Sistem Cerdas Deteksi Kebocoran dan Estimasi Lokasi pada Simulator Jaringan Pipa Menggunakan Artificial Neural Network (ANN) Berbasis Data Pressure Drop

Mahmudah, Juwita Maulidina (2026) Sistem Cerdas Deteksi Kebocoran dan Estimasi Lokasi pada Simulator Jaringan Pipa Menggunakan Artificial Neural Network (ANN) Berbasis Data Pressure Drop. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 2042221038-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Sistem perpipaan merupakan infrastruktur vital dalam berbagai sektor industri, seperti distribusi air bersih, minyak, dan gas. Namun demikian, kebocoran pada pipa sering terjadi akibat faktor usia material, kondisi lingkungan, maupun kerusakan mekanis. Kebocoran ini dapat menyebabkan kerugian signifikan, baik secara ekonomi maupun lingkungan. Metode deteksi konvensional seperti inspeksi visual dan metode akustik masih memiliki keterbatasan dalam mendeteksi kebocoran berskala kecil, terutama pada lokasi yang sulit dijangkau. Oleh karena itu, penelitian ini bertujuan untuk mengembangkan sistem cerdas pendeteksi kebocoran pipa berbasis artificial neural network (ANN), dengan memanfaatkan data pressure drop dan laju aliran fluida untuk mendeteksi sekaligus mengestimasi lokasi kebocoran secara lebih akurat. Metode yang digunakan dalam penelitian ini melibatkan pemasangan tiga unit pressure transmitter dan dua unit flow transmitter pada jaringan pipa PVC berdiameter ½ inci untuk memantau parameter tekanan dan laju aliran fluida. Data dikumpulkan interval 1 detik melalui pengujian kondisi normal dan kondisi kebocoran pada empat lokasi kebocoran dengan variasi bukaan valve sebesar 100%, 90%, 80%, 70%, 60%, dan 50% yang kemudian diproses menggunakan algoritma backpropagation yang diimplementasikan pada arsitektur multilayer perceptron (MLP). Model ANN dibangun menggunakan lima neuron input yang merepresentasikan kondisi tidak terjadi kebocoran, kebocoran pada valve 1, valve 2, valve 3, dan valve 4, dua hidden layer dengan konfigurasi 64, 32 neuron, dan 5 output berupa kelas kebocoran. Dataset dibagi menjadi 80% data pelatihan dan 20% data pengujian dengan parameter pelatihan berupa learning rate sebesar 0,001, batch size 32, menggunakan optimizer Adam, dropout 0,2 dan 101 epoch. Nilai training accuracy mencapai sekitar 97%, dengan nilai loss 0,0849.
============================================================================================================================================
Pipeline systems are vital infrastructure in various industrial sectors, such as clean water, oil, and gas distribution. However, pipeline leakage frequently occurs due to material ageing, environmental conditions, and mechanical damage. Such leakage can cause significant losses, both economically and environmentally. conventional detection methods, such as visual inspection and acoustic methods, still have limitations in detecting small-scale leaks, particularly in locations that are difficult to access. Therefore, this study aims to develop an intelligent pipeline leakage detection system based on an artificial neural network (ANN) by utilizing pressure drop and fluid flow rate data to detect and estimate leakage locations more accurately. The method used in this study involved the installation of three pressure transmitters and two flow transmitters on a ½ inch diameter PVC pipeline network to monitor pressure and fluid flow rate parameters. Data were collected at one-second intervals through testing under normal conditions and leakage conditions at four leakage locations with valve opening variations of 100%, 90%, 80%, 70%, 60%, and 50%, which were then processed using the backpropagation algorithm implemented in a multilayer perceptron (MLP) architecture. The ANN model was developed using five input neurons representing the no leakage condition, leakage at valve 1, valve 2, valve 3, and valve 4, two hidden layers with configurations of 64 and 32 neurons, and five outputs in the form of leakage classes. The dataset was divided into 80% training data and 20% testing data, with training parameters consisting of a learning rate of 0.001, a batch size of 32, the Adam optimizer, a dropout rate of 0.2, and 101 epochs. The training accuracy reached approximately 97%, with a loss value of 0.0849.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	Deteksi Kebocoran Pipa, Artificial Neural Network , Pressure Drop, Laju Aliran Fluida, Backpropagation, Pipeline Leakage Detection, Artificial Neural Network, Pressure Drop, Fluid Flow Rate, Backpropagation
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > TA Engineering (General). Civil engineering (General) > TA1573 Detectors. Sensors T Technology > TA Engineering (General). Civil engineering (General) > TA357 Computational fluid dynamics. Fluid Mechanics
Divisions:	Faculty of Vocational > Instrumentation Engineering
Depositing User:	Juwita Maulidina Mahmudah
Date Deposited:	05 Aug 2026 01:20
Last Modified:	05 Aug 2026 01:20
URI:	http://repository.its.ac.id/id/eprint/143708

Actions (login required)

View Item