Indarwatin, Arin Siska (2019) Penyelesaian Persamaan Diferensial Parsial Dengan Pendekatan Artificial Neural Network. Sarjana thesis, Universitas Brawijaya.
Abstract
Penyelesaian persamaan diferensial parsial merupakan model matematis yang diterapkan untuk menyelesaikan kasus fisis yang didasarkan pada prinsip atau hukum tertentu yang menyatakan keadaan alami suatu materi. Pada penelitian ini, dibuat suatu model yang digunakan untuk menentukan solusi persamaan diferensial parsial dengan pendekatan artificial neural network. Ide ini didasari dari konsep feedforward neural network yang memetakan fungsi tertentu secara maju tanpa ada koneksi feedback. Model ini dibagi menjadi dua bagian yaitu bagian fungsi yang menghitung masalah syarat batas dan keadaan awal sistem dengan parameter tetap, dan bagian model neural network dengan pengaturan parameter bobot yang diubah-ubah. Dalam penelitian ini, dilakukan penghitungan kasus persamaan Laplace dan persamaan difusi dengan syarat batas Dirichlet dan Neumann, dengan arsitektur single layer feedforward neural network yang berisi 10 neuron pada hidden layernya. Kemudian hasil pendekatan neural network dibandingkan dengan solusi eksak dan solusi numerik biasa. Dari hasil perbandingan tiap model, penyelesaian persamaan diferensial parsial memiliki akurasi yang mendekati solusi numerik, sehingga dapat dijadikan sebagai alternatif penyelesaian persamaan diferensial pada kasus syarat batas Dirichlet.
English Abstract
Solving Partial Differential Equation (PDE) is a mathematics modelling which is applied to solve physics cases based on natural principles phenomena and the laws regarding the property of matters. This study was conducted to get a model to determine the solution of partial differential equation with artificial neural network approximation. The notion of this study was based on the concept of feedforward neural network, mapping any function in forward without any feedback connection. The model was divided into two parts; the first part satisfies the boundary condition and the initial conditions of the problem, while the second part contains the neural network architecture with adjustable weight parameter which was constructed to approximate the solution of PDE. This study presented the method to solve Laplace equation and diffusion equation in the case of Dirichlet and Neumann boundary conditions. The solution was modeled by constructing a single layer feedforward neural network with 10 neurons in a hidden layer. Then, the result of PDE approximation was compared with exact solution and ordinary numerical method. From the comparison of each model, it was concluded that solving PDE with artificial neural network can be an alternative to determine the solution through its accuracy as it is comparable with exact solution in Dirichlet boundary problem.
Other obstract
-
Item Type: | Thesis (Sarjana) |
---|---|
Identification Number: | SKR/MIPA/2019/261/051911041 |
Uncontrolled Keywords: | Feedforward neural network, persamaan diferensial parsial, Syarat batas, Persamaan Laplace, Persamaan Difusi, Feedforward neural network, Different partial equation, Boundary conditions, Laplace equation, Diffusion Equation |
Subjects: | 500 Natural sciences and mathematics > 515 Analysis > 515.3 Differential calculus and equations > 515.35 Differential equations |
Divisions: | Fakultas Matematika dan Ilmu Pengetahuan Alam > Fisika |
Depositing User: | Budi Wahyono Wahyono |
Date Deposited: | 10 Aug 2020 08:03 |
Last Modified: | 24 May 2022 02:12 |
URI: | http://repository.ub.ac.id/id/eprint/179634 |
Text
Arin Siska Indarwatin.pdf Download (3MB) |
Actions (login required)
View Item |