Perancangan Sistem Deteksi Kebocoran Air Pada Pipeline Ultrasonic Water Flowmeter Berbasis Random Forest

Budiono, Safira Farsya Putri (2025) Perancangan Sistem Deteksi Kebocoran Air Pada Pipeline Ultrasonic Water Flowmeter Berbasis Random Forest. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009211140-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Kebocoran air pada jaringan distribusi rumah tangga menjadi salah satu penyebab utama tingginya angka kehilangan air (Non-Revenue Water) di Indonesia. Penelitian ini bertujuan
untuk mengetahui faktor-faktor yang memengaruhi performa deteksi kebocoran serta merancang sistem deteksi berbasis ultrasonic water flowmeter dan algoritma random forest. Data dikumpulkan melalui eksperimen sistem aliran tertutup (close loop) yang mensimulasikan tiga kondisi nyata: Normal, Bocor, dan Tidak Ada Aktivitas. Setelah dilakukan rekayasa fitur
seperti rolling mean, z-score, dan durasi, model dilatih menggunakan pendekatan gabungan threshold dan pola (pattern-based) serta diseimbangkan dengan SMOTE. Hasil evaluasi
menunjukkan bahwa model mencapai akurasi 97%, recall 97%, FPR 1,6%, dan FNR 3,1%. Sistem juga dilengkapi antarmuka GUI berbasis Python yang memudahkan pemantauan secara real-time. Model menunjukkan performa terbaik dalam mengenali kondisi kebocoran, terutama dalam mencegah false negative yang berdampak serius. Dengan performa tinggi, sistem ini
dapat menjadi solusi praktis, akurat, dan ekonomis dalam deteksi kebocoran air rumah tangga serta memiliki potensi implementasi lebih luas di bidang konservasi sumber daya air.

====================================================================================================================================

Water leakage in household distribution networks is one of the primary causes of high Non-Revenue Water (NRW) levels in Indonesia. This study aims to identify the factors influencing leakage detection performance and to design a leakage detection system using an ultrasonic water flowmeter and Random Forest algorithm. Data were collected through a
closed-loop experimental setup simulating three realistic water usage conditions: Normal, Leakage, and No Activity. Feature engineering was applied using rolling mean, z-score, and flow duration, and the model was trained using a combined threshold and pattern-based approach. Class imbalance was addressed using SMOTE oversampling. The validation results show that the model achieved 97% accuracy, 97% recall, a false positive rate of 1.6%, and a
false negative rate of 3.1%. A user-friendly GUI was developed using Python to monitor flow conditions and classification results in real-time. The system demonstrated high effectiveness in detecting leakage events, especially in reducing false negatives, which are critical for minimizing water loss. With its high performance,simplicity, and cost-efficiency, this system offers a promising solution for household water leakage monitoring and can contribute to
broader water resource conservation initiatives.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kebocoran air, Random Forest, Ultrasonic Flowmeter ======================================================================================================================== Water Leakage, Random Forest, Ultrasonic Flowmeter
Subjects:	T Technology > TD Environmental technology. Sanitary engineering > TD233 Water consumption
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Safira Farsya Putri Budiono
Date Deposited:	05 Aug 2025 03:53
Last Modified:	05 Aug 2025 03:53
URI:	http://repository.its.ac.id/id/eprint/127190

Actions (login required)

View Item