Soft computing techniques to estimate the uniaxial compressive strength of mortar incorporated with cement kiln dust

In the current study, the effect of cement kiln dust on the fresh and hardened cement mortar was investigated; chemical analysis of the binder was also included. Cement kiln dust (CKD) content (% by dry weight of cement) was from 0 to 100%. The fresh property of the cement was evaluated by flow table test, and the evaluation of hardened properties was based on the compressive strength, flexural strength, and stress–strain behavior of the mortar. Standard sand was used in the current study with a ratio of sand-to-binder ( s / b ) 3 and a water-to-binder ratio of 0.5. Also, cement and CKD were characterized based on microstructure tests, X-ray diffraction (XRD) and scanning electron microscopy, and thermogravimetric analysis to determine the weight loss of CKD and cement under high temperature. Three distinct models (linear regression model (LR), adaptive regression spline (MARS), and artificial neural network (ANN)) were utilized to generate predictive models to estimate the compressive strength of CKD-modified cement mortar, the current study data and 162 collected data from different research studies were used in the model development. The collected data were a combination of datasets with different w / b , s / b , CKD, silicon dioxide content in the binder (SiO 2 ), calcium oxide content in the binder (CaO, %), maximum size of fine aggregate (MSA, mm), and curing times of the samples (t, days). Additionally, the coefficient of determination ( R 2 ), scatter index, mean absolute error, and mean absolute percentage error were used to evaluate the effectiveness of the generated models. According to the results of experimental work, increasing CKD content decreased the compressive strength and flexural strength of cement mortar. Furthermore, the modeling analysis showed that the ANN model was better than the LR and MARS model for predicting the compressive strength of CKD-modified cement mortar.