Cement production is an energetically demanding process and one of the most important sources of CO2 emission in the world. Therefore, replacing part of the cement with more environmentally friendly supplementary materials such as natural zeolite (NZ) is of great importance. The small strain shear modulus (G0) of soils is an essential parameter in many aspects of geotechnical engineering. In the present study, ninety-two bender element tests on loose sandy soils grouted with cement and NZ was conducted to obtain equations for predicting the G0 of grouted sands. Multiple linear regression (MLR) and Group Method of Data Handling (GMDH) polynomial NN were used to estimate the G0 of zeolite–cement grouted sands. The G0 was modeled as a function of the average sand grain size (D50) as well as the percentages of cement replacement with zeolite (Z) and water to cementitious material ratios (W/CM) of grout. Also, the parameter active compounds (AC) was considered as one of the input parameters instead of the simultaneous effect of W/CM and Z. The results showed that both MLR and GMDH models, in which AC was considered as an input parameter to predict the G0 of grouted sands, has a much better performance than the same models with W/CM and Z as input parameters. The GMDH-based equations are more efficient than the MLR-based equations. Applying the AC as an input parameter to predict the G0 of grouted sands by the GMDH models leads to about 35–41% improvement in G0 prediction. Therefore, the AC is a very effective parameter for estimating the G0 of the grouted sands. The GMDH-based predicted G0 is significantly affected by changes in zeolite percentage and the effect of sand particles’ size variation on the accuracy and variation of the predicted G0 is less than the other two parameters (W/CM and Z).
Read full abstract