Estimation of the liquefaction potential of soils of Dalaman Basin/Muğla (SW Anatolia-Turkey) using Multilayer Perceptron Feed Forward Machine Learning Technique.

Soil liquefaction is one of the secondary effects of earthquakes and defined as the decrease in the strength and stiffness of saturated soil because of the increase in pore water pressure and resulting decrease in the effective stress under dynamic loads such as earthquakes. Soil liquefaction is controlled by earthquake magnitude, groundwater level, depth of the soil layer, acceleration, soil type, depositional environment, age of soil deposit and fine particle percentage. SW Anatolian Region is controlled by different fault mechanisms including normal, strike-slip and thrust faults including Gökova Fault Zone (GFZ), Muğla (MF) and Yatağan faults (YF) of Muğla-Yatağan Fault Zone (MYFZ), Milas–Ören Fault Zone (MOFZ), Fethiye–Burdur zone (FBZ) and Hellenic Arc which are probable to generate earthquakes with great magnitudes in a time span important for population (100years). Dalaman Basin is an extensional sedimentary basin (Delta Environment) in SW Anatolian Region which is controlled by numerous active basin margin normal faults in the close proximity to the Fethiye-Burdur Fault Zone and Hellenic Arc which makes it important in terms of soil liquefaction. Machine learning techniques have been started to be used in estimation of the liquefaction potential and this study aims to estimate liquefaction potential of sandy-silty soil layers in the Dalaman Basin using Multilayer Perceptron (MLP) feed forward machine learning technique. This method includes the generation of equation using depth, SPT blow numbers, fine particle content, groundwater level, total and effective stresses, maximum acceleration, earthquake magnitudes and CSR information with liquefaction case histories after 1999 Kocaeli and Taiwan Earthquakes and estimation of the liquefaction potentials of the soils of Dalaman Basin with this pre-generated equation. Results clearly shows that Multilayer Perceptron Machine learning technique is useful in estimation of liquefaction potential.

This study has been produced from PhD thesis named as “Determination of the geo-engineering properties and liquefaction potential of the Quaternary deposits of Dalaman-Muğla/SW Anatolia”.