Uniaxial compressive strength (UCS) of rock is of great use in drilling and stimulation of oil and gas wells such as wellbore stability analysis and fracturing operations whereas UCS is a key parameter that can be used to increase the efficiency of drilling and stimulation operations. Artificial Neural Network (ANN) is a novel approach for solving engineering problems. ANNs, like people, learn by example. They use input–output parameters to be trained to recognize the correct relationship. These methods are able to consider all effective parameters simultaneously and also develop and learn from the field data (due to existing errors and uncertainties) directly. In this study, the P-wave velocity, density and porosity were used as inputs to predict UCS as output. These input and output parameters are related to experimental studies on sandstone cores. Obtained results of artificial neural network compare with real results and their correlation has been validated. The results show that the proposed ANN methods could be applied as a new acceptable method for the prediction of UCS. By developing neural networks approaches with these input parameters, UCS can be determined in well planning which leads to optimization and cost reduction.
Uniaxial compressive strength (UCS) is a key parameter in wellbore stability, mud weight window analysis and fracturing operations. UCS is usually determined through a uniaxial or unconfined compression test in a laboratory. While this test method appears to be relatively simple, it is time-consuming, comparably costly, and requires carefully prepared rock samples. High-quality core samples are needed for the application of UCS in laboratory. However, such cores cannot always be extracted from weak, highly fractured, thinly bedded and/or block-in-matrix rocks [1-3]. Additional difficulties exist concerning the extraction of good quality samples, either from an outcrop in the field or from a large block in the laboratory. Weak to very weak rocks may deteriorate during coring and fail to yield good quality samples. For these reasons, many researchers have been undertaken to examine the UCS of rocks. Some of these researches have involved direct laboratory testing investigations and some others have deal with empirical and statistical works. The general tendency to predict UCS of intact rocks is to use simpler, quicker, and less costly rock tests such as the Schmidt rebound hammer, point load test, impact strength, and sonic velocity . These methods, such as point load index (PLI) and Schmidt Hardness Index (SH), are often employed when it is impractical to prepare test samples according to the standards suggested by ISRM  or ASTM . Moreover, because of the large range of conversion coefficients (from 10 to 50) between UCS and PLI [7-10], and large error range in the estimation of UCS from SH due to the compact energy used  the drawbacks for estimation of UCS seems inevitable. Many researchers have been conducting investigations to predict UCS from non-destructive testing methods such as the sound velocity, porosity, Fuzzy model, use of correlation of mineralogical and textural characteristics, etc [12-21].