Application of machine learning algorithm for the estimation of time-dependent strength of basic oxygen furnace slag-treated soil

Abstract

The main purpose of this study is to predict the time-dependent strength of BOF slag-treated dredged soil using four machine learning (ML) algorithms (random forests, multi-layer perceptron, support vector regression, k-nearest neighbors). These models were trained using a dataset developed from the published literature. The slag type, slag content, water content, and curing time were used as input values. Here, the curing time was divided into three stages according to the magnitude of strength development. Among the algorithms, the multi-layer perceptron (MLP) was selected as the optimal model, and its predicted strength was compared with that of BOF slag-treated soil calculated by the previous empirical equation. In addition, MLP accurately predicted the strength of BOF slag-treated soil compared with that of the empirical equation. Consequentially, ML algorithms had higher applicability for estimating of the time-dependent strength of BOF slag-treated soils.

Keywords

• BOF slag
• Dredged clay
• Time-dependent strength
• Prediction
• Machine learning algorithm
• Empirical equation