Perancangan Sistem Deteksi Lokasi Kebocoran Air Berbasis Feed-Forward Neural Network Pada Jaringan Distribusi Air di Perumda Air Minum Tugu Tirta Malang

Fernanda, Mohammad Alfaris (2023) Perancangan Sistem Deteksi Lokasi Kebocoran Air Berbasis Feed-Forward Neural Network Pada Jaringan Distribusi Air di Perumda Air Minum Tugu Tirta Malang. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311940000089-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (4MB) | Request a copy

Abstract

Berdasarkan SDGs PBB poin ke-enam, Air dan sanitasi yang cukup merupakan salah satu hak asasi manusia. Pengadaan dan penyediaan air bersih dilakukan oleh PERUMDA Air Minum. Dalam pelayanannya, kuantitas, kualitas, dan kontinuitas suplai perlu dijaga. Fenomena kehilangan air, menghambat preservasi ketiga aspek tersebut. Kehilangan air fisik adalah kehilangan air yang disertai munculnya aliran air riil seperti kebocoran. Target kehilangan air nasional berada pada angka 20% sedangkan rata-rata kehilangan air nasional adalah 40%. Sehingga, dibutuhkan metode deteksi lokasi kebocoran air pada jaringan distribusi air. Pada penelitian kali ini, diajukan suatu metode berbasis jaringan syaraf tiruan dengan algoritma I-ELM untuk mendeteksi lokasi dan ukuran kebocoran. Pertama, dilakukan pemilihan DMA sebagai objek yang akan diteliti. Lalu, spesifikasi DMA tersebut diambil untuk dilakukan pemodelan menggunakan EPANET. Kemudian, setelah DMA selesai dimodelkan dan divalidasi, dilakukan pemodelan kebocoran menggunakan emitter coefficient dengan rentangan kebocoran 0,5-1 LPS. Saat kebocoran disimulasikan, diambil tekanan pada 17 junctions yang telah ditentukan. Kemudian, titik kebocoran divariasikan lokasi dan ukurannya disepanjang tiga ruas pipa yang dipilih. Data tekanan diambil untuk setiap variasi kebocoran. Disisi lain, sistem deteksi lokasi dan ukuran bocor berbasis I-ELM dikembangkan. Data tekanan dengan 17 fitur disertai dengan dua variabel target berupa lokasi dan ukuran bocor diinputkan ke dalam sistem untuk dijadikan data training. Data testing didapatkan dari splitting set data training. Didapatkan hasil prediksi lokasi kebocoran dari data uji ketiga ruas pipa, memiliki deviasi rata-rata secara berturut-turut sebesar 3,7577, 2,4310, dan 0,5630. Didapatkan pula hasil prediksi ukuran bocor dari data uji ketiga ruas pipa, memiliki deviasi rata-rata secara berturut-turut sebesar 0,0807, 0,0585, dan 0,0543. Secara keseluruhan, didapatkan kemampuan prediksi lokasi bocor dengan RMSE overall sebesar 0,0516 dan RMSE overall untuk prediksi ukuran bocor sebesar 0,0696. Sistem deteksi lokasi bocor ini dibatasi dengan kondisi awal DMA yang ideal, dan hanya memiliki satu titik bocor.
==========================================================================================================================================
Based on the United Nations Sustainable Development Goals (SDGs), the sixth point emphasizes the importance of adequate water and sanitation, which is considered a fundamental human right. The provision of clean water is carried out by the Drinking Water Regional Public Company (PERUMDA Air Minum). In its service, the quantity, quality, and continuity of water supply need to be maintained. Water loss phenomenon hinders the preservation of these three aspects. Physical water loss refers to the loss of water accompanied by the emergence of actual water flow, such as leakage. The national target for water loss is 20%, while the national average for water loss is 40%. Therefore, a method is needed to detect the location of water leaks in the water distribution network. In this research, a neural network-based method using the I-ELM algorithm is proposed to detect the location and size of leaks. First, a District Metered Area (DMA) is selected as the object of study. The specifications of the chosen DMA are extracted to perform modeling using EPANET software. After the DMA is modeled and validated, leakage modeling is conducted using emitter coefficients with a range of 0.5-1 LPS (Liters per Second). When the leaks are simulated, pressure data is collected from 17 predetermined junctions. The leak points are varied in location and size along three selected pipeline sections. Pressure data is collected for each leak variation. On the other hand, an I-ELM-based leak location and size detection system is developed. Pressure data with 17 features, along with two target variables representing the leak location and size, are inputted into the system for training data. The testing data is obtained by splitting the training dataset. The results of leak location prediction from the testing data of the three pipeline sections have average deviations of 3.7577, 2.4310, and 0.5630, respectively. The results of leak size prediction from the testing data of the three pipeline sections have average deviations of 0.0807, 0.0585, and 0.0543, respectively. Overall, the leak location prediction has an overall RMSE (Root Mean Square Error) of 0.0516, and the leak size prediction has an overall RMSE of 0.0696. The leak detection system is limited to ideal initial conditions of the DMA and only considers a single leak point.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	I-ELM, Kebocoran Air, Lokasi Bocor, Tekanan, Ukuran Bocor, RMSE I-ELM, Leakage Location, Leakage Size, Pressure, RMSE, Water Leakage
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TD Environmental technology. Sanitary engineering > TD481 Water distribution systems
Divisions:	Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Mohammad Alfaris Fernanda
Date Deposited:	27 Jul 2023 04:39
Last Modified:	27 Jul 2023 04:39
URI:	http://repository.its.ac.id/id/eprint/99517

Actions (login required)

View Item