Sistem Monitoring berbasis IOT untuk Predictive Maintenance Generator Set Menggunakan Machine Learning

Apriliya, Mady Deka and , Prof. Ir. Supriyono, ST, MT, Ph.D and , Joko Sedyono, ST, M.Eng, Ph.D (2025) Sistem Monitoring berbasis IOT untuk Predictive Maintenance Generator Set Menggunakan Machine Learning. Thesis thesis, Universitas Muhammadiyah Surakarta.

	PDF (Naskah Publikasi) artikel publikasi mady deka apriliya.pdf Download (1MB)
	PDF (Halaman Depan) halaman depan ok.pdf Download (1MB)
	PDF (Bab I) Bab I.pdf Download (249kB)
	PDF (Bab II) Bab II.pdf Restricted to Repository staff only Download (582kB)
	PDF (Bab III) Bab III.pdf Restricted to Repository staff only Download (1MB)
	PDF (Bab IV) Bab IV.pdf Restricted to Repository staff only Download (2MB)
	PDF (Bab V) Bab V.pdf Restricted to Repository staff only Download (221kB)
	PDF (Daftar Pustaka) daftar pustaka.pdf Download (223kB)
	PDF (Lampiran) lampiran tesis.pdf Restricted to Repository staff only Download (4MB)
	PDF (Surat Pernyataan Publikasi) surat pernyataan publikasi mady deka apriliya.pdf Restricted to Repository staff only Download (648kB)

Abstract

Generator sets are a crucial component of modern infrastructure as a backup or primary power source. However, traditional reactive or preventive maintenance is often inefficient and prone to human error, which can lead to unexpected downtime and high operational costs. As a solution, this study proposes the development of a predictive maintenance (PdM) system based on the Internet of Things (IoT) and Machine Learning (ML) to monitor generator set conditions in real time. The system is designed using an ESP32 microcontroller and various sensors to collect critical data such as voltage, current, temperature, oil pressure, engine speed, vibration, fuel level, and operating hours. The collected data is then sent and displayed via a web dashboard using the Node-RED platform. This study also implements and compares three Machine Learning algorithms—Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN)—to predict potential generator set failures. The implemented IoT-based monitoring system demonstrates stable and reliable performance, with adequate sensor data acquisition accuracy. Based on the model comparison, the Random Forest algorithm performed best, with an overall accuracy of 98.47%. This model was highly effective in classifying Anomaly and Normal conditions, and demonstrated significant improvement in detecting Failure classes with few false negatives. Meanwhile, the SVM and KNN models achieved an accuracy of around 83%, but showed weakness in detecting Failure conditions, with the SVM only successfully detecting 42% and the KNN only 56% of actual failure cases. This study successfully demonstrated that a predictive maintenance (PdM) system based on the Internet of Things (IoT) and Machine Learning (ML), with the Random Forest model as a predictor, can improve generator set reliability through early failure detection, reducing unplanned downtime, and optimizing maintenance costs. This provides an important reference for implementing similar technologies in industry

Item Type:	Thesis (Thesis)
Uncontrolled Keywords:	Predictive Maintenance, Internet of Things, Machine Learning, Generator Set, Random Forest
Subjects:	T Technology > TI Industrial Engineering T Technology > TJ Mechanical engineering and machinery
Divisions:	Fakultas Teknik > S2 Teknik Mesin
Depositing User:	MADY DEKA APRILIYA
Date Deposited:	15 Aug 2025 08:43
Last Modified:	15 Aug 2025 08:43
URI:	http://eprints.ums.ac.id/id/eprint/138286

Actions (login required)

View Item