Pengembangan Sistem Deteksi dan Lokalisasi Kebocoran Pipa Berbasis Deep Learning Residual-Network pada Jaringan Distribusi Air di Perumda Air Minum Tugu Tirta Malang

Prabowo, Zhiya Ulhaq (2024) Pengembangan Sistem Deteksi dan Lokalisasi Kebocoran Pipa Berbasis Deep Learning Residual-Network pada Jaringan Distribusi Air di Perumda Air Minum Tugu Tirta Malang. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009201156-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (4MB) | Request a copy

Abstract

Berdasarkan ringkasan SDGs 6.2 tahun 2021, bahwa 46% dari populasi dunia kekurangan air yang tersanitasi. PERUMDA Air Minum Tugu Tirta merupakan salah perusahaan yang berkontribusi dalam penyediaan dan distribusi air bersih. Kehilangan air fisik merupakan tantangan dalam distribusi air bersih yang dialami perusahaan. Demi mengatasi tantangan, diperlukan metode deteksi lokasi kebocoran pada jaringan distribusi air. Penelitian kali ini akan diajukan metode berbasis deep learning residual-network (ResNet) untuk mendeteksi ukuran dan lokasi kebocoran yang terjadi. Pertama, dilakukan pemilihan DMA sebagai objek penelitian. Spesifikasi DMA diambil untuk dilakukan pemodelan dengan software WaterCAD dan kemudian divalidasi serta dilakukan pemodelan kebocoran dengan mengatur Emitter Coefficient serta lokasi kebocorannya. Ukuran kebocoran adalah rendah 0,3 l/s, sedang 0,6 l/s, dan tinggi 1,2 l/s. Ketika kebocoran disimulasikan, data tekanan diambil pada empat titik sensor yang telah ditentukan pada DMA. Kemudian, dilakukan variasi lokasi dengan variasi ukuran yang sama. Data tekanan aktual dari DMA juga diambil dengan menggunakan sensor tekanan yang dipasang pada pressure point dari jaringan. Sistem deteksi dan lokalisasi berbasis ResNet dikembangkan dengan menggunakan data simulasi dalam proses pelatihan sistem. Dari proses pelatihan tersebut didapatkan hasil performa akurasi deteksi ukuran sebesar 98.62% dan lokalisasi sebesar 98.16%, serta F1-Score deteksi ukuran sebesar 98.97% dan lokalisasi sebesar 98.17%. Akan tetapi, pada percobaan dengan data aktual, performa dapat disebut kurang baik, dimana akurasi deteksi ukuran dan lokalisasi sebesar 75%, serta F1-Score deteksi ukuran sebesar 85.71%. dan lokalisasi sebesar 85.71%. =================================================================================================================================
According to the 2021 SDGs 6.2 summary, 46% of the world's population lacks access to sanitized water. PERUMDA Air Minum Tugu Tirta is one of the companies contributing to the provision and distribution of clean water. Physical water loss is a significant challenge in the distribution of clean water faced by the company. To address this challenge, a method for detecting leak locations in the water distribution network is necessary. This study proposes a deep learning residual-network (ResNet)-based method to detect the size and location of leaks. First, a District Metered Area (DMA) was selected as the research object. The DMA specifications were used for modeling with WaterCAD software. The model was validated, and leak modeling was performed by adjusting the Emitter Coefficient and the location of the leaks. The leak sizes were categorized as low (0.3 l/s), medium (0.6 l/s), and high (1.2 l/s). During the leak simulation, pressure data was collected at four predetermined sensor points within the DMA. Various locations with the same size variations were tested. Actual pressure data from the DMA was also collected using pressure sensors installed at pressure points in the network. The ResNet-based detection and localization system was developed using simulation data for the system training process. The training process yielded a detection size accuracy is 98.62% and a localization accuracy is 98.16%, with an F1-Score for detection size is 98.97% and for localization is 98.17%. However, when tested with actual data, the performance was less satisfactory, with a detection size and localization accuracy of 75%, and the F1-Score for detection size and localization is 85.71%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deep Learning, F1-Score, Leak Localization, ResNet, Water Leak, Kebocoran Air, Lokalisasi Bocor
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > TD Environmental technology. Sanitary engineering > TD481 Water distribution systems
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Zhiya Ulhaq Prabowo
Date Deposited:	30 Jul 2024 01:47
Last Modified:	30 Jul 2024 01:47
URI:	http://repository.its.ac.id/id/eprint/109409

Actions (login required)

View Item