Prediksi Wave-induced Liquefaction dengan Artificial Neural Network dan Wide Genetic Algorithm

Kristianto, Dwi (2017) Prediksi Wave-induced Liquefaction dengan Artificial Neural Network dan Wide Genetic Algorithm. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
5115201041-Master-Theses.pdf
Restricted to Repository staff only

Download (2MB) | Request a copy

Abstract

Fenomena Liquefaction telah menjadi perhatian peneliti sejak era 1970an, mulai dari pemodelan analitik, pemodelan dengan Finite Element Method (FEM) sampai beragamnya jenis pengujian di laboratorium. Salah satu bidang dalam penelitian liquefaction adalah memprediksi lokasi kedalaman tanah tempat terjadinya liquefaction. Kerumitan pemodelan analitik, kebutuhan untuk melakukan validasi laboratorium dan inspeksi lapangan, telah membuka peluang untuk pengembangan pemodelan yang mudah, praktis, murah sekaligus valid. Dalam penelitian ini, pemodelan Artificial Neural Network (ANN) digunakan untuk memprediksi kedalaman maksimum liquefaction. ANN dipilih karena ANN mampu memodelkan interaksi komputasi paralel pada otak melalui proses learning terhadap data. Metode Back Propagation (BP) adalah metode pelatihan yang paling banyak digunakan, walaupun memiliki kelemahan mudah terjebak dalam local optimum dan tidak stabil. Untuk mengatasi masalah tersebut, metode optimasi Genetic Algorithm (GA) digunakan dalam proses pelatihan ANN. GA adalah metode optimasi yang menirukan proses evolusi, seleksi, rekombinasi dan mutasi yang terjadi di alam. Walaupun telah banyak digunakan, GA memiliki beberapa kelemahan yaitu premature convergence dan local optimum. Dalam penelitian ini dilakukan modifikasi terhadap GA, antara lain: operasi seleksi Wide Tournament, operasi rekombinasi BLX-α Multi-Parent, operasi mutasi Aggregate Mate Pool dan operasi Direct Mutation- Recombination. Modifikasi GA bertujuan untuk meningkatkan keragaman populasi, memperluas cakupan pencarian solusi dan menemukan global optimum lebih mudah dan cepat. Global optimum GA adalah konfigurasi ANN terbaik dengan error prediksi MdAPE terkecil. ================================================================== Liquefaction phenomenon has been the attention of researchers since the 1970s, ranging from analytical modeling, Finite Element Method (FEM) modeling to various types laboratory testing. One of its research field is to predict the depth where soil liquefaction occurred. The hassle of analytic modeling, repetitive laboratory testing and expensive field inspections, has opened up opportunities to develop simple, practical, inexpensive and valid modeling. In this research, Artificial Neural Network (ANN) regression modeling is used to predict the maximum depth of liquefaction. ANN selected among other Artificial Intelligence methods because its ability to model the interaction of brain parallel computing through iterative learning process of inputted data. Back Propagation (BP) is one of widely used training method for ANN. Although it has some weaknesses such as easily trapped in local optimum and unstable results, it is most widely used one. To overcome BP weaknesses, Genetic Algorithm (GA) is being used for ANN learning process. GA is optimization method that mimics the process of evolution, selection, recombination and mutation that occurs in nature. Although GA is a well known method, it also has some weaknesses, such as premature convergence and local optimum. In this research, GA is heavily modified to overcome its weaknesses. The modifications are Wide Tournament selection, Multi-Parent BLX-α recombination, Aggregate mutation and Direct Mutation-Recombination operation. The modifications aim to increase the population diversity, expanding the coverage area of solution search and find global optimum solution more easily and quickly. Global optimum solution of GA is best ANN configuration which has smallest MdAPE prediction errors.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Wave-induced Liquefaction; Algoritma Genetika Lebar; Jaringan Saraf Tiruan; seleksi Wide Tournament; rekombinasi BLX-α Multi Induk; mutasi Aggregate Mate Pool; Direct Mutation-Recombination; Artificial Neural Network; Wide Genetic Algorithm; Wide Tournament selection; BLX-α Multi-Parent recombination; Aggregate mutation
Subjects: Q Science > QA Mathematics
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Z Bibliography. Library Science. Information Resources > ZA Information resources > ZA4050 Electronic information resources
Divisions: Faculty of Information Technology > Information System > (S2) Master Theses
Depositing User: DWI KRISTIANTO
Date Deposited: 31 Mar 2017 04:23
Last Modified: 22 Dec 2017 01:19
URI: http://repository.its.ac.id/id/eprint/2664

Actions (login required)

View Item View Item