Integrasi Algoritma Pruned Exact Linear Time (PELT) dan Bidirectional Gated Recurrent Unit (Bi-GRU) pada Sistem Prediksi Remaining Useful Life Bearing Berbasis Condition-Aware

Musyaffa, Naufal Fathan Abiyyu (2026) Integrasi Algoritma Pruned Exact Linear Time (PELT) dan Bidirectional Gated Recurrent Unit (Bi-GRU) pada Sistem Prediksi Remaining Useful Life Bearing Berbasis Condition-Aware. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5007221126-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (11MB) | Request a copy

Abstract

Bearing merupakan komponen kritis pada mesin berputar yang kegagalannya dapat menyebabkan downtime tidak terencana dan kerugian operasional. Pendekatan prediksi Remaining Useful Life (RUL) konvensional umumnya menggunakan seluruh umur bearing sebagai data masukan tanpa membedakan fase sehat dan fase degradasi, sehingga informasi yang digunakan belum sepenuhnya merepresentasikan perkembangan kerusakan. Penelitian ini mengembangkan sistem prediksi RUL berbasis condition-aware melalui segmentasi fase degradasi menggunakan algoritma Pruned Exact Linear Time (PELT) pada Health Indicator (HI) berbasis Autoencoder, kemudian memprediksi RUL menggunakan model Bidirectional Gated Recurrent Unit (Bi-GRU). Penelitian menggunakan sinyal getaran horizontal dari dataset XJTU-SY, sedangkan performa model dievaluasi menggunakan Mean Absolute Error (MAE), Root Mean Square Error (RMSE), dan bias. Hasil penelitian menunjukkan bahwa pendekatan yang diusulkan menghasilkan rata-rata MAE sebesar 9.2 menit, lebih rendah dibandingkan pendekatan tanpa segmentasi fase degradasi yang menghasilkan rata-rata MAE sebesar 154.1 menit. Validasi tambahan pada dataset PRONOSTIA juga menghasilkan MAE sebesar 11.83 menit, menunjukkan bahwa pipeline yang diusulkan tetap dapat diterapkan pada dataset dengan karakteristik data yang berbeda. Hasil tersebut menunjukkan bahwa segmentasi fase degradasi mampu meningkatkan akurasi prediksi RUL dengan memfokuskan proses pembelajaran pada informasi degradasi yang relevan.
========================================================================================================================================
Bearings are critical components in rotating machinery whose failures can lead to unplanned downtime and significant operational losses. Conventional Remaining Useful Life (RUL) prediction approaches generally utilize the entire bearing lifetime without distinguishing between healthy and degradation phases, causing the input data to contain information that is not fully representative of the degradation process. This study develops a condition-aware RUL prediction system by integrating the Pruned Exact Linear Time (PELT) algorithm for degradation phase segmentation on an Autoencoder-based Health Indicator (HI) and the Bidirectional Gated Recurrent Unit (Bi-GRU) model for RUL prediction. The proposed method was evaluated using horizontal vibration signals from the XJTU-SY bearing dataset, with prediction performance assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and bias. The proposed approach achieved an average MAE of 9.2 minutes, outperforming the approach without degradation phase segmentation, which yielded an average MAE of 154.1 minutes. Additional validation on the PRONOSTIA dataset resulted in an MAE of 11.83 minutes, demonstrating that the proposed pipeline can be applied to datasets with different acquisition characteristics. These results indicate that degradation phase segmentation improves RUL prediction accuracy by enabling the model to focus on degradation-related information.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	RUL, Condition-Aware, Health Indicator, PELT, Bi-GRU.
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ174 Maintenance and repair of machinery
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Mechanical Engineering > 21201-(S1) Undergraduate Thesis
Depositing User:	Naufal Fathan Abiyyu Musyaffa
Date Deposited:	03 Aug 2026 10:22
Last Modified:	03 Aug 2026 10:22
URI:	http://repository.its.ac.id/id/eprint/142380

Actions (login required)

View Item