Quantitative and qualitative analysis of lost circulation in natural and induced fractured formations: the integration of operational conditions and geomechanical parameters

Lost circulation is one of the most hazardous and costly problems encountered while drilling operation not only in naturally fractured formations but also in hydraulically induced fractured ones. To be more precise, mud losses into the formation may occur at hydrostatic mud pressures in excess of the formation pressure which can be defined as a kind of wellbore instability. In general, there are many operational parameters affect lost circulation that the important ones are mud type, rheological parameters of drilling mud and mud weight. Moreover, some geomechanical issues are influential on lost circulation such as in situ stresses, pore pressure, rock strength as well as pre-existing natural fractures and their orientation. Artificial neural networks (ANNs) are prime candidates to integrate the vast numbers of effective parameters for predicting such events. In spite of many advantages, ANNs have some drawbacks such as its tendency to memorise training data rather than learning to generalise from trend and local optima. In order to deal with these problems, Support vector regression (SVR) has been introduced as a robust novel approach. Although ANN has great capability in predicting lost circulation, SVR yields relatively more accurate results.