Pengembangan Model Anomaly Detection untuk Monitoring Amine Pump Pada Proyek LEADS di PT Pertamina EP Cepu

Sarwono, Wirandito (2025) Pengembangan Model Anomaly Detection untuk Monitoring Amine Pump Pada Proyek LEADS di PT Pertamina EP Cepu. Project Report. [s.n.]. (Unpublished)

Text 5025221088-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Proyek ini berfokus pada implementasi teknologi anomaly detection untuk memantau kinerja amine pump di fasilitas Matindok guna meningkatkan efisiensi operasional dan mencegah kerusakan peralatan, menggunakan data deret waktu dari sistem LEADS milik Pertamina EP Cepu – Matindok Field yang memantau aktivitas dua unit pompa amina, 340P1001 dan 340P1002, mencakup parameter operasional seperti tekanan, suhu, laju aliran, serta posisi katup. Metodologi CRISP-DM diterapkan, dengan pemilihan algoritma Long Short-Term Memory (LSTM) Autoencoder dalam tahap pemodelan karena efektivitasnya dalam menangani data deret waktu dan mendeteksi anomali, di mana model dilatih menggunakan data operasi normal dan anomali terdeteksi ketika nilai reconstruction error meningkat secara signifikan. Hasil deteksi anomali ini kemudian divisualisasikan menggunakan platform Business Intelligence Power BI untuk menampilkan laporan dan dasbor interaktif, bertujuan memberikan peringatan dini mengenai potensi kegagalan pompa, meningkatkan keandalan peralatan, memperpanjang umur aset, meningkatkan keselamatan, dan meminimalkan kerugian akibat downtime yang tidak terduga.
=====================================================================================================================================
This project focuses on the implementation of anomaly detection technology to monitor the performance of amine pumps at the Matindok facility to improve operational efficiency and prevent equipment damage, using time-series data from the LEADS system owned by Pertamina EP Cepu – Matindok Field which monitors the activity of two amine pump units, 340P1001 and 340P1002, covering operational parameters such as pressure, temperature, flow rate, and valve position. The CRISP-DM methodology is applied, with the selection of the Long Short-Term Memory (LSTM) Autoencoder algorithm in the modeling stage due to its effectiveness in handling time-series data and detecting anomalies, where the model is trained using normal operation data and anomalies are detected when the reconstruction error value increases significantly. These anomaly detection results are then visualized using the Power BI Business Intelligence platform to display reports and interactive dashboards, aiming to provide early warnings regarding potential pump failures, improve equipment reliability, extend asset life, enhance safety, and minimize losses due to unexpected downtime.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	deteksi anomali, amine pump, LSTM Autoencoder, predictive maintenance, Pertamina EP Cepu.
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Wirandito Sarwono
Date Deposited:	06 Dec 2025 03:57
Last Modified:	06 Dec 2025 03:57
URI:	http://repository.its.ac.id/id/eprint/128869

Actions (login required)

View Item