Experimental and Machine Learning-Based Prediction of Bamboo Geocell-Reinforced Pond Ash Behaviour under Triaxial Loading

Abstract

The disposal of pond ash generated from coal-based thermal power plants has become a serious environmental concern, while the depletion of natural construction materials has encouraged the use of sustainable alternatives in geotechnical engineering. This study investigates the unconsolidated undrained triaxial behaviour of unreinforced and bamboo-reinforced pond ash specimens. Triaxial tests were conducted on specimens of 38 mm diameter and 76 mm height under confining pressures of 100, 200, and 300 kPa. Two bamboo reinforcement forms, namely planar reinforcement and unit geocell reinforcement, were used. Geocells of 25 mm diameter with heights of 2.5, 5, 7.5, and 10 mm, along with planar reinforcements of 25 mm diameter and 0.5 mm thickness, were placed at mid-height as single-layer reinforcement and at 1/3rd height as double-layer reinforcement. Machine learning-based regression modelling was also incorporated to predict peak deviator stress using confining pressure, reinforcement type, geocell height, number of layers, and reinforcement location as input variables. The peak deviator stress increased from 929.19 kPa for unreinforced pond ash to 1701.21 kPa for pond ash reinforced with 10 mm double-layer bamboo geocell at 300 kPa confining pressure, showing an improvement of 83.09%. The 10 mm bamboo geocell increased friction angle and cohesion by 10.38% and 195.03%, respectively, for double-layer reinforcement. Polynomial regression showed the best prediction performance, with R² = 0.998, RMSE = 16.24 kPa, MAE = 12.80 kPa, and MAPE = 1.35%. The findings indicate that bamboo geocell-reinforced pond ash is a sustainable geomaterial, and machine learning can support preliminary strength prediction.