Soil Shear Strength Prediction Model based on Artificial Neural Network and Osprey Optimization Algorithm

Abstract

Soil shear strength is an essential parameter in geotechnical and civil engineering that is measured and used when building structures like retaining walls, pavements, and dams. The two factors that determine a soil sample's shear strength are internal friction and cohesiveness. When soil is subjected to a load, its shear strength determines how well it can tolerate internal movement and slippage. Thus, an infrastructure's capacity to endure damage is determined by its shear strength. Lab calculations may be used to determine a soil sample's shear strength, which is influenced by several variables including moisture content, plastic index, and liquid limit. However, estimating soil shear strength in labs is time-consuming and expensive due to instrument handling issues and lengthy measurement methods for consistent and precise data. Thus, soil shear strength may be precisely and quickly computed using artificial intelligence.In this research, a soil shear strength prediction model is designed with the help of artificial neural network (ANN) and osprey optimization algorithm (OOA). The ANN algorithm is utilized for reduce the error between the actual and predicted value in proposed model in order to enhance the accuracy of the model whereas osprey optimization algorithm is utilized for determine the best weight values of the ANN algorithm based on the objective function. In this work, root mean square error (RMSE) is taken as the objective function. Besides that, pre-processing of the dataset is done using the principal component analysis (PCA) in orders to reduce the dataset dimension. The simulation evaluation is done on the standard dataset which contains 12 attributes related to soil and one attribute is related to strength. The result shows that the proposed model achieves the minimum error between actual and predicted value. Besides that, convergence rate graph shows the osprey optimization algorithm is quickly find the optimal results. Finally, different parameters such as RMSE, MAPE, MAE, and determination of coefficient is measured for the proposed method and result shows that RMSE, MAPE, and MAE value is lower and determination of coefficient higher value over the existing models.