Scholarlink

<p>This study evaluates the stabilization of expansive black cotton (BC) soil for pavement subgrade applications using cement partially substituted with flue gas desulfurization (FGD) gypsum, an industrial by-product. A comprehensive experimental program was conducted to assess the geotechnical, mechanical and durability performance of treated soil through Atterberg limits, compaction, unconfined compressive strength (UCS), California bearing ratio (CBR), permeability, fatigue response and microstructural characterization. Stabilized specimens were prepared with varying cement and FGD gypsum dosages and cured for different durations. The results demonstrated substantial improvement in the engineering behaviour of BC soil after stabilization. Plasticity characteristics were significantly reduced, while strength, permeability and fatigue resistance were markedly enhanced relative to untreated soil. UCS values ranged up to 585&#xa0;kPa depending on mix composition and curing duration, highlighting the combined influence of binder composition and curing period. Among the mixtures investigated, the blend containing 6% cement and 3% FGD gypsum exhibited the most balanced and superior overall performance. All stabilized mixes satisfied the minimum CBR requirement (≥ 8%) specified for pavement subgrade materials in IRC:37–2018. To complement the experimental investigation, an artificial neural network (ANN) model was developed to predict UCS using cement content, FGD gypsum dosage and curing period as input variables. The optimal 3–8–1 ANN architecture, trained using the Quasi-Newton algorithm with BFGS Hessian approximation and Brent’s adaptive learning-rate optimization, converged within 16 epochs and yielded low mean squared errors of 0.0635 and 0.0678 for training and validation, respectively. The ANN model was selected due to its capability to capture complex nonlinear relationships between stabilization parameters and strength characteristics. However, the model is developed based on a limited dataset and is applicable only within the parameter ranges studied. Future work should include comparison with other machine learning models and validation using extended datasets to improve generalization. The study demonstrates the potential for a sustainable and technically viable approach for resilient pavement subgrade development, contributing to SDGs 9, 11, 12 and 13.</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

Multiscale experimental characterization and ANN-based modeling of cement–FGD gypsum stabilized black cotton soil for subgrade applications

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • Sridhar Halasinahally Ningegowda,
  • Shiva Kumar Govindaraju,
  • Chidananda M. Linganagoudar,
  • Ujwal. Mallaiah Sudhamani,
  • Poornachandra Pandit

摘要

This study evaluates the stabilization of expansive black cotton (BC) soil for pavement subgrade applications using cement partially substituted with flue gas desulfurization (FGD) gypsum, an industrial by-product. A comprehensive experimental program was conducted to assess the geotechnical, mechanical and durability performance of treated soil through Atterberg limits, compaction, unconfined compressive strength (UCS), California bearing ratio (CBR), permeability, fatigue response and microstructural characterization. Stabilized specimens were prepared with varying cement and FGD gypsum dosages and cured for different durations. The results demonstrated substantial improvement in the engineering behaviour of BC soil after stabilization. Plasticity characteristics were significantly reduced, while strength, permeability and fatigue resistance were markedly enhanced relative to untreated soil. UCS values ranged up to 585 kPa depending on mix composition and curing duration, highlighting the combined influence of binder composition and curing period. Among the mixtures investigated, the blend containing 6% cement and 3% FGD gypsum exhibited the most balanced and superior overall performance. All stabilized mixes satisfied the minimum CBR requirement (≥ 8%) specified for pavement subgrade materials in IRC:37–2018. To complement the experimental investigation, an artificial neural network (ANN) model was developed to predict UCS using cement content, FGD gypsum dosage and curing period as input variables. The optimal 3–8–1 ANN architecture, trained using the Quasi-Newton algorithm with BFGS Hessian approximation and Brent’s adaptive learning-rate optimization, converged within 16 epochs and yielded low mean squared errors of 0.0635 and 0.0678 for training and validation, respectively. The ANN model was selected due to its capability to capture complex nonlinear relationships between stabilization parameters and strength characteristics. However, the model is developed based on a limited dataset and is applicable only within the parameter ranges studied. Future work should include comparison with other machine learning models and validation using extended datasets to improve generalization. The study demonstrates the potential for a sustainable and technically viable approach for resilient pavement subgrade development, contributing to SDGs 9, 11, 12 and 13.