Finazis, Riski Fitra (2019) Pemodelan Polishing Unit Berbasis Jaringan Syaraf Tiruan di PT Saka Indonesia Pangkah Limited. Other thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
02311540000085 - Undergraduate_Theses.pdf Download (2MB) | Preview |
Abstract
Pemodelan pada polishing unit yang merupakan bagian dari sistem produced water treatment PT Saka Indonesia Pangkah Limited (PT.SIPL) dilakukan menggunakan Jaringan Syaraf Tiruan (JST). JST dipergunakan untuk mempermudah pemodelan sistem pada polishing unit yaitu pada proses fisika serta biologis yang kompleks dan sulit dimodelakn secara matematis. JST yang digunakan berstruktur Multi-layer feedforward, dengan algoritma Levenberg-Marquardt. Model JST memiliki 4 input node pada input layer yaitu temperatur, pH, COD, TSS dan 1 output node pada output layer berupa COD effluent. Metode untuk menentukan hidden node terbaik belum ada sehingga dilakukan 20 variasi hidden node, untuk mencari nilai RMSE validasi yang terbaik. Hasil simulasi menunjukkan bahwa nilai RMSE terbaik berada pada JST dengan struktur 4-13-1 (Input layer-hidden layer- output layer). Nilai RMSE yang dihasilkan bernilai 14.7919, yang mana sudah lebih baik dari model JST referensi yang bernilai 59.48. Model JST yang dipergunakan mampu menghasilkan nilai COD dibawah kadar maksimum baku mutu air limbah kegiatan produksi MIGAS fasilitas on-shore dengan rentang nilai 32.4555 - 97.84738 (mg/l). Sehingga dinamika proses pada polishing unit dapat ditunjukkan model JST 4-13-1 dengan parameter yang mempengaruhi yaitu temperatur, pH, COD, dan TSS. ================================================================================================================================
Modeling of polishing units which are part of produced water treatment system by PT Saka Indonesia Pangkah Limited (PT.SIPL) is carried out using Artificial Neural Networks (ANN). ANN is used to facilitate system modeling in polishing units, namely physical and biological processes that are complex and difficult to model mathematically. ANN is used with a multi-layer feedforward structure, with the Levenberg-Marquardt algorithm. ANN model has 4 input nodes at the input layer, namely temperature, pH, COD, TSS and 1 output node in the output layer consisting of COD waste. The method for determining the best hidden node must be done 20 variations of hidden nodes, to find the best RMSE validation value. The simulation results show the best RMSE value on ANN with structure 4-13-1 (input layer-hidden layer-output layer). The resulting RMSE value is 14.7919, which is better than the reference ANN model which is worth 59.48. ANN model that is used is able to produce COD values below the maximum level of waste water quality standards oil & gas production on-shore facilities with a range values 32.4555 - 97.84738 (mg/l). So that the process dynamic of the polishing unit can be shown in the ANN 4-13-1 model with parameters that affect the temperature, pH, COD, and TSS.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Produced water; Polishing Unit; Jaringan Syaraf Tiruan |
| Subjects: | T Technology > TD Environmental technology. Sanitary engineering > TD433 Water treatment plants |
| Divisions: | Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis |
| Depositing User: | Riski Fitra Finazis |
| Date Deposited: | 24 Sep 2024 05:42 |
| Last Modified: | 24 Sep 2024 05:42 |
| URI: | http://repository.its.ac.id/id/eprint/67730 |
Actions (login required)
 |
View Item |