Application of Response Surface Method in Predicting and Optimizing the Engineering Properties of Activated Metakaolin Treated Non-Lateritic Soil

Abstract

This study focuses on using response surface methodology (RSM) to predict and optimize the unconfined compressive strength (UCS) and California bearing ratio (CBR) of activated metakaolin (MK) treated non-lateritic soil for road construction purposes. The experimental results of various blends compacted using British standard light, West African standard, and British standard heavy methods were used to create a useful model for overall response variation. The design consists of two design factors, MK and sodium hydroxide (SH), with MK and SH as independent variables, and UCS and CBR as the responses. Predictive equations for the responses were obtained using the independent variables. Statistical analysis and analysis of variance for all responses showed that quadratic models were successful in predicting the UCS and CBR of activated MK-treated non-lateritic soil with R² (0.9835-0.9999), Adj R² (0.9718-0.9999), and Pred R² (0.8776-0.9998). F-values are greater than the critical F-value (3.59), indicating that the factors have a significant effect on the model behaviour. The P value of the models was less than P_α(0.05), indicating that the factors are significant in predicting the responses. Furthermore, optimized factors were predicted to obtain optimal values for UCS and CBR that met the Nigerian General Specification for road base course usage. These predictions were validated, and a good correlation was observed between the experimental and predicted values, as judged by the absolute relative percent error (0.0232–1.1628). The proposed models are capable of predicting the UCS and CBR values, which can help make early decisions during the construction process.