Application of artificial neural networks for predicting soil settlement in geotechnical applications with plastic waste reinforcement above buried pipes

Abstract The inevitability of employing shallow buried pipes in urban areas, roadways, and subways is attributed to the progress of development and population expansion. This paper investigates the impact of recycled plastic bottle utilization on soil settlement above buried pipes under static loads and employs a two-hidden-layer artificial neural network (ANN) model to accurately predict settlement values. Experimental measurements of settlement are conducted under various reinforcement conditions and applied pressures and resulted a dataset of 72 data points, which was divided into 70% for training and 30% for testing using a holdout validation approach. The results demonstrate significant reductions in settlement with plastic waste reinforcement, with mattress depth to width of the loading steel plate reinforcement ratios u/B = 0.5, u/B = 1.0, and u/B = 1.5 exhibiting settlement reductions of 0.25 mm, 2.3 mm, and 4.5 mm, respectively, compared to the unreinforced condition. The ANN model, configured with two hidden layers of 10 and 6 neurons respectively, had used the hyperbolic tangent (tanh) activation function and trained with the Levenberg–Marquardt algorithm. The R 2 values reached 0.9990 for training and 0.9965 for testing, while the Root Mean Square Error (RMSE) was maintained at 0.021% for training and 0.034% for testing which indicates minimal deviation between predicted-observed settlements. The findings highlight the practical significance of plastic bottle reinforcement as an efficient and sustainable soil improvement technique for minimizing settlement above buried pipes. Despite the high accuracy, the study acknowledges limitations related to static loading conditions, sandy SP soil type, and a relatively small experimental dataset. Future research is recommended to explore dynamic and cyclic loading scenarios, assess long-term PET degradation effects, and validate the reinforcement approach under diverse soil conditions.