A robust hybrid near-real-time model for prediction of drilling fluids filtration; Engineering with Computers; Vol. 41

Dettagli Bibliografici
Parent link:Engineering with Computers.— .— Berlin: Springer Nature
Vol. 41.— 2025.— 26 p.
Altri autori: Davoodi Sh. Shadfar, Al-Shargabi M. A. T. S. Mokhammed Abdulsalam Takha Sallam, Wood D. A. David, Mehrad M. Mohammad, Rukavishnikov V. S. Valery Sergeevich
Riassunto:Title screen
Efficient drilling operations demand careful preparation, especially to ensure the optimal filtration characteristics of drilling fluids to prevent issues like formation damage. Monitoring changes in fluid filtration volume (FV) is critical for maintaining wellbore stability, preventing rock damage, and lowering drilling costs. This study employs deep learning (DL) models, specifically Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) hybridized with the grey wolf optimizer (GWO), to accurately predict FV using daily measurements of fluid density (FD) and Marsh funnel viscosity (MFV). A field dataset of 1260 records from 17 wells in southwest Iran is analyzed to explore the relationship between fluid properties and their impact on FV. Results reveal that the LSTM-GWO model outperforms other algorithms, achieving the lowest root mean squared error (RMSE) of 1.0950 mL for FV prediction on test data. In comparison, LSTM and CNN models record higher RMSE values of 1.9963 mL and 2.2862 mL, respectively, whereas a hybrid CNN-GWO model achieves an RMSE of 1.3551 mL. Sensitivity analysis indicates that the LSTM-GWO model is efficient at capturing relationships between inputs and FV, suggesting an ability to learn complex non-linear patterns from the input data. Further analyses confirmed the robustness of the hybrid models, revealing that they exhibited greater resilience compared to basic DL models. The proposed hybrid DL approach presents a promising methodology for accurate and near-real-time drilling fluid FV prediction, addressing current limitations in traditional empirical and/or analytical methods
Текстовый файл
AM_Agreement
Lingua:inglese
Pubblicazione: 2025
Soggetti:
Accesso online:https://doi.org/10.1007/s00366-025-02113-3
Natura: Elettronico Capitolo di libro
KOHA link:https://koha.lib.tpu.ru/cgi-bin/koha/opac-detail.pl?biblionumber=679533

MARC

LEADER 00000naa0a2200000 4500
001 679533
005 20250407142323.0
090 |a 679533 
100 |a 20250407d2025 k||y0rusy50 ba 
101 0 |a eng 
102 |a DE 
135 |a drcn ---uucaa 
181 0 |a i   |b  e  
182 0 |a b 
183 0 |a cr  |2 RDAcarrier 
200 1 |a A robust hybrid near-real-time model for prediction of drilling fluids filtration  |f Shadfar Davoodi, Mohammed Al-Shargabi, David A. Wood [et al.] 
203 |a Текст  |b визуальный  |c электронный 
283 |a online_resource  |2 RDAcarrier 
300 |a Title screen 
320 |a References: 71 tit 
330 |a Efficient drilling operations demand careful preparation, especially to ensure the optimal filtration characteristics of drilling fluids to prevent issues like formation damage. Monitoring changes in fluid filtration volume (FV) is critical for maintaining wellbore stability, preventing rock damage, and lowering drilling costs. This study employs deep learning (DL) models, specifically Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) hybridized with the grey wolf optimizer (GWO), to accurately predict FV using daily measurements of fluid density (FD) and Marsh funnel viscosity (MFV). A field dataset of 1260 records from 17 wells in southwest Iran is analyzed to explore the relationship between fluid properties and their impact on FV. Results reveal that the LSTM-GWO model outperforms other algorithms, achieving the lowest root mean squared error (RMSE) of 1.0950 mL for FV prediction on test data. In comparison, LSTM and CNN models record higher RMSE values of 1.9963 mL and 2.2862 mL, respectively, whereas a hybrid CNN-GWO model achieves an RMSE of 1.3551 mL. Sensitivity analysis indicates that the LSTM-GWO model is efficient at capturing relationships between inputs and FV, suggesting an ability to learn complex non-linear patterns from the input data. Further analyses confirmed the robustness of the hybrid models, revealing that they exhibited greater resilience compared to basic DL models. The proposed hybrid DL approach presents a promising methodology for accurate and near-real-time drilling fluid FV prediction, addressing current limitations in traditional empirical and/or analytical methods 
336 |a Текстовый файл 
371 0 |a AM_Agreement 
461 1 |t Engineering with Computers  |c Berlin  |n Springer Nature 
463 1 |t Vol. 41  |v 26 p.  |d 2025 
610 1 |a электронный ресурс 
610 1 |a труды учёных ТПУ 
610 1 |a Filtration volume 
610 1 |a Hybrid deep-learning 
610 1 |a Fluid density 
610 1 |a Marsh funnel viscosity 
610 1 |a Long short-term memory 
610 1 |a Grey wolf optimizer 
701 1 |a Davoodi  |b Sh.  |c specialist in the field of petroleum engineering  |c Research Engineer of Tomsk Polytechnic University  |f 1990-  |g Shadfar  |9 22200 
701 1 |a Al-Shargabi  |b M. A. T. S.  |c specialist in the field of petroleum engineering  |c Engineer of Tomsk Polytechnic University  |f 1993-  |g Mokhammed Abdulsalam Takha Sallam  |9 22768 
701 1 |a Wood  |b D. A.  |g David 
701 1 |a Mehrad  |b M.  |g Mohammad 
701 1 |a Rukavishnikov  |b V. S.  |c Director of the Center for Training and Retraining of Oil and Gas Specialists, Associate Professor of Tomsk Polytechnic University, Candidate of Technical Sciences  |c Engineer of Tomsk Polytechnic University  |f 1984-  |g Valery Sergeevich  |9 17614 
801 0 |a RU  |b 63413507  |c 20250407 
850 |a 63413507 
856 4 |u https://doi.org/10.1007/s00366-025-02113-3  |z https://doi.org/10.1007/s00366-025-02113-3 
942 |c CF