Microtremor-based Soil Liquefaction Potential Assessment Using Artificial Neural Network

Abstract

The soil liquefaction assessments that are widely used in recent practices demand a soil parameters obtained from Standard Penetration Test (SPT) or Cone Penetration Test (CPT) This research suggests a new approach to do soil liquefaction assessment with the microtremor data which is easier and more economic The analysis presented in this research consists of 93 spots within the southern part of Tainan City Taiwan where borehole data has been obtained and microtremor data is recorded The final model obtained from Artificial Neural Network (ANN) will explain the relationship between microtremor parameters as the input and liquefaction potential as the output Input parameters of microtremor data consists of (1) amount of high peak occurrence (2) amount of low peak occurrence (3-7) frequencies of each peak (8) amount of valley occurrence (9-11) frequencies of each valley with the total of 11 input parameters The output target is derived from SPT data to calculate soil liquefaction Potential Index (LPI) that can be categorized into (1) Non-liquefiable (LPI5) Variation in ANN training covers different amounts of input parameters ranging from 3 to 11 parameters The best performance is shown in 10 input parameters (except input 8) with the success rate of 85 2% The result shows that microtremor observation with the help of ANN can predict liquefaction potential

Date of Award	2018 Feb 21
Original language	English
Supervisor	Jian-Hong Wu (Supervisor)