Nurtaslimi, Ahmad Azkiya Nurtaslimi (2026) Sistem Penentuan Lifetime Motor Induksi Tiga Fasa Berdasarkan Data Getaran Bearing Menggunakan Long Short-Term Memory. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 2040211093-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
2040211093-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy |
Abstract
Kerusakan bearing pada motor induksi merupakan tantangan kritis di lingkungan industri karena sering memicu kegagalan sistemik akibat keterbatasan data run-to-failure dan tidak optimalnya pemantauan kondisi secara berkelanjutan. Penelitian ini mengusulkan sistem prognostik berbasis sinyal getaran menggunakan sensor MPU-6050 yang dikombinasikan dengan ekstraksi fitur Degradation Energy Indicator (DEI) pada frekuensi karakteristik bearing serta pembentukan Health Index (HI) berbasis Mahalanobis Distance (MD) untuk merepresentasikan degradasi secara kuantitatif pada kondisi data terbatas. Analisis run-to-failure pada bearing tipe 6201-2RS (dataset Imp1) dan NSK 6205 (dataset P5) menunjukkan peningkatan energi getaran yang progresif, dengan rasio MD faulty terhadap healthy mencapai 17,59 serta penurunan HI secara bertahap dari mendekati 1 hingga di bawah ambang degradasi 0,0876, yang menandai fase akhir umur pakai bearing. Model Long Short-Term Memory (LSTM) diterapkan untuk memprediksi tren degradasi HI melalui peramalan multi-step, dengan konfigurasi optimal menghasilkan performa prediksi yang stabil dan akurat, ditunjukkan oleh nilai MAE sebesar 0,0085–0,0087, RMSE sebesar 0,0108–0,0109, serta MAPE terendah sebesar 0,49% pada dataset Imp1_2. Integrasi model linear degradation memungkinkan estimasi lifetime atau Remaining Useful Life (RUL) secara real-time, di mana hasil estimasi pada bearing NSK 6205 menunjukkan sisa umur operasional sebesar 83,54 jam, yang lebih konservatif dan aman dibandingkan perhitungan teoritis L₁₀ sebesar 102,90 jam dalam memetakan batas akhir usia aktual 120 jam. Validasi empiris menunjukkan bahwa kegagalan bearing terdeteksi saat suhu operasi melampaui 85 °C dan intensitas getaran melebihi 9 m/s².
==================================================================================================================================
Bearing failures in induction motors remain a critical challenge in industrial environments, often leading to systemic breakdowns due to the scarcity of run-to-failure data and the lack of continuous condition monitoring. This study proposes a vibration-based prognostic system using an MPU-6050 sensor, integrating Degradation Energy Indicator (DEI) features extracted at bearing characteristic frequencies with a Health Index (HI) constructed using Mahalanobis Distance (MD) to quantitatively represent bearing degradation under limited data conditions. Run-to-failure analysis on 6201-2RS bearings (Imp1 dataset) and NSK 6205 bearings (P5 dataset) demonstrates a progressive increase in vibration energy, with the MD faulty-to-healthy ratio reaching 17.59 and a gradual decline in HI from values close to 1 to below the degradation threshold of 0.0876, indicating the terminal phase of bearing life. A Long Short-Term Memory (LSTM) model is employed to predict the temporal degradation trend of HI through multi-step forecasting, where the optimal configuration achieves stable and accurate performance with MAE values of 0.0085–0.0087, RMSE values of 0.0108–0.0109, and a minimum Mean Absolute Percentage Error (MAPE) of 0.49% on the Imp1_2 dataset. The integration of a linear degradation model enables real-time estimation of Remaining Useful Life (RUL), yielding an estimated RUL of 83.54 hours for the NSK 6205 bearing, which is more conservative and safer than the theoretical L₁₀ life of 102.90 hours when mapped against the empirical failure limit of 120 hours. Experimental validation confirms that bearing failure is detected when operating temperatures exceed 85 °C and vibration intensity surpasses 9 m/s².
Actions (login required)
 |
View Item |