Strength Prediction Of Cement Stabilized Clay Using Artificial Neural Networks

Suleiman Khatrush, Zainab ALmahmoudi

Artificial Neural Networks (ANN), cement stabilization, unconfined compressive strength, clay soil

Stabilization of clay soils is necessary in many civil engineering projects in order to increase strength, reduce settlement and also for other special purposes, especially when weak soils do exist. Soil treatment with cement is one of the most commonly used method, it is proved to be an efficient and effective chemical stabilization method due to its economic advantages and ease of use. In this research, Test data sets with a wide range of parameters were implemented in an Artificial Neural Networks (ANN) program in order to evaluate the unconfined compressive strength of cementations clay soils. The data were collected from the selected published laboratory experimental investigations conducted by many researchers to study the effect of various parameters on the strength improvement of cement treated clay. The selected data were chosen to represent a wide range of clayey soils obtained from different places around the world. The predictive model was developed using the artificial neural network tools in MATLAB software. The artificial neural network (ANN) technique was applied using the Levenberg-Marquardt algorithm to develop a model that predicts the unconfined compressive strength of cement treated clayey soils. The number of data sets for this study were (429) collected from (15) previous research studies. Eight input parameters were chosen as follows: Liquid limit (LL)%, Plasticity index (PI)%, Clay fraction (CF)%, Sand (S)%, silt (M)%, water content (Wc) % , curing time (Tm) in days and cement content (Cc)% The unconfined compressive strength (UCS) was chosen as one output parameter, then the data was normalized using the min-max method. In order to evaluate the performance of the predictive model developed in this study, the statistical analysis of the model was performed using regression (R2), mean square error (MSE), root mean square error (RMSE), and coefficient of efficiency (CE). Good correlation was obtained with regression (R2) of 0.897. Sensitivity analysis indicate that the cement content is mostly affecting the resulting soil strength (UCS) followed by water content, curing time and liquid limit. The rest of the variables show relatively lower impact.

https://dx.doi.org/10.31873/IJEAS.12.08.01