b vakilinia - Petroleum and Gas Engineering Department, Science and Research Branch, Islamic Azad University, Tehran, Iran
s Jamshidi - Chemical and Petroleum Engineering Department, Sharif University of Technology, Tehran, Iran

Drilling fluid is one of the most important parts of drilling operation, which takes about 30% of the total operation cost. Lost circulation is the uncontrolled flow of the whole or a part of the drilling fluid into the formation and there are two types of this phenomenon: Induced and Natural fluid loss. Since lost circulation may cause a lot of problems such as loss of drilling time and excess in cost, formation damage, stuck pipe, flow of oil and gas into the well (kick) and blowout, it can be mentioned as one of the most serious problems that can arise during drilling operation. Therefore, detecting the amount and the type of lost circulation can contribute to the elimination of these problems.Artificial Neural Network (ANN) is a parallel system and a mathematical model of biological neural networks and human brain neurons, which is capable of learning. It discovers quickly the complex relationships between parameters with very little error and provides a reasonable answer.In this study an accurate diagnosis of different types of lost circulation and the amount of each type, specially induced one, which has been impossible so far, is estimated by using an ANN. The considered effective drilling parameters of mud loss which were used in ANN are mud weight, mud viscosity, yield point, gel strength, solid content, rate of pump and pump pressure, obtained from 15 wells drilling information of one of the Iranian gas fields, 10 wells used for network training and the other 5 wells used for testing. The network has the R-square of 0.9737, 0.9733 and 0.9779 for training, validation and testing, respectively. According to the obtained errors, by comparing the network results and drilling information, ANN turns out to be a reliable tool for predicting the lost circulation.

Drilling fluid, Lost Circulation, Natural loss, Induced loss, Artificial Neural Network, R_square