Deteksi Korosi Besi dalam Beton Menggunakan 1D Convolution

Sutantyo, Evan Kinanda (2024) Deteksi Korosi Besi dalam Beton Menggunakan 1D Convolution. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024201070-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (12MB) | Request a copy

Abstract

Korosi besi dalam beton dapat menimbulkan risiko signifikan terhadap daya tahan dan integritas struktural infrastruktur sipil. Studi ini memperkenalkan pendekatan inovatif untuk mendeteksi korosi besi dalam beton menggunakan jaringan saraf konvolusional satu dimensi (1D-CNN). Beton dibuat dengan ukuran panjang 30 cm, lebar 20 cm, dan tinggi 10 cm dengan mutu K125, kemudian direndam dalam air garam 3\% dan dialiri listrik menggunakan adaptor 12V 0.5A untuk mempercepat proses korosi (efek elektrolisis). Pengambilan data dilakukan menggunakan LiteVNA 64 yang terkalibrasi untuk mendapatkan file S1P dan S2P dari sampel beton yang korosi dan tidak korosi. Data S1P dan S2P yang telah dikonversi ke format npy digunakan sebagai input untuk model 1D-CNN guna melatih dan menguji deteksi korosi pada besi beton. Model 1D-CNN bekerja dengan mengekstraksi fitur dari data input melalui lapisan konvolusional dan pooling, kemudian melakukan deteksi menggunakan lapisan dense terakhir untuk menentukan adanya korosi atau tidak pada besi dalam beton. Hasil pengujian menunjukkan bahwa model 1D-CNN mampu mendeteksi kondisi korosi pada besi beton dengan akurasi yang tinggi, sebagaimana ditunjukkan oleh metrik evaluasi seperti f1 score, precision, recall, dan accuracy.
==================================================================================================================================
Iron corrosion in concrete can pose significant risks to the durability and structural integrity of civil infrastructure. This study introduces an innovative approach for the detection of iron corrosion within concrete using a one-dimensional convolutional neural network (1D-CNN). The concrete was made 30 cm long, 20 cm wide, and 10 cm high of K125 grade, then immersed in 3% salt water and electrified using a 12V 0.5A adapter to accelerate the corrosion process (electrolysis effect). Data was collected using a calibrated LiteVNA 64 to obtain S1P and S2P files of the corroded and non-corroded concrete samples. The S1P and S2P data converted to npy format were used as input for the 1D-CNN model to train and test the detection of corrosion in rebar. The 1D-CNN model works by extracting features from the input data through convolutional and pooling layers, then performing detection using the last dense layer to determine whether or not there is corrosion of the steel in the concrete. The test results show that the 1D-CNN model is able to detect the corrosion condition of rebar with high accuracy, as indicated by evaluation metrics such as f1 score, precision, recall, and accuracy.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Corrosion detection, iron corrosion, concrete, electrolysis, 1D-CNN, LiteVNA 64, Deteksi korosi, korosi besi, beton, elektrolisis
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) T Technology > TA Engineering (General). Civil engineering (General) > TA1573 Detectors. Sensors
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	EVAN KINANDA SUTANTYO
Date Deposited:	24 Sep 2024 01:47
Last Modified:	24 Sep 2024 01:47
URI:	http://repository.its.ac.id/id/eprint/115671

Actions (login required)

View Item