Prediction of critical total drawdown in sand production from gas wells: Machine learning approach

Fahd Saeed Alakbari, Mysara Eissa Mohyaldinn, Mohammed Abdalla Ayoub, Ali Samer Muhsan, Said Jadid Abdulkadir, Ibnelwaleed A. Hussein, Abdullah Abduljabbar Salih

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)


Sand production is a critical issue in petroleum wells. The critical total drawdown (CTD) is an essential indicator of the onset of sand production. Although some models are available for CTD prediction, most of them are proven to lack accuracy or use commercial software. Furthermore, the previous correlations have not studied the trend analysis to verify the correct relationships between the parameters. Therefore, this study aims to build accurate and robust models for predicting CTD using response surface methodology (RSM) and support vector machine (SVM). The RSM is utilized to obtain the equation without using any software. The SVM model is an alternative method to predict the CTD with higher accuracy. This study used 23 datasets to develop the proposed models. The CTD is a strong function of the total vertical depth, cohesive strength, effective overburden vertical stress, and transit time with correlation coefficients (R) of 0.968, 0.963, 0.918, and −0.813. Different statistical methods, that is, analysis of variance (ANOVA), F-statistics test, fit statistics, and diagnostics plots, have shown that the RSM correlation has high accuracy and is more robust than correlations reported in the literature. Moreover, trend analysis has proven that the proposed models ideally follow the correct trend. The RSM correlation decreased the average absolute percent relative error (AAPRE) by 12.7% compared to all published correlations' AAPRE of 22.6%–30.4%. The SVM model has shown the lowest AAPRE of 6.1%, with the highest R of 0.995. The effects of all independent variables on the CTD are displayed in three-dimensional plots and showed significant interactions.

Original languageEnglish
Pages (from-to)2493-2509
Number of pages17
JournalCanadian Journal of Chemical Engineering
Issue number5
Publication statusPublished - May 2023
Externally publishedYes


  • critical total drawdown
  • machine learning
  • sand control
  • sand management

ASJC Scopus subject areas

  • General Chemical Engineering


Dive into the research topics of 'Prediction of critical total drawdown in sand production from gas wells: Machine learning approach'. Together they form a unique fingerprint.

Cite this