Pembuatan Indeks Kesehatan (Health Index) Degradasi Bearing Untuk Memperkirakan Sisa Usia Pakai Berdasarkan Pemrosesan Fitur

Ginting, Epindonta (2024) Pembuatan Indeks Kesehatan (Health Index) Degradasi Bearing Untuk Memperkirakan Sisa Usia Pakai Berdasarkan Pemrosesan Fitur. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5007201140-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (7MB) | Request a copy

Abstract

Peralatan rotary digunakan secara luas di berbagai industri, termasuk industri otomotif, dirgantara, petrokimia, dan pembangkit listrik. Kegagalan bearing menimbulkan risiko signifikan terhadap mesin-mesin industri di berbagai sektor. Oleh karena itu, monitoring kondisi bearing dan prediksi umur bearing sangat penting dilakukan. Penelitian ini mengusulkan pendekatan baru untuk memprediksi sisa umur pakai (Remaining Useful Life/RUL) bearing dengan menggabungkan jaringan Long Short-Term Memory (LSTM) dengan health index. Data getaran dikumpulkan dari taper roller bearing yang secara sengaja dirusak dengan mesin Electrical Discharge Machining (EDM). Data vibrasi dikumpulkan menggunakan Wilcoxon Mac800 pada frekuensi 25,6 kHz. Health index diperoleh melalui proses yang melibatkan ekstraksi fitur statistik, pemilihan 5 fitur dengan korelasi tertinggi, Principal Component Analysis (PCA) untuk reduksi dimensi, dan smoothing. Pendekatan ini divalidasi pada dataset kerusakan buatan dan dataset bantalan NASA yang menunjukkan mode kegagalan serupa. Hasil penelitian menunjukkan peningkatan substansial dalam akurasi prediksi RUL ketika menggunakan health index yang dirancang dibandingkan dengan menggunakan data getaran secara langsung. Model prediksi yang dibangun memiliki akurasi prediksi Mean Absolute Error (MAE) sebesar 0,003592 dan Root Mean Squared Error (RMSE) sebesar 0,0051547. Selain itu, penelitian ini juga menunjukkan pentingnya data run-to-failure untuk menghasilkan prediksi RUL yang lebih akurat dikarenakan model sudah dilatih dengan pola degradasi secara menyeluruh.
==================================================================================================================================
Rotary equipment is widely used in various industries, including automotive, aerospace, petrochemical, and power generation. Bearing failures pose significant risks to industrial machinery across these sectors. Therefore, bearing condition monitoring and RUL prediction are crucial tasks. This research proposes a novel approach for RUL prediction of bearings by combining Long Short-Term Memory (LSTM) networks with health indices. Vibration data was collected from taper roller bearings intentionally damaged using Electrical Discharge Machining (EDM). Vibration data was acquired using Wilcoxon Mac800 at a frequency of 25.6 kHz. Health indices were obtained through a process involving feature extraction, selection of the five most correlated features, Principal Component Analysis (PCA) for dimension reduction, and smoothing. This approach was validated on a synthetic fault dataset and a NASA bearing dataset exhibiting similar failure modes. Results demonstrate a substantial improvement in RUL prediction accuracy when using the designed health indices compared to using raw vibration data directly. The developed prediction model achieved a Mean Absolute Error (MAE) of 0.003592 and a Root Mean Squared Error (RMSE) of 0.0051547. Additionally, the study highlights the importance of run-to-failure data in generating more accurate RUL predictions as the model is trained on the complete degradation pattern.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Prognosis kerusakan, jaringan memori jangka panjang dan pendek (LSTM), bearing gelinding (rolling bearing), mesin berputar, getaran; Damage prognosis, long and short term memory (LSTM) networks, rolling bearings, rotating machines, vibration Fault prognosis, long and short-term memory network (LSTM), rolling bearing, rotating machinery, vibration.
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:	Epindonta Ginting
Date Deposited:	29 Jul 2024 03:19
Last Modified:	29 Jul 2024 03:19
URI:	http://repository.its.ac.id/id/eprint/109263

Actions (login required)

View Item