PENGEMBANGAN WEBSITE UNTUK KLASIFIKASI KERUSAKAN KOLOM BETON MENGGUNAKAN AI PADA PROYEK PEMBANGUNAN X

BINTANG AKBAR ADI PRADANA, . (2026) PENGEMBANGAN WEBSITE UNTUK KLASIFIKASI KERUSAKAN KOLOM BETON MENGGUNAKAN AI PADA PROYEK PEMBANGUNAN X. Diploma thesis, UNIVERSITAS NEGERI JAKARTA.

	Text COVER.pdf Download (3MB)
	Text BAB I.pdf Download (3MB)
	Text BAB II.pdf Restricted to Registered users only Download (762kB) | Request a copy
	Text BAB III.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB IV.pdf Restricted to Registered users only Download (3MB) | Request a copy
	Text BAB V.pdf Restricted to Registered users only Download (469kB) | Request a copy
	Text DAFTAR PUSTAKA.pdf Download (487kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (3MB) | Request a copy

Abstract

Pengawasan kerusakan kolom beton pada proyek konstruksi masih dilakukan secara manual. Kelemahan berupa subjektivitas penilaian antar pihak quality control ditemukan pada metode tersebut. Perbedaan hasil identifikasi sering dijumpai pada objek kerusakan yang sama. Risiko keselamatan juga dihadapi oleh pihak quality control saat menjangkau kolom di ketinggian. Dokumentasi hasil pengawasan belum tersimpan dalam sistem yang seragam dan terpusat. Sistem deteksi otomatis berbasis teknologi computer vision dibutuhkan untuk mengatasi keterbatasan tersebut. Metode Research and Development (R&D) dengan model 4D (Define, Design, Develop, Disseminate) diterapkan dalam penelitian ini. Studi kasus dilakukan pada Proyek Pembangunan X di Kabupaten Bekasi. Bangunan tersebut terdiri dari empat lantai dengan 24 kolom beton yang rusak. Lima jenis kerusakan dideteksi, yaitu honeycomb, cracks, spalling, corrosion, dan segregation. Setiap jenis dibagi menjadi tiga tingkat keparahan sehingga terbentuk 15 kelas deteksi. Dataset sebanyak 3.000 gambar asli diperluas menjadi 5.400 gambar melalui augmentasi. Pembagian dataset dilakukan dengan rasio 80% latih, 10% validasi, dan 10% uji. Hasil evaluasi model YOLOv12 menunjukkan nilai precision sebesar 0,865 dan recall sebesar 0,807. Nilai mAP50 tercatat sebesar 0,853 dan mAP50-95 tercatat sebesar 0,704. Sistem website dikembangkan dengan FastAPI sebagai backend serta Vite dan React sebagai frontend. Model diekspor dalam format ONNX dan di-hosting melalui PythonAnywhere. Uji kelayakan dilakukan oleh dua validator. Produk dinyatakan layak dengan revisi oleh validator ahli TI. Produk dinyatakan layak tanpa revisi oleh validator ahli bangunan. Penelitian ini menghasilkan website deteksi kerusakan kolom beton berbasis YOLOv12. Sistem dirancang untuk memberi hasil penilaian yang objektif. Fitur unggah gambar, deteksi otomatis, bounding box, dan klasifikasi keparahan disediakan. Halaman Beranda, Analisis, Riwayat, Notifikasi, Akun, dan Pengaturan disediakan. Riwayat pemeriksaan, monitoring perbaikan, dan ekspor laporan juga tersedia. Ekspor laporan disediakan dalam format CSV, IFC Overlay, dan PDF. Sistem ini digunakan sebagai alat bantu awal dalam pengawasan kolom beton. Risiko kerja pada area tinggi dikurangi melalui analisis gambar. ***** Concrete column damage inspection on construction projects is still performed manually. Subjective assessment among inspectors is identified as a weakness of this method. Different identification hasils are often found on the same damage object. Safety risks are also faced by inspectors when accessing columns at elevated locations. Inspection documentation has not been stored in a uniform and centralized system. An automated detection system based on computer vision technology is needed to address these limitations. The Research and Development (R&D) method with the 4D model (Define, Design, Develop, Disseminate) was applied in this research. A case study was conducted on Construction Project X in Bekasi Regency. The building consisted of four floors with 24 damaged concrete columns. Five types of damage were detected: honeycomb, cracks, spalling, corrosion, and segregation. Each type was divided into three severity levels, hasiling in 15 detection classes. A dataset of 3,000 original images was expanded to 5,400 images through augmentation. The dataset was split into 80% training, 10% validation, and 10% testing data. The YOLOv12 model evaluation showed a precision value of 0.865 and a recall value of 0.807. The mAP50 was recorded at 0.853 and the mAP50-95 at 0.704. The website system was developed using FastAPI as the backend and Vite with React as the frontend. The model was exported in ONNX format and hosted through PythonAnywhere. A feasibility test was conducted by two validators. The product was declared feasible with revisions by the IT expert validator. The product was declared feasible without revisions by the building expert validator. This research produced a YOLOv12-based concrete column damage detection website. The system was designed to generate objective inspection hasils. Image unggah, automatic detection, bounding boxes, and severity classification are provided. The website includes Home, Analysis, Riwayat, Notification, Account, and Settings pages. Inspection Riwayat, repair monitoring, and report export are also provided. Reports can be exported in CSV, IFC Overlay, and PDF formats. The system is used as an initial tool for concrete column inspection. Work risks in elevated areas are reduced through image-based analysis.

Item Type:	Thesis (Diploma)
Additional Information:	1). Adhi Purnomo, M.T. ; 2). Rezi Berliana Yasinta, M.T.
Subjects:	Teknologi dan Ilmu Terapan > Teknik Sipil Teknologi dan Ilmu Terapan > Konstruksi Bangunan Teknologi dan Ilmu Terapan > Teknik Komputer
Divisions:	FT > D IV Teknologi Rekayasa Konstruksi Bangunan Gedung
Depositing User:	Bintang Akbar Adi Pradana .
Date Deposited:	21 Aug 2026 03:48
Last Modified:	21 Aug 2026 03:48
URI:	http://repository.unj.ac.id/id/eprint/72639

Actions (login required)

View Item