dc.contributor.advisor | Sui, Dan | |
dc.contributor.advisor | Wiktorski, Ekaterina | |
dc.contributor.author | Frankiewicz, Jakub Tomasz | |
dc.date.accessioned | 2019-12-18T10:23:34Z | |
dc.date.available | 2019-12-18T10:23:34Z | |
dc.date.issued | 2019-06-15 | |
dc.identifier.uri | http://hdl.handle.net/11250/2633873 | |
dc.description | Master's thesis in Petroleum Engineering. | nb_NO |
dc.description.abstract | The oil and gas industry, especially its upstream part generates a massive amount of data. The proper data collection and processing are the vital elements of reducing the non-productive time and increasing the drilling operations efficiency.
The major part of each well program is the drill bits selection. It is the most important tool which does slicing or crushing downhole and highly affects the overall drilling performance. However, drill bit selection is mostly accomplished through lessons learned from previous runs as well as bit grading after each run. These methods are highly subjective and usually based on the engineer’s experience.
The abundance of field data with data analytics and machine learning capabilities are a perfect combination for creating reliable data-driven models. The main objective of this study is to create robust models that are able to classify the formation based on drilling parameters as well as estimate the bit dull grading based on drilling parameters and the formation. In order to achieve the aforementioned goals, the disclosed Volve filed dataset was meticulously processed and analyzed.
The models were created for each of the well sections by using the Python, especially the pandas and scikit-learn libraries. However, after running the first simulation, models usually showed unsatisfactory accuracy. In order to increase models performance, the code was written to find the best parameter for each machine learning technique. Even though the bit dull grading model has a valid algorithm, the input parameters are hard to find, due to the lack of literature and patterns.
Obtained results proved that the machine learning technique may be successfully implemented to solve the everyday problems in the oil and gas industry. Moreover, the outcome should help in the well planning process, enables to decrease the number of trips and improves overall drilling phase efficiency. The process could eliminate the trial and error drill bits selection and ensure more efficient and effective decision-making process. | nb_NO |
dc.language.iso | eng | nb_NO |
dc.publisher | University of Stavanger, Norway | nb_NO |
dc.relation.ispartofseries | Masteroppgave/UIS-TN-IEP/2019; | |
dc.subject | petroleumsteknologi | nb_NO |
dc.subject | petroleum engineering | nb_NO |
dc.subject | boreteknologi | nb_NO |
dc.subject | Bit Dull Grading | nb_NO |
dc.subject | Python | nb_NO |
dc.subject | formation classification | nb_NO |
dc.subject | data analytics | nb_NO |
dc.subject | machine learning | nb_NO |
dc.title | The Application of Data Analytics and Machine Learning for Formation Classification and Bit Dull Grading Prediction | nb_NO |
dc.type | Master thesis | nb_NO |
dc.subject.nsi | VDP::Technology: 500::Rock and petroleum disciplines: 510::Petroleum engineering: 512 | nb_NO |