important parameters in drilling operations. The main purpose of it is to bring the cuttings from the formation to the surface. For this purpose, the drilling mud must reach a certain viscosity value in order to transport the cuttings to the surface. The viscosity, which has a significant effect on the easy transport of crumbs to the surface, should be observed regularly during drilling. These measurements are made to determine if the drilling mud is efficient. In the present study, estimation of the viscosity of the drilling mud with the Barite, Calcium Carbonate and Olivine were investigated by using artificial neural network. Spud type muds were prepared according to American Petroleum Institute (API) Spec. 13Astandart and then, these materials were added in different amounts (1-6 wt%). Rheological and filtration analysis of the muds were done according to American Petroleum Institute (API) standards. The developed neural network architecture is trained by the limited experimental data and the estimation performance of it is tested with the data not used in training. The results obtained from the viscometer and artificial neural network estimation were compared with each other and they showed sufficient agreement for viscosity estimation of drilling mud. It is observed that the average percentage error in estimation of the drilling mud viscosity was found to be less than 2%. According to the results, the designed artificial neural network structure has very successful prediction performance and it can say that ANN could be used with directly estimate the viscosity or other rheological parameters without any more experimental procedures after training the network with adequate samples.