Azhari, Inayatullah Muhammad (2025) Perancancangan Dan Implementasi Sistem Diagnosis Kerusakan Bearing Pada Conveyor Idler Menggunakan Convolutional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 2040211053-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
2040211053-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Sistem konveyor pada industri pertambangan digunakan untuk memindahkan batubara dari lokasi penambangan hingga proses pemuatan ke kapal tongkang. Keandalan sistem konveyor menjadi prioritas utama untuk menjaga kelancaran proses pertambangan secara kontinu dan mencegah terjadinya shutdown operasional. Salah satu penyebab utama shutdown adalah kerusakan pada bearing idler, yang dapat menimbulkan kerugian finansial. Proyek akhir ini bertujuan untuk mengembangkan sistem diagnosis kerusakan bearing pada conveyor idler secara real-time menggunakan Continuous Wavelet Transform (CWT) dan Convolutional Neural Network (CNN). CWT digunakan untuk mentransformasi sinyal getaran ke dalam domain waktu-frekuensi, sehingga memungkinkan analisis terhadap sinyal non-stasioner. Sementara itu, CNN digunakan untuk mengklasifikasikan kondisi bearing berdasarkan pola dari hasil transformasi. Data getaran diperoleh dari sensor akselerometer tiga sumbu (X, Y, dan Z) pada kecepatan putar shaft idler sebesar 120 RPM. Sistem ini diimplementasikan untuk diagnosis tiga kondisi bearing, yaitu corrosion, normal, dan plastic deformation. Seluruh proses diagnosis divisualisasikan melalui dashboard untuk mendukung proses monitoring. Hasil implementasi menunjukkan bahwa model CNN mampu mengenali kondisi normal dan plastic deformation dengan confidence sebesar 1.00 pada ketiga sumbu, dan menghasilkan total 32 prediksi. Model yang digunakan berasal dari percobaan ketiga, dengan akurasi sebesar 91,67% pada data validasi maupun data pengujian. Namun, sistem belum mampu mengenali kondisi bearing corrosion, karena seluruh data corrosion diklasifikasikan sebagai kondisi normal dengan confidence sebesar 1.00. Hal ini menunjukkan bahwa pola fitur pada data corrosion memiliki kemiripan dengan kondisi normal, sehingga model tidak dapat membedakan fitur sinyal dari kedua kondisi bearing tersebut. Fitur alarm hanya aktif ketika sistem mendeteksi kondisi plastic deformation. Proses diagnosis mengalami delay selama 6 detik, yang disebabkan oleh segmentasi sebanyak 120 data dan mekanisme voting pada setiap sumbu. Oleh karena itu, pengembangan lebih lanjut diperlukan untuk meningkatkan kemampuan model CNN dalam mengenali kondisi bearing corrosion dan mengurangi delay salama proses diagnosis.
========================================================================================================================================
Conveyor systems in the mining industry are used to transport coal from the mining site to the loading process onto barges. The reliability of conveyor systems is a top priority to maintain the continuity of mining operations and prevent operational shutdowns. One of the main causes of shutdowns is damage to idler bearings, which can result in financial losses. This final project aims to develop a real-time bearing damage diagnosis system for conveyor idlers using Continuous Wavelet Transform (CWT) and Convolutional Neural Network (CNN). CWT is used to transform vibration signals into the time-frequency domain, enabling analysis of non-stationary signals. Meanwhile, CNN is used to classify bearing conditions based on patterns from the transformation results. Vibration data is obtained from a three-axis accelerometer sensor (X, Y, and Z) at an idler shaft rotation speed of 120 RPM. This system is implemented for the diagnosis of three bearing conditions: corrosion, normal, and plastic deformation. The entire diagnosis process is visualized through a dashboard to support the monitoring process. The implementation results show that the CNN model is capable of recognizing normal conditions and plastic deformation with a confidence level of 1.00 on all three axes, producing a total of 32 predictions. The model used was from the third experiment, with an accuracy of 91.67% on both the validation and testing data. However, the system was unable to recognize bearing corrosion conditions, as all corrosion data was classified as normal conditions with a confidence level of 1.00. This indicates that the feature patterns in the corrosion data are similar to those in normal conditions, so the model cannot distinguish the signal features between the two bearing conditions. The alarm feature is only active when the system detects plastic deformation conditions. The diagnosis process experiences a delay of 6 seconds, caused by the segmentation of 120 data points and the voting mechanism on each axis. Therefore, further development is needed to enhance the CNN model's ability to recognize bearing corrosion conditions and reduce the delay during the diagnosis process.
Actions (login required)
 |
View Item |