Sistem Deteksi Anomali Getaran Pada Blower Sentrifugal Menggunakan Metode Long Short-Term Memory Autoencoder

Anwar, Muhammad Fairuz Fawaidul (2026) Sistem Deteksi Anomali Getaran Pada Blower Sentrifugal Menggunakan Metode Long Short-Term Memory Autoencoder. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221119-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Blower sentrifugal merupakan komponen penting dalam proses produksi bahan kimia di PT LTG. Sistem monitoring yang telah ada hanya menampilkan kondisi getaran tanpa analisis
kondisi secara otomatis sehingga potensi terjadinya kerusakan sulit terdeteksi lebih awal. Penelitian ini mengembangkan sistem deteksi anomali menggunakan algoritma Long-Short Term Memory Autoencoder (LSTM-AE) sebagai pengembangan dari sistem yang telah ada serta menentukan konfigurasi arsitektur terbaik berdasarkan nilai F1-Score. Model dikembangkan menggunakan data getaran kondisi normal pada tiga sumbu kemudian deteksi anomali dilakukan berdasarkan nilai reconstruction error yang dibandingkan dengan nilai threshold. Hasil penelitian menunjukkan bahwa model LSTM-AE dapat mendeteksi seluruh data anomali tanpa menghasilkan false negative. Pengujian pertama menghasilkan F1-Score sebesar 99,13 persen, sedangkan pengujian kedua meningkat menjadi 99,75 Persen. Hasil tersebut menunjukkan bahwa algoritma LSTM- AE memiliki performa yang sangat baik dan efektif untuk diterapkan sebagai metode deteksi anomali pada sistem monitoring kondisi sehingga dapat memberikan informasi peringatan lebih awal sebelum terjadinya kerusakan serta mendukung penerapan pemeliharan di era industri 4.0.
========================================================================================================================================
The centrifugal blower is an essential component in the chemical production process at PT LTG. The existing monitoring system only displays vibration conditions without automatic condition analysis, making it difficult to detect potential damage early. This study develops an anomaly
detection system using the Long Short-Term Memory Autoencoder (LSTM-AE) algorithm to improve the existing system, and it determines the optimal architectural configuration based on the F1-Score. The model was developed using normal vibration data across three axes, and anomaly detection is then performed based on the reconstruction error value compared against a threshold. The results indicate that the LSTM-AE model can detect all anomalous data without producing false negatives. The first test yielded an F1-Score of 99.13%, while the second test
increased to 99.75%. These results demonstrate that the LSTM-AE algorithm performs excellently and is highly effective when implemented as an anomaly detection method in
condition monitoring systems. Consequently, it can provide early warning information before damage occurs and support the application of maintenance strategies in the Industry 4.0 era.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Blower Sentrifugal, Sistem Deteksi Anomali, LSTM-AE, Getaran, Centrifugal Blower, Anomaly Detection System, LSTM-AE, Vibration.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA935 Vibration
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Muhammad Fairuz Fawaidul Anwar
Date Deposited:	04 Aug 2026 07:45
Last Modified:	04 Aug 2026 07:45
URI:	http://repository.its.ac.id/id/eprint/143751

Actions (login required)

View Item