Case Study: Stuck Pipe Analysis for Deviated Wells

Research output: ThesisMaster's Thesis

Standard

Case Study: Stuck Pipe Analysis for Deviated Wells. / Holoda, Gergely.
2015.

Research output: ThesisMaster's Thesis

Bibtex - Download

@mastersthesis{75ac5e629c534a66b8583d3596a41f32,
title = "Case Study: Stuck Pipe Analysis for Deviated Wells",
abstract = "Stuck pipe problems, within that, differential sticking problems is probably the greatest drilling problem worldwide in terms of time and financial costs. If once the drillstring stuck, a timely and costly freeing procedure need to regain the moving ability. The company supporting this thesis provided data, brooked three differential sticking in the same area. To free the pipe, took more than 10 days overall, so finding the reasons is essential. The company assumed that there is a connection between these problems, and the aim of this thesis is to find it or them. Work started with literature review from all over the industry{\textquoteright}s history, and then a summary was made of the influencing factors of differential sticking, introduced every one of them in detail and how could modify them to avoid sticking. The thesis contains a computational prediction with neural network modeling to help to prove that the applicable modifications were right, and the sticking would not have occurred in the same situation again. Neural network modeling is a powerful data model that is able to capture and represent complex input-output relationships. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform {"}intelligent{"} tasks similar to those performed by the human brain. The true power and advantage of neural networks lies in their ability to represent both linear and non-linear relationships and in their ability to learn these relationships directly from the data being modeled. To able to use this powerful data model, the first step was to determine the parameters, what the author would like to use in the model. It was a complex task, because not all of the necessary data was available, some of them had to be calculated from other available data. The generated network at the end made up from 17406 datasets, of them 20% were used to test the model. The results of testing phase ended with 0.0575% bad prediction, so the prediction was quite punctual. During the analysis of sticking situations, the factors, which influence differential sticking were investigated. Among other things well trajectory, mud program, mud formulation, flow pattern, solids control, BHA composition. The applied modifications finally were put into the model and prediction was made for these cases. The neural network model predicted all cases as non-sticking situation, so it could stated that reasons of sticking were found.",
keywords = "Feststecken von Bohrgest{\"a}nge, Feststecken durch Druckunterschiede, Analyse, K{\"u}nstlich-Neuronalen-Netzwerk, Modellieren, stuck pipe, differential sticking, analysis, neural network, modeling",
author = "Gergely Holoda",
note = "embargoed until 30-11-2020",
year = "2015",
language = "English",

}

RIS (suitable for import to EndNote) - Download

TY - THES

T1 - Case Study: Stuck Pipe Analysis for Deviated Wells

AU - Holoda, Gergely

N1 - embargoed until 30-11-2020

PY - 2015

Y1 - 2015

N2 - Stuck pipe problems, within that, differential sticking problems is probably the greatest drilling problem worldwide in terms of time and financial costs. If once the drillstring stuck, a timely and costly freeing procedure need to regain the moving ability. The company supporting this thesis provided data, brooked three differential sticking in the same area. To free the pipe, took more than 10 days overall, so finding the reasons is essential. The company assumed that there is a connection between these problems, and the aim of this thesis is to find it or them. Work started with literature review from all over the industry’s history, and then a summary was made of the influencing factors of differential sticking, introduced every one of them in detail and how could modify them to avoid sticking. The thesis contains a computational prediction with neural network modeling to help to prove that the applicable modifications were right, and the sticking would not have occurred in the same situation again. Neural network modeling is a powerful data model that is able to capture and represent complex input-output relationships. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform "intelligent" tasks similar to those performed by the human brain. The true power and advantage of neural networks lies in their ability to represent both linear and non-linear relationships and in their ability to learn these relationships directly from the data being modeled. To able to use this powerful data model, the first step was to determine the parameters, what the author would like to use in the model. It was a complex task, because not all of the necessary data was available, some of them had to be calculated from other available data. The generated network at the end made up from 17406 datasets, of them 20% were used to test the model. The results of testing phase ended with 0.0575% bad prediction, so the prediction was quite punctual. During the analysis of sticking situations, the factors, which influence differential sticking were investigated. Among other things well trajectory, mud program, mud formulation, flow pattern, solids control, BHA composition. The applied modifications finally were put into the model and prediction was made for these cases. The neural network model predicted all cases as non-sticking situation, so it could stated that reasons of sticking were found.

AB - Stuck pipe problems, within that, differential sticking problems is probably the greatest drilling problem worldwide in terms of time and financial costs. If once the drillstring stuck, a timely and costly freeing procedure need to regain the moving ability. The company supporting this thesis provided data, brooked three differential sticking in the same area. To free the pipe, took more than 10 days overall, so finding the reasons is essential. The company assumed that there is a connection between these problems, and the aim of this thesis is to find it or them. Work started with literature review from all over the industry’s history, and then a summary was made of the influencing factors of differential sticking, introduced every one of them in detail and how could modify them to avoid sticking. The thesis contains a computational prediction with neural network modeling to help to prove that the applicable modifications were right, and the sticking would not have occurred in the same situation again. Neural network modeling is a powerful data model that is able to capture and represent complex input-output relationships. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform "intelligent" tasks similar to those performed by the human brain. The true power and advantage of neural networks lies in their ability to represent both linear and non-linear relationships and in their ability to learn these relationships directly from the data being modeled. To able to use this powerful data model, the first step was to determine the parameters, what the author would like to use in the model. It was a complex task, because not all of the necessary data was available, some of them had to be calculated from other available data. The generated network at the end made up from 17406 datasets, of them 20% were used to test the model. The results of testing phase ended with 0.0575% bad prediction, so the prediction was quite punctual. During the analysis of sticking situations, the factors, which influence differential sticking were investigated. Among other things well trajectory, mud program, mud formulation, flow pattern, solids control, BHA composition. The applied modifications finally were put into the model and prediction was made for these cases. The neural network model predicted all cases as non-sticking situation, so it could stated that reasons of sticking were found.

KW - Feststecken von Bohrgestänge

KW - Feststecken durch Druckunterschiede

KW - Analyse

KW - Künstlich-Neuronalen-Netzwerk

KW - Modellieren

KW - stuck pipe

KW - differential sticking

KW - analysis

KW - neural network

KW - modeling

M3 - Master's Thesis

ER -