Performance of Artificial Neural Network and Modified Gravitational Search Algorithm Models to Predict Vibration Response of Geocell Reinforced Based Weak Sand

Abstract

The use of a fast evolving artificial intelligence technology (AIT) to forecast the vibration response of maritime soil based on geocells is explored in this research. The vibration response is represented by an indicator called peak particle velocity (PPV). For the purpose of predicting PPV, the artificial intelligence techniques Artificial Neural Network (ANN) and Modified Gravitational Search Algorithm (MGSA) are employed. To create the dataset for the model, a number of field vibration tests were first performed over the geocell-reinforced beds. PPV variation was investigated by varying the test variables—footing embedment, dynamic load, infill material modulus, width, and depth of geocell mattress placement—during the test performed. The various statistical indicators were determined in order to evaluate the prediction performance of a constructed model. Plate load results on geocell-reinforced foundation beds have been used to validate the proposed hybrid ANN-MGSA model. High accuracy and consistency were found when the findings of the ANN-EHO, JSA, MOA, and RNN method were compared, particularly at predicted and actual resolution levels. A parametric sensitivity has also been examined in order to better understand the behaviour of geocell-reinforced structures.