Perancangan Model Predictive Control Pada Kolom Distilasi Biner Menggunakan Pemodelan Neural Network

Putra, DIo Alif (2021) Perancangan Model Predictive Control Pada Kolom Distilasi Biner Menggunakan Pemodelan Neural Network. Undergraduate thesis, INSTITUT TEKNOLOGI SEPULUH NOPEMBER.

[img] Text
02311740000080-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (1MB) | Request a copy

Abstract

Proses pemisahan menggunakan kolom distilasi pada industri perminyakan dan kimia di seluruh dunia diperkirakan mencapai 95%. Untuk memperoleh produk distilasi dengan kemurnian tinggi sangat sulit untuk dikendalikan jika menggunakan pengontrol konvensional karena sifat kolom yang non-linear. Oleh karena itu, dilakukan perancangan model neural network untuk menggambarkan proses pada kolom distilasi biner yang sifatnya non-linear, dan dilakukan perancangan Model Predictive Control (MPC) untuk mengendalikan produk distilat/top product dan produk bawah/bottom product dari kolom distilasi biner. Pemodelan neural network MLP (Multi Layer Perceptron) dirancang dengan struktur 3-layer feed-forward neural network. Model neural network yang dirancang dapat memodelkan proses kolom distilasi biner dengan baik, dimana didapatkan nilai MSE pelatihan sebesar 0,000164715 untuk fraksi mol produk distilat (XD), dan nilai MSE pelatihan sebesar 0,00019959 untuk fraksi mol produk bawah (XB), serta hasil akurasi maksimal pengujian prediksi pemodelan neural network diperoleh sebesar 1,0. Sistem kontrol MPC memiliki performansi yang baik dalam hal mengendalikan kemurnian produk distilat (XD) dan kemurnian produk bawah (XB). Pengendali MPC untuk mengendalikan fraksi mol produk distilat mendapatkan nilai settling time sebesar 25 detik, error steady state sebesar 0,010147% dan maximum overshoot sebesar 0,8091%. Pengendali MPC untuk mengendalikan fraksi mol produk bawah mendapatkan nilai settling time sebesar 15 detik, error steady state sebesar 0,05263% dan maximum overshoot sebesar 0,2439% ======================================================================================================= The separation process using distillation columns in the petroleum and chemical industries worldwide is estimated at 95%. To obtain high-purity distillation products, it is very difficult to control using conventional controllers because of the non-linear nature of the column. Therefore, a neural network model is designed to describe the non-linear process in a binary distillation column, and a Model Predictive Control (MPC) is designed to control the distillate product/top product and bottom product from the binary distillation column. The MLP (Multi Layer Perceptron) neural network model is designed with a 3-layer feed-forward neural network structure. The designed neural network modelling can model the binary distillation column process well enough, where the training MSE value is 0.000164715 for the mole fraction of the distillate product (XD), and training MSE value is 0.00019959 for the mole fraction of bottom product (XB) and the maximum accuracy of the neural network modelling prediction test is 1.0. MPC has good performance in terms of controlling the purity of the distillate product (XD) and the purity of the bottom product (XB). MPC controller for controlling the mole fraction of the distillate product obtained a settling time value of 25 seconds, error steady state of 0,010147% and maximum overshoot of 0,8091%. MPC controller for controlling the mole fraction of the bottom product obtained a settling time value of 15 seconds, error steady state of 0.05263% and a maximum overshoot of 0.2439%.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Distillation column, top product, bottom product, neural network model, Model Predictive Control, Kolom Distilasi, Produk Distilat, Produk Bawah
Subjects: Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
T Technology > TA Engineering (General). Civil engineering (General) > TA660.C6 Columns
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: Dio Alif Putra
Date Deposited: 24 Aug 2021 03:15
Last Modified: 24 Aug 2021 05:03
URI: https://repository.its.ac.id/id/eprint/89103

Actions (login required)

View Item View Item