Search

While many factors influence the success of a given well, the permeability of the surrounding formation is one of the most important properties to understand the nature of any reservoir and to be utilized for effective oil and gas drilling. Gathering data from well logs for different wells can be highly expensive and time-consuming. The goal of this work is to find the best artificial intelligent model which can predict the permeability values with minimum error while saving time and money. Therefore, accurately estimating is highly beneficial to use such a model for further field and engineering applications. In this project, a trial was accomplished through a Machine Learning (ML) approach using several modules of Artificial Intelligent including ANFIS and ANN to examine and build a permeability prediction model based on nine (9) well-logging parameters taken from well-logging data measured at a borehole in carbonate rock. The permeability was predicted from well-log data using Artificial Intelligent (AI) technique. Field data were recorded at one borehole, where all logs are correlated together. After obtaining results, the prediction model can be considered successful, it is highly recommended to utilize ANFIS- Genfis2 as it gives outstanding results as the correlation coefficient training was 1.0 and testing was 0.9347 compared with ANFIS-Genfis1 which was not satisfying with training correlation coefficient of 1.0 and testing 0.4073, including a significant reduction in the percentage error of 14.3% compared of 301%, and utilize ANN with a double layer not single, as the result of single layer showed a correlation coefficient of 0.9337 in training and 0.9924 in testing. In addition, single layer method showed higher error compared with double layer. Conclusively, it is recommended to apply the model with other data obtained from the same reservoir, to minimize the number of unneeded data, enhance the measurement performance by avoiding human errors, and develop other relationships between a set of parameters that can result in a better and most effective prediction model. In novelty, utilizing and studying the output of this trial application of the machine learning approach will summarize the best models and techniques for predicting many important reservoir properties such as Permeability. The number of well logging parameters is high and has been statically analyzed to increase the resolution of the input data. Building this prediction model will increase the recovered amount from the subsurface and will lead to significant cost savings in drilling and exploration operational

The rapidity distribution of well-identified particles such as pions, kaons, protons and their antiparticles measured in AGS, SPS, and BRAHMS experiments (Au+Au collisions), at various energies spanning from higher energies s NN = 200 down to lower energies 2 GeV, are compared with that obtained from huge statistical ensembles of 100, 000 events generated from the Cosmic Ray Monte Carlo (CRMC) EPOS event-generator. All these data are then compared to the results calculated in the frame work of Landau hydrodynamical model. The EPOS event-generator can describe the experimental data of the rapidity distribution for all mentioned particles at all considered energies. This leads to use the EPOS event-generator to produce the rapidity distribution for various particles at energies where there are no experimental data yet to compare with. Such a prediction might serve as an input for future experiments such as FAIR and NICA for instance. A successful comparison between the Landau hydrodynamical approach and the computational CRMC EPOS 1.99 model in describing the rapidity distribution for the mentioned particles is shown in the presence of the experimental data. We conclude that the Landau hydrodynamical model is capable of describing both experimental data and EPOS 1.99 event-generator results, and they are in a good agreement.