Scholarlink

<p>The traditional shale brittleness evaluation method based on elastic parameters and mineral composition cannot fully consider the influence of laminated features on on-site operations, and laminated features have a key impact on crack propagation and engineering sweet spot evaluation. Therefore, these traditional methods are not suitable for classifying and evaluating the engineering optimal points in layered shale reservoirs. In this study, we conducted scratch tests on the entire wellbore section of layered shale, obtaining high-resolution continuous strength profiles and finely characterizing strength fluctuations under bedding control. The optimal configuration of hidden layers and learning rates was determined through Bayesian hyperparameter optimization, and then a long short-term memory (LSTM) neural network was applied to integrate scratch test data with well logging curves. This combination enables the model to simultaneously consider the cyclic non-uniformity of compressive strength and the anisotropy of thin laminae. The scratch test results show that the clayey shale exhibits a higher crack density (approximately 8 cracks/cm, with an average crack width of about 3&#xa0;mm), displaying distinct high-frequency thin interlayer structures across the entire scratch surface. Numerous damaged surfaces appear on both sides of the scratch, forming a “fishbone” pattern. In contrast, the feldspar–quartz and mixed shales exhibit lower crack densities (around 2 cracks/cm, with an average crack width of about 0.8&#xa0;mm), resulting in relatively smoother scratch surfaces. This paper provides insights into the macroscopic mechanical properties of laminated shale and delivers a more reliable tool for brittleness evaluation and sweet spot characterization, offering valuable guidance for optimizing shale oil and gas development and engineering decisions.</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

Bayesian-Optimized LSTM for Full-Borehole Brittleness Prediction by Integrating Scratch Data and Wireline Logging Data

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • Liu Yang,
  • Zhenchuan Han,
  • Yunhui Lu,
  • Ce Li,
  • Xuli Guo,
  • Hongjian Wang

摘要

The traditional shale brittleness evaluation method based on elastic parameters and mineral composition cannot fully consider the influence of laminated features on on-site operations, and laminated features have a key impact on crack propagation and engineering sweet spot evaluation. Therefore, these traditional methods are not suitable for classifying and evaluating the engineering optimal points in layered shale reservoirs. In this study, we conducted scratch tests on the entire wellbore section of layered shale, obtaining high-resolution continuous strength profiles and finely characterizing strength fluctuations under bedding control. The optimal configuration of hidden layers and learning rates was determined through Bayesian hyperparameter optimization, and then a long short-term memory (LSTM) neural network was applied to integrate scratch test data with well logging curves. This combination enables the model to simultaneously consider the cyclic non-uniformity of compressive strength and the anisotropy of thin laminae. The scratch test results show that the clayey shale exhibits a higher crack density (approximately 8 cracks/cm, with an average crack width of about 3 mm), displaying distinct high-frequency thin interlayer structures across the entire scratch surface. Numerous damaged surfaces appear on both sides of the scratch, forming a “fishbone” pattern. In contrast, the feldspar–quartz and mixed shales exhibit lower crack densities (around 2 cracks/cm, with an average crack width of about 0.8 mm), resulting in relatively smoother scratch surfaces. This paper provides insights into the macroscopic mechanical properties of laminated shale and delivers a more reliable tool for brittleness evaluation and sweet spot characterization, offering valuable guidance for optimizing shale oil and gas development and engineering decisions.