Pemantauan Kondisi Berbasis Data Untuk Gas Admission Valve Pada Mesin Dual-Fuel Empat-Tak Dengan Machine Learning

Naufan, Muhammad Rashif (2026) Pemantauan Kondisi Berbasis Data Untuk Gas Admission Valve Pada Mesin Dual-Fuel Empat-Tak Dengan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6010241036-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Penelitian ini bertujuan untuk membangun model prediksi kegagalan berbasis machine learning. Studi dilakukan pada PLN Indonesia Power Pusat Listrik Tenaga Diesel Gas 200 MW Pesanggaran Bali. Permasalahan yang dihadapi adalah tingginya gangguan pada Solenoid Operated Gas Admission Valve, yang mengakibatkan loss production sebesar 1.058,68 MWh pada periode 2024–2025. Strategi yang diterapkan masih bersifat corrective dan preventive, sehingga belum mampu mendeteksi kerusakan secara dini. Data penelitian berasal dari Supervisory Control and Data Acquisition-Human Machine Interface mesin. Data mentah dilakukan pemrosesan awal agar menghasilkan data yang layak dan dimodelkan ke dalam algoritma machine learning yaitu Random Forest, Gradient Boosting dan XGBoost. Selanjutnya model dievaluasi untuk diperoleh algoritma yang terbaik. Dua pendekatan yang dilakukan adalah klasifikasi dan residual analisis pada model yang berbeda, pada skenario yang berbeda, resolusi data yang berbeda dan teknik labelling yang berbeda. Hasil pengujian menunjukkan skenario terbaik diperoleh dari pendekatan model klasifikasi dengan resolusi data 10 menit, labelling event trip ±2 jam. Hasil pengujian deteksi dini menunjukkan model klasifikasi dengan algoritma XGBoost yang terbaik. Evaluasi kemampuan deteksi dini menunjukkan bahwa model berhasil mendeteksi seluruh kejadian gangguan pada data training dan testing. Pada data testing, XGBoost memberikan lead time peringatan dini masing-masing 64 menit dan 37 menit sebelum terjadinya gangguan.
=======================================================================================================================================
This study aims to develop a machine learning-based failure prediction model. The research was conducted at PLN Indonesia Power, Pesanggaran 200 MW Diesel Gas Power Plant in Bali. The primary issue addressed is the high frequency of failures in the Solenoid Operated Gas Admission Valve, which resulted in a production loss of 1,058.68 MWh during the 2024–2025 period. The existing maintenance strategy is still mainly corrective and preventive and therefore has not been able to detect potential failures at an early stage. The research utilized data retrieved from the engine's Supervisory Control and Data Acquisition-Human Machine Interface (SCADA-HMI). The raw data underwent preprocessing to ensure data quality before being modelled using machine learning algorithms, specifically Random Forest, Gradient Boosting, and XGBoost. The models were subsequently evaluated to determine the optimal algorithm. Two distinct approaches were implemented: classification and residual analysis, tested across different models, scenarios, data resolutions, and labelling techniques. The test results show that the best scenario is obtained from the classification model approach with a 10-minute data resolution and labelling event trips ±2 hours. The early detection evaluation results show that the best classification model is XGBoost. The model successfully detected all fault events in both training and testing data. In the testing dataset, XGBoost provided early warning lead times of 64 minutes and 37 minutes before the fault events occurred.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Deteksi anomali, Pembangkit listrik tenaga gas, Supervised learning, Aplikasi machine learning, Klasifikasi, Regresi, Resisual, Moving average, anomaly detection, gas-fired power plant, machine learning application, classification, regression, residual
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26101-(S2) Master Thesis
Depositing User:	Muhammad Rashif Naufan
Date Deposited:	29 Jul 2026 06:50
Last Modified:	29 Jul 2026 06:50
URI:	http://repository.its.ac.id/id/eprint/139544

Actions (login required)

View Item