Perancangan Sistem Pengendali Komposisi Pada Kolom Demethanizer Dan Deethanizer Berbasis Artificial Neural Network (ANN)

Putra, I Putu Reza Rizkana (2021) Perancangan Sistem Pengendali Komposisi Pada Kolom Demethanizer Dan Deethanizer Berbasis Artificial Neural Network (ANN). Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

Text 02311740000030-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2023. Download (2MB) | Request a copy

Abstract

Kolom distilasi seperti demethanizer dan deethanizer merupakan sistem multi-input multi-output (MIMO) yang memiliki karakteristik nonlinear. Dalam tugas akhir ini dirancang sistem Model Predictive Control (MPC) berbasis neural network. Input kontrol dibangkitkan berdasarkan output yang telah diprediksi dengan menggunakan model neural network. Model neural network dibuat berdasarkan hasil pengambilan data open loop dan dibagi menjadi dua model NN MISO. Model NN MISO 1 menunjukkan hasil MSE pelatihan sebesar 0,00008792 dengan hasil pengujian regresi (R) sebesar 0,9957. Model NN MISO 2 menunjukkan hasil MSE pelatihan sebesar 0,00001868 dengan hasil pengujian regresi (R) sebesar 0,9785. Perancangan pengendali NNMPC 1 untuk mengendalikan komposisi methane dihasilkan performa dengan settling time sebesar 36 detik, error steady state sebesar 0,001053%, dan maximum overshoot sebesar 3,921%. Pengendali NNMPC 2 untuk mengendalikan komposisi ethane dihasilkan performa dengan settling time sebesar 45 detik, error steady state sebesar 0,0007%, dan maximum overshoot sebesar 1,8911%. Pengujian disturbance juga dilakukan dengan hasil settling time sebesar 233 detik, error steady state sebesar 0,00021%, dan maximum overshoot sebesar 0,1381% untuk pengendali NNMPC 1. Pengendali NNMPC 2 menghasilkan performa dengan hasil settling time sebesar 220 detik, error steady state sebesar 0,00032%, dan maximum overshoot sebesar 0,0793%.
======================================================================================================
The distillation columns such as demethanizer dan deethanizer are multi-input multi-output (MIMO) systems which has nonlinear characteristics. In this final project, a neural network-based Model Predictive Control (MPC) system is designed. The control input is generated based on the predicted output using a neural network model. Neural network model is based on results of open loop data. The NN MISO 1 model shows MSE results of 0.00008792 with R value of 0.9957. The NN MISO 2 model shows MSE results of 0.00001868 with R value of 0.9785. NNMPC 1 is used to control the methane composition resulted in performance with 36 seconds settling time, 0.001053% ess, and 3.921% maximum overshoot. NNMPC 2 controller to control the ethane composition resulted in a performance with 45 seconds settling time, 0.0007% ess, and 1.8911% maximum overshoot. The disturbance test was also carried out with the results of 233 seconds settling time, 0.00021% ess, and 0.1381% maximum overshoot for the NNMPC 1 controller. NNMPC 2 controller produced a performance of 220 seconds settling time, 0.00032% ess, and 0.0793% maximum overshoot.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Demethanizer column, Deethanizer column, neural network, Neural Network Model Predictive Control, Kolom demethanizer, Kolom deethanizer, Neural network, Neural Network Model Predictive Control.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TJ Mechanical engineering and machinery > TJ217.6 Predictive Control
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	I Putu Reza Rizkana Putra
Date Deposited:	01 Sep 2021 05:18
Last Modified:	01 Sep 2021 05:18
URI:	http://repository.its.ac.id/id/eprint/90809

Actions (login required)

View Item