Pemodelan Keandalan Hot Well Pump Berbasis Data Proses Dengan Metode Principal Component Analysis Pada Pembangkit Listrik Tenaga Panas Bumi

Setianingtias, Regina Putri (2026) Pemodelan Keandalan Hot Well Pump Berbasis Data Proses Dengan Metode Principal Component Analysis Pada Pembangkit Listrik Tenaga Panas Bumi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221052-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Hot well pump merupakan komponen vital dalam siklus uap kondensat Pembangkit Listrik Tenaga Panas Bumi (PLTP) yang berfungsi memompa kondensat dari kondenser menuju cooling tower. Gangguan pada hot well pump dapat menurunkan output daya pembangkit hingga 50%, sehingga pemantauan kondisi peralatan secara berkelanjutan menjadi kebutuhan kritis. Pendekatan pemodelan keandalan konvensional yang mengandalkan data historis kegagalan memiliki keterbatasan karena kejadian trip pada hot well pump relatif jarang terjadi. Penelitian ini mengusulkan metode pemodelan keandalan berbasis data proses yang tersedia secara kontinu melalui sistem Distributed Control System (DCS). Metode Principal Component Analysis (PCA) diterapkan untuk mereduksi dimensi lima variabel proses yaitu discharge pressure, bearing temperature, pump vibration, motor current, dan motor voltage, menjadi komponen utama yang merepresentasikan pola operasi normal hot well pump. Skor Principal Component (PC) terpilih kemudian digabungkan menjadi satu Condition Indicator (CI) yang merepresentasikan besarnya penyimpangan kondisi operasi terhadap pola normal. Distribusi lognormal dipilih sebagai model probabilistik terbaik secara relatif berdasarkan uji goodness-of-fit. Model keandalan statis dihitung berdasarkan konsep stress-strength dengan CI sebagai stress dan persentil ke-99 data normal sebagai constant strength. Keandalan dinamis dihitung secara kumulatif dari akumulasi nilai keandalan statis harian untuk menggambarkan tren degradasi terhadap waktu. Hasil penelitian ini menunjukkan bahwa model 1 dengan lima variabel proses menghasilkan tiga PC utama dengan variansi kumulatif 75.17% dan mampu mendeteksi penurunan keandalan yang konsisten menjelang kejadian trip akibat vibrasi tinggi, dengan rata – rata CI pada data normal sebesar 1.567 dibandingkan 550.841 pada data trip. Model 2 yang disederhanakan menjadi tiga variabel dominan, menghasilkan tiga PC dengan variansi kumulatif 100% serta rata – rata CI sebesar 1.576 pada data normal dan 147.364 pada data trip, menunjukkan bahwa penyederhanaan jumlah variabel tetap mempertahankan sensitivitas indikator. Kedua model menghasilkan model keandalan statis dengan koefisien determinasi (R2) berturut – turut sebesar 0.9793 dan 0.8873. Seluruh hasil pemodelan diintegrasikan dalam antarmuka interaktif berbasis MATLAB App Designer untuk mendukung pemantauan kondisi hot well pump secara visual dan berkelanjutan. Penelitian ini berkontribusi pada SDGs ke-9, yaitu Industry, Innovation, and Infrastructure, melalui pengembangan sistem pemantauan berbasis data inovatif.
===================================================================================================================================
The hot well pump is a vital component in the steam–condensate cycle of a Geothermal Power Plant (GPP), responsible for pumping condensate from the condenser to the cooling tower. Disturbances in the hot well pump can reduce the power plant’s output by up to 50%, making continuous condition monitoring a critical requirement. Conventional reliability modeling approaches that rely on historical failure data have limitations because trip events in hot well pumps occur relatively infrequently. Therefore, this study proposes a process data-based reliability modeling method using data continuously available through the Distributed Control System (DCS). Principal Component Analysis (PCA) was applied to reduce the dimensionality of five process variables—discharge pressure, bearing temperature, pump vibration, motor current, and motor voltage—into principal components representing the normal operating pattern of the hot well pump. The selected principal component scores were then combined into a single Condition Indicator (CI) representing the magnitude of deviation from the normal operating pattern. The lognormal distribution was selected as the relatively best probabilistic model based on the Anderson-Darling goodness-of-fit test. Static reliability was calculated based on the stress-strength concept, treating CI as a random stress variable and the 99th percentile of normal data as the constant strength. Dynamic reliability was computed cumulatively from daily static reliability values to illustrate degradation trends over time. Results show that model 1, using five process variables, yielded three principal components with a cumulative variance of 75.17% and was capable of detecting a consistent reliability decline prior to the trip event on October 10, 2023, caused by high vibration, with a mean CI of 1.567 for normal data compared to 550.841 for trip data. Model 2 simplified to the three most dominant variables, yielded three principal components with a cumulative variance of 100% and mean CI values of 1.576 for normal data and 147.364 for trip data, showing that reducing the number of variables still preserved the indicator's sensitivity in distinguishing normal from trip conditions. Both models produced static-reliability regression fits with R² values of 0.9793 and 0.8873, respectively. All modeling results were integrated into an interactive MATLAB App Designer interface to support continuous and visual condition monitoring of the hot well pump. This research contributes to SDG 9, that is Industry, Innovation, and Infrastructure, through the development of an innovative data-driven monitoring system.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Data Proses, Hot Well Pump, Keandalan, Pembangkit Listrik Tenaga Panas Bumi, Principal Component Analysis ============================================================ Geothermal Power Plant, Hot Well Pump, Principal Component Analysis, Process Data, Reliability
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA169 Reliability (Engineering)
Divisions:	Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Regina Putri Setianingtias
Date Deposited:	30 Jul 2026 07:42
Last Modified:	30 Jul 2026 07:46
URI:	http://repository.its.ac.id/id/eprint/139976

Actions (login required)

View Item