Variable Significance Determination Utilizing Extended CHAID Method in Fiber Improvement of Fine Soils

Abstract

Adding fiber to reinforce soil is a technique widely used at geotechnical works due to the possibility of a considerable increase in soil’s mechanical properties. One way to evaluate such improvement is to compare the features before and after the addition, utilizing the unconfined compressive strength (UCS) test. However, the results of this test can be biased by soil variability, fibers characteristics, and test errors. The objective of the present work is to analyze which variable most influences the results of UCS tests and evaluate the relationship between the soil strain and its strength alterations due to fiber. In this way, the extended chi-square automatic interaction detector (CHAID) method was adopted on UCS samples obtained from a database established from papers, theses, and dissertations. This process concluded that the type of fiber used is the variable with the highest significance among the studied ones. Furthermore, the regression of the polypropylene (PP) demonstrated a linear tendency and a strong correlation coefficient, validating the hypothesis that the sample’s strength has an immediate relation with its strain.