Scholarlink

<p>Shear wave velocity (Vs) and static Young’s modulus (Est) are important parameters in petroleum engineering. Many empirical models have been introduced to determine these parameters, but each of these is appropriate for a specific area. This study, aims to present and develop 9 deep learning (DL) algorithms including Multi-Layer Perceptron (MLP), Long Short-Term Memory networks (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN) and hybrid algorithms including LM (LSTM + MLP), GM (GRU + MLP), CM (CNN + MLP), CG (CNN + GRU) and CL (CNN + LSTM) to estimate these parameters using well logs data. The values of Vs and Est were estimated using the nine algorithms and data from four selected logs. Some statistical measures like, mean square error (MSE), root mean square error (RMSE) and coefficient of determination (R<sup>2</sup>) were employed to validate the results. The obtained results indicate that all proposed DL models achieved sufficient accuracy in predicting Vs and Est ​values. However, the hybrid CL and CG algorithms outperformed the others, delivering the most precise predictions. A comparative analysis of the CL and CG algorithms reveals that their results are nearly identical, however, the CG algorithm offers certain advantages over the CL algorithm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid Deep Learning Frameworks for Predicting Shear Wave Velocity and Young’s Modulus from Well Log Data

  • 期刊论文

  • 发表日期：

    2026年1月6日

查看全文
  • Farhad Mollaei,
  • Ali Moradzadeh,
  • Reza Mohebian

摘要

Shear wave velocity (Vs) and static Young’s modulus (Est) are important parameters in petroleum engineering. Many empirical models have been introduced to determine these parameters, but each of these is appropriate for a specific area. This study, aims to present and develop 9 deep learning (DL) algorithms including Multi-Layer Perceptron (MLP), Long Short-Term Memory networks (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN) and hybrid algorithms including LM (LSTM + MLP), GM (GRU + MLP), CM (CNN + MLP), CG (CNN + GRU) and CL (CNN + LSTM) to estimate these parameters using well logs data. The values of Vs and Est were estimated using the nine algorithms and data from four selected logs. Some statistical measures like, mean square error (MSE), root mean square error (RMSE) and coefficient of determination (R2) were employed to validate the results. The obtained results indicate that all proposed DL models achieved sufficient accuracy in predicting Vs and Est ​values. However, the hybrid CL and CG algorithms outperformed the others, delivering the most precise predictions. A comparative analysis of the CL and CG algorithms reveals that their results are nearly identical, however, the CG algorithm offers certain advantages over the CL algorithm.