Prediksi Remaining Useful Lifetime pada Mesin Raw Mill menggunakan Transformer-Gated Convolutional Unit

Muliatama, Berliana Putri (2023) Prediksi Remaining Useful Lifetime pada Mesin Raw Mill menggunakan Transformer-Gated Convolutional Unit. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06111940000013-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (6MB) | Request a copy

Abstract

Di era Industri 4.0, Internet of Thing (IoT) dapat dimanfaatkan untuk pengambilan keputusan di bidang pemeliharaan (maintenance), yakni dengan memanfaatkan Predictive Maintenance (PdM). Salah satu task dari Predictive Maintenance (PdM) adalah pengestimasian Remaining Useful Lifetime (RUL) dari mesin. Salah satu mesin yang memiliki peran penting adalah mesin raw mill pada industri semen, sehingga estimasi RUL untuk mesin ini dinilai perlu untuk dilakukan. Prediksi RUL dalam penelitian ini menggunakan algoritma deep learning dengan model berbasis Transformer yang memanfaatkan Gated Convolutional Unit (GCU) sebagai lapisan ekstraksi fitur lokalnya. Penggunaan Transformer dapat memanfaatkan mekanisme parallel computation sehingga mengurangi waktu komputasi tetapi mampu menangkap ketergantungan jangka panjang. Kemudian, GCU digunakan untuk menekankan kontribusi konteks lokal dari data yang digunakan pada ekstraksi fitur. Dalam penelitian ini, digunakan data multisensor dari mesin raw mill salah satu pabrik semen di Indonesia pada 2015. Hasil simulasi menunjukkan bahwa model T-GCU dapat memprediksi RUL mesin dengan baik berdasarkan observasi nilai RMSE-nya terhadap ground truth RUL mesin raw mill. Selanjutnya, hasil penelitian ini dapat dijadikan sebagai dasar untuk menjadwalkan maintenance sebelum terjadi kegagalan.
===============================================================================================================================
In the industrial era of 4.0, Internet Of Things (IoT) can be used to make data driven decision in maintenance, which is using Predictive Maintenance (PdM). Predictive Maintenance (PdM) is being used to estimate Remaining Useful Lifetime (RUL) from the machine. One of important machine in cement industry is raw mil lmachine, so estimating the RUL of this machine carry great importance. RUL prediction in this research will using deep learning with Transfomrer model that leverage Gated Convolutional Unit (GCU) for local extraction layer. THe use of Transformer with the hope of using it parallel computation mechanism capability to reduce computation time but able to capture long-term dependency in the data. Also, GCU is used to highlight local context in the data on the feature extraction. This research goal is to be used as reference on maintenance scheduling if there is failure indication from obtained RUL using Transformer-Gating Convolutional Unit method and analyze the estimated RUL result that is obtained from the prediction. In this research, multi sensor data from the raw mill is used from one of the Indonesia cement factory in 2015. With different range of time, T-GCU can predict machine’s RUL value with good performance indicated by the RMSE value observed from the raw mill machine ground truth RUL value.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deep Learning, Transformer Model, Predictive Maintenance, Remaining Useful Lifetime, Raw Mill Machine, Mesin Raw Mill
Subjects:	Q Science > QA Mathematics > QA273.6 Weibull distribution. Logistic distribution. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Berliana Putri Muliatama
Date Deposited:	30 Aug 2023 02:32
Last Modified:	30 Aug 2023 02:32
URI:	http://repository.its.ac.id/id/eprint/104195

Actions (login required)

View Item