Prediction of Soil Plasticity Index with the Use of Regression Analysis and Artificial Neural Networks: A specific case for Bakırköy District

Akbay Arama Z. , Yücel M., Akın M. S. , Nuray S. E. , Alten O.

6th International Conference on Harmony Search, Soft Computing and Applications, İstanbul, Türkiye, 16 - 17 Temmuz 2020, ss.1-2

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.1-2


Consistency limits are simple and precious soil parameters that are directly representing the mineralogical and physical properties and indirectly used to estimate the strength and rigidity characteristics of fine grained soils. The consistency tests are assumed as the easiest and the basic tests of geotechnical engineering and the application details of the tests are depending on the experience of the operator and the environment conditions. Therefore, the reliability and correctness of the tests are still in the concern of geotechnical studies. Besides, the development of information and computer technologies also present alternative solution perspectives for the attainment of geotechnical properties. In this connection, the regression analysis and artificial neural network applications become the most preferred techniques to estimate the consistency limits with the use of an only a variable.  In this study, it is aimed to obtain the consistency properties of high plastic clays that are procured from the site investigation reports conducted for Bakırköy district (Istanbul, Turkey) depending on only the evaluation of liquid limit test results. A database is arranged with the use of approximately 450 consistency limit tests. Linear and multi-linear regression analyses and artificial neural network applications were conducted with the use of different variants such as depth and fine content to determine the plastic limit and the plasticity index value. As a result, comparisons of the applied techniques are done to query the appropriateness of the methods and to find the approximation rate to the actual value. Consequently, the success of the applied techniques was tried to be shown with the developed equation set.