Sistem Deteksi Anomali Air Pressure Dan Motor Current Pada Air Supply System Di Platform Offshore Menggunakan Metode BILSTM Autoencoder

Sabilulhaq, Dzikry Ihsanuddin (2026) Sistem Deteksi Anomali Air Pressure Dan Motor Current Pada Air Supply System Di Platform Offshore Menggunakan Metode BILSTM Autoencoder. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221128-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Air supply system merupakan salah satu sistem utilitas pada platform offshore yang menyediakan udara bertekanan untuk mendukung sistem pneumatik, instrumentasi, dan peralatan proses. Penurunan tekanan udara, perubahan karakteristik arus motor kompresor, maupun gangguan operasi lainnya dapat menyebabkan penurunan kinerja sistem dan meningkatkan risiko kerusakan peralatan. Pemantauan yang diterapkan pada umumnya masih menggunakan air pressure switch dan alarm berbasis batas atas bawah sehingga hanya mendeteksi kondisi ketika parameter telah melewati nilai ambang. Pendekatan tersebut belum mampu mengidentifikasi perubahan pola operasi yang masih berada dalam rentang normal namun mengindikasikan terjadinya degradasi sistem. Penelitian ini bertujuan mengembangkan sistem deteksi anomali berbasis Long Short-Term Memory (LSTM) Autoencoder untuk mengidentifikasi penyimpangan pola operasi pada data time series tekanan udara dan arus motor air compressor. Model dilatih menggunakan data operasi normal sehingga mampu mempelajari karakteristik perilaku sistem tanpa memerlukan data berlabel atau unsupervised learning. Proses deteksi dilakukan dengan menghitung reconstruction error antara data aktual dan hasil rekonstruksi model, kemudian membandingkannya dengan nilai threshold yang diperoleh menggunakan pendekatan mean dan standard deviation. Kinerja model dievaluasi menggunakan Receiver Operating Characteristic (ROC) dan Precision-Recall Area Under the Curve (PR-AUC) pada data uji yang telah diseimbangkan. Hasil evaluasi menunjukkan nilai ROC-AUC sebesar 0,9949 untuk model tekanan dan 0,9525 untuk model arus motor, sedangkan nilai PR-AUC masing-masing sebesar 0,9674 dan 0,8846. Hasil tersebut menunjukkan bahwa model mampu membedakan pola operasi normal dan anomali berdasarkan nilai reconstruction error, serta dapat digunakan sebagai dasar pemantauan kondisi air supply system pada platform offshore.
=====================================================================================================================================
Air supply system is one of the essential utility on offshore platforms, providing compressed air to support pneumatic systems, instrumentation, and process equipment. A reduction in air pressure, variations in compressor motor current characteristics, or other operational disturbances may degrade system performance and increase the risk of equipment failure. Conventional monitoring methods primarily rely on air pressure switches and threshold-based alarms, which are only capable of detecting abnormal conditions after monitored parameters exceed predefined upper or lower limits. Consequently, these approaches are unable to identify subtle changes in operational patterns that remain within the normal operating range but may indicate the onset of system degradation. This study aims to develop an anomaly detection system based on a Long Short-Term Memory (LSTM) Autoencoder to identify deviations in the operational patterns of air pressure and air compressor motor current data. Proposed model is trained on normal operating data, enabling it to learn the intrinsic behavior of the system without requiring labeled anomaly data through an unsupervised learning approach. Anomaly detection is performed by calculating the reconstruction error between the actual observations and the reconstructed outputs generated by the model, followed by comparison with a threshold determined using the mean and standard deviation of the reconstruction error. The model performance is evaluated using the Receiver Operating Characteristic Area Under the Curve (ROC-AUC) and Precision–Recall Area Under the Curve (PR-AUC) metrics on a balanced test dataset. The evaluation results demonstrate ROC-AUC values of 0.9949 for the air pressure model and 0.9525 for the compressor motor current model, while the corresponding PR-AUC values are 0.9674 and 0.8846, respectively. These findings indicate that the proposed model effectively distinguishes normal and anomalous operating patterns based on reconstruction error and can serve as a reliable foundation for condition monitoring of air supply systems on offshore platforms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	BiLSTM Autoencoder, Deteksi Anomali, Predictive Maintenance, Air Compressor, Pressure dan Current Anomali
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.62 Decision support systems T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TA Engineering (General). Civil engineering (General) > TA1573 Detectors. Sensors T Technology > TA Engineering (General). Civil engineering (General) > TA169.6 Fault location (Engineering)
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Dzikry Ihsanuddin Sabilulhaq
Date Deposited:	05 Aug 2026 04:27
Last Modified:	05 Aug 2026 04:27
URI:	http://repository.its.ac.id/id/eprint/143982

Actions (login required)

View Item