Karmina, Vasya Ayu (2025) Prediksi Fitur Health Index Bearing Boiler Feedwater Pump Turbine Pada Sistem Boiler PLTU Menggunakan Fitur Nonlinier dan Metode Temporal Convolutional Network. Other thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
5026211091-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy |
Abstract
Pemeliharaan prediktif penting untuk meningkatkan keandalan dan efisiensi peralatan industri, khususnya di pembangkit listrik tenaga uap (PLTU), karena kerusakan yang tidak terdeteksi dapat mengakibatkan downtime yang panjang dan biaya perbaikan yang tinggi. Tugas akhir ini berfokus pada prediksi health index bearing Boiler Feedwater Pump Turbine (BFPT), yang merupakan komponen dalam sistem boiler pada PLTU. Data getaran akan diekstraksi fitur nonliniernya menggunakan Detrended Fluctuation Analysis untuk mengetahui nilai eksponen skala per hari. Nilai eksponen skala yang terus naik seiring berjalannya waktu dapat menunjukkan tanda kegagalan pada mesin. Fitur tersebut akan digunakan sebagai health index untuk dijadikan input ke dalam model prediksi Temporal Convolutional Network (TCN), arsitektur deep learning yang mampu secara efektif menangkap pola lokal sekaligus memiliki bidang reseptif yang fleksibel. Performa model terbaik ada di komponen bearing 1 vibration X dengan nilai MAPE sebesar 6.6% dan MAE sebesar 0.1144. Selain itu, performa model terbaik cenderung ke komponen bearing 1 dengan nilai MAPE 6.66% untuk vibration X dan 6.71% untuk vibration Y. Performa model untuk bearing 2 mendapat nilai MAPE yang lebih tinggi, yaitu sebesar 15.92% untuk vibration X dan 9.34% untuk vibration Y. Konfigurasi parameter menunjukkan bahwa akurasi prediksi dapat ditingkatkan dengan menaikkan nilai num_blocks, filter, dan dropout sementara nilai num_blocks dipertahankan kecil. Setiap model juga memiliki kecepatan pelatihan kurang dari satu menit. Temuan ini diharapkan dapat memberikan kontribusi positif bagi pengembangan pemeliharaan prediktif pada mesin.
=======================================================================================================================================
Predictive maintenance is important to improve the reliability and efficiency of industrial equipment, especially in steam power plants, because undetected damage can result in long downtime and high repair costs. This final project focuses on predicting the health index of Boiler Feedwater Pump Turbine (BFPT) bearings, which are components in the boiler system in PLTU. The vibration data will be extracted for nonlinear features using Detrended Fluctuation Analysis (DFA) to determine the value of the scale exponent per day. The value of the scale exponent that continues to increase over time can indicate a sign of failure in the machine. This feature will be used as a health index to be used as input into the Temporal Convolutional Network (TCN) prediction model, a deep learning architecture that is able to effectively capture local patterns while having a flexible receptive field. The best model performance is in the bearing component 1 vibration X with a MAPE value of 6.6% and MAE of 0.1144. In addition, the best model performance tends to be for bearing component 1 with a MAPE value of 6.66% for vibration X and 6.71% for vibration Y. The model performance for bearing 2 gets a higher MAPE value, which is 15.92% for vibration X and 9.34% for vibration Y. The parameter configuration shows that accuracy prediction can be improved by increasing the num_blocks, filter, and dropout values while the num_blocks value is kept small. Each model also has a training speed of less than one minute. These findings are expected to provide a positive contribution to the development of predictive maintenance on machines.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | pembangkit listrik tenaga uap, pemeliharaan prediktif, health index, detrended fluctuation analysis, temporal convolutional network, steam power plants, predictive maintenance, health index, detrended fluctuation analysis, temporal convolutional network |
Subjects: | T Technology > T Technology (General) > T174 Technological forecasting |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis |
Depositing User: | Vasya Ayu Karmina |
Date Deposited: | 28 Jul 2025 08:06 |
Last Modified: | 28 Jul 2025 08:06 |
URI: | http://repository.its.ac.id/id/eprint/122617 |
Actions (login required)
![]() |
View Item |