Sistem Inspeksi Kualitas Tutup Botol Pestisida Menggunakan Metode Convolutional Neural Network

Ramadhan, Adn Nove (2024) Sistem Inspeksi Kualitas Tutup Botol Pestisida Menggunakan Metode Convolutional Neural Network. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 2040201124-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (12MB) | Request a copy

Abstract

Integritas kemasan memegang peranan krusial dalam industri manufaktur pestisida sebagai indikator kualitas produk akhir. Perusahaan Pestisida, yang berlokasi di Mojokerto dan beroperasi sejak tahun 2005, menghadapi tantangan signifikan dalam memastikan keutuhan dan keamanan produk pestisida yang diproduksi. Proyek Akhir ini bertujuan untuk mengembangkan dan mengimplementasikan sistem inspeksi otomatis untuk meningkatkan kualitas kontrol tutup botol pestisida. Proses produksi melibatkan beberapa tahap, termasuk pemuatan botol secara manual, pengisian otomatis, penutupan tutup botol menggunakan mesin, dan pemanasan aluminium seal dengan mesin Induction Cap Sealing. Saat ini, inspeksi dilakukan dengan sensor ultrasonik untuk mendeteksi keberadaan aluminium seal dan inspeksi manual untuk mendeteksi cacat fisik tutup botol. Data produksi harian menunjukkan adanya 1,27% unit tutup botol yang ditolak (reject), disebabkan oleh berbagai faktor seperti tutup botol yang miring, kontaminasi pestisida, kegagalan penempatan aluminium seal, dan kerusakan pada tutup botol. Untuk mengatasi masalah ini, sebuah model deteksi berbasis Convolutional Neural Network (CNN) dengan model YOLOv5 telah berhasil dikembangkan untuk mengklasifikasikan tutup botol ke dalam beberapa kategori, yaitu tutup normal, rusak, terkontaminasi, miring dan tutup terlalu tinggi. Metodologi yang digunakan melibatkan pelatihan model dengan dataset gambar yang telah dianotasi, seta pengujian performa model dari jarak 5 cm hingga 30 cm. Berdasarkan analisis hasil pengujian, sistem menunjukkan akurasi yang tinggi dengan nilai precision, recall, dan F1 score yang optimal pada jarak 15 cm hingga 25 cm, mencapai nilai 1 untuk semua metrik pada jarak tersebut. Hasil ini menunjukkan bahwa sistem inspeksi yang diusulkan mampu meningkatkan akurasi dan efisiensi dalam proses kontrol kualitas tutup botol pestisida, berpotensi mengurangi tingkat penolakan dan meningkatkan kualitas produk akhir.
=================================================================================================================================
Packaging integrity plays a crucial role in the pesticide manufacturing industry as an indicator of the final product's quality. The Pesticide Company, located in Mojokerto and operating since 2005, faces significant challenges in ensuring the integrity and safety of the produced pesticides. This Final Project aims to develop and implement an automatic inspection system to enhance the quality control of pesticide bottle caps. The production process involves several stages, including manual bottle loading, automatic filling, machine capping, and heating the aluminum seal with an Induction Cap Sealing machine. Currently, inspection is perfored using ultrasonic sensors to detect the presence of the aluminum seal and manual inspection to detect physical defects in the bottle caps. Daily production data shows a 1.27% rejection rate of bottle caps due to various factors such as tilted caps, pesticide contamination, aluminum seal placement failure, and damaged caps. To address this issue, a detection model based on Convolutional Neural Network (CNN) using the YOLOv5 model has been successfully developed to classify bottle caps into several categories: normal caps, damaged, contaminated, tilted and too high. The methodology involved training the model with an annotated image dataset and testing the model's performance at distances ranging from 5 cm to 30 cm. Based on the analysis of the test results, the system demonstrated high Accuracy with optimal precision, recall, and F1 score values at distances of 15 cm to 25 cm, achieving a value of 1 for all metrics at these distances. These results indicate that the proposed inspection system can improve the Accuracy and efficiency of the pesticide bottle cap quality control process, potentially reducing the rejection rateand enhancing the final product's quality.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	Convolutional Neural Network, Integritas Kemasan, Inspeksi Tutup Botol, Product Integrity, Bottle Cap Inspection.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK6592.A9 Automatic tracking. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7878 Electronic instruments
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Adn Nove Ramadhan
Date Deposited:	26 Sep 2024 06:00
Last Modified:	26 Sep 2024 06:00
URI:	http://repository.its.ac.id/id/eprint/115690

Actions (login required)

View Item