Failure Detection and Diagnosis Based on Machine Learning and Statistical Based Approach (PT Sreeya Sewu Case Study)

Christianta, Samuel Aditya (2024) Failure Detection and Diagnosis Based on Machine Learning and Statistical Based Approach (PT Sreeya Sewu Case Study). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010201126-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (2MB) | Request a copy

Abstract

Penelitian ini mengeksplorasi penerapan model deteksi kegagalan berbasis pembelajaran mesin dan pendekatan berbasis statistik untuk mengurangi downtime tak terencana di PT Sreeya Sewu, sebuah perusahaan produksi makanan yang berfokus pada peningkatan produktivitas di bidang operasi dan pemeliharaan. Studi ini berfokus pada penerapan model deteksi anomali dalam konteks pembelajaran tanpa pengawasan untuk mengidentifikasi gejala awal kegagalan. Dengan kemampuan identifikasi kegagalan, PT Sreeya Sewu dapat mengurangi downtime tak terencana yang berakibat pada peningkatan produktivitas di bidang pemeliharaan dan operasi. Untuk menangkap ketergantungan temporal dan sekuensial dalam data deret waktu, jaringan potensial yang dieksplorasi dalam penelitian ini adalah jaringan Long Short Term Memory (LSTM) Auto Encoder untuk model berbasis pembelajaran mesin dan Multivariate Exponentially Weighted Moving Average (MEWMA) untuk model berbasis statistik. Selain itu, penelitian ini menerapkan Exponentially Weighted Moving Average (EWMA) univariat untuk melakukan diagnosis kegagalan berdasarkan pola anomali sebelum kegagalan terjadi. Temuan ini diproyeksikan akan menguntungkan perusahaan dengan menyoroti konteks deteksi anomali dalam deteksi kegagalan, perbandingan model untuk melakukan deteksi kegagalan, dan maksimisasi data terbatas yang tersedia untuk memberikan pandangan awal tentang kemampuan model.
==============================================================================================================================
This research explores the implementation of a failure detection model based on machine learning and a statistical-based approach to reduce unplanned downtime for PT Sreeya Sewu, a food production company focusing on increasing productivity in the operation and maintenance area. The study focuses on implementing an anomaly detection model in an unsupervised learning context to identify early symptoms of failure. With the identification of failure capability, PT Sreeya Sewu can reduce unplanned downtime that results in increased productivity in the maintenance and operation area. To capture temporal and sequential dependencies in time-series data, a potential network explored in this research is the Long Short Term Memory (LSTM) Auto Encoder network for machine learning-based model and Multivariate Exponentially Weighted Moving Average (MEWMA) for statistical-based model. In addition, the research implements the univariate Exponentially Weighted Moving Average (EWMA) to conduct failure diagnosis based on the anomalous pattern before the failure exists. The findings are projected to benefit the company by highlighting the anomaly detection context in failure detection, the model comparison to conduct failure detection, and the maximization of available limited data to provide a preliminary view of model capability.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Failure Detection, Anomaly Detection, Machine Learning, LSTM Auto Encoder, EWMA, Failure Diagnosis
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA169 Reliability (Engineering) T Technology > TA Engineering (General). Civil engineering (General) > TA169.5 Failure analysis T Technology > TS Manufactures > TS174 Maintainability (Engineering) . Reliability (Engineering)
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Samuel Aditya Christianta
Date Deposited:	21 Aug 2024 03:54
Last Modified:	21 Aug 2024 03:54
URI:	http://repository.its.ac.id/id/eprint/112641

Actions (login required)

View Item