M. Vahdani - Faculty of Civil Engineering, K. N. Toosi University of Technology, Tehran, Iran
M. Ghazavi - Faculty of Civil Engineering, K. N. Toosi University of Technology, Tehran, Iran
M. Roustaei - Civil Engineering Department, Islamic Azad University, Qazvin Branch, Qazvin, Iran

The present study was based on a promoting statistical method known as response surface method (RSM). RSM has been applied as an efficient method to optimize many physical applications in industry for more than two decades. In the current study, the RSM was utilized as a platform to develop models as a function of some prescribed input factors to predict mechanical properties (responses) of frozen soils (i.e. peak tensile/compressive strength, elasticity modulus). Besides, RSM makes it possible to find significant factors and probable interactions as well. A widespread literature review was conducted and three case studies were chosen to evaluate the performance of the RSM in developing precise models and finally an optimum experiment. For each case study, less than half of the available data (an average of ۴۰.۸%) was employed to develop models and the remaining part was employed to evaluate the validity of derived models. A comparison between predicted and measured data showed a good agreement with a significant level of ۰.۰۵. This indicates that upon using the model a hundred times to predict an specific property for different input factors, the maximum five predictions may diverge from the measured values with ± confidence interval. In addition, some contours were plotted to give a comprehensive presentation of any probable correlations between investigated properties and input factors. Based on the developed models with an average correlation coefficients (R۲) of ۹۳.۶۹, temperature was found to be the most significant factor affecting the mechanical properties of frozen fine soil, while the dry density was not as effective as the temperature.

Frozen soil, Mechanical properties, Prediction models, response surface method, Statistical optimization