Sistem Klasifikasi Untuk Mendeteksi Adanya Penggumpalan Pupuk Urea Pada Mesin Hopper Menggunakan Metode K-Nearest Neighbors

Brillianto, Muhammad Rakha (2025) Sistem Klasifikasi Untuk Mendeteksi Adanya Penggumpalan Pupuk Urea Pada Mesin Hopper Menggunakan Metode K-Nearest Neighbors. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040211085_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Pemantauan kondisi penggumpalan pupuk Urea pada mesin Hopper TKD710D di PT Petrokimia Gresik masih dilakukan secara berkala oleh operator. Metode ini cukup memakan waktu dan tidak selalu akurat karena bergantung pada pengamatan visual. Padahal, penggumpalan dapat terjadi saat suhu dan kelembapan pupuk Urea berada di luar batas standar perusahaan, yang berisiko mengganggu mutu produk dan kelancaran distribusi. Penelitian ini mengembangkan sistem klasifikasi untuk membantu pemantauan tersebut secara otomatis menggunakan algoritma K-Nearest Neighbors (K-NN). Sistem membaca data suhu dan kelembapan dari sensor, mengolahnya melalui Node-RED, menyimpan data ke dalam database MySQL, dan menampilkan hasil klasifikasi secara langsung melalui antarmuka website. Kondisi pupuk dikategorikan ke dalam tiga kelas, yaitu Tidak Menggumpal, Potensi Menggumpal, dan Menggumpal. Pengujian model dilakukan menggunakan validasi silang 5- Fold Cross Validation, dengan hasil akurasi sebesar 93,79% dan nilai F1-score yang stabil untuk tiap kelas. Selain itu, sistem diuji langsung di lapangan sebanyak 10 kali dan menunjukkan hasil klasifikasi yang sesuai dengan kondisi fisik pupuk Urea di hopper. Sistem juga mampu memberikan notifikasi secara otomatis saat kondisi menggumpal terdeteksi.
==========================================================================================================================================
Monitoring the caking condition of urea fertilizer in the Hopper TKD710D at PT Petrokimia Gresik is still carried out manually by operators. This method can be timeconsuming and is not always accurate, as it relies on visual observation. In fact, caking may occur when the temperature and humidity of the fertilizer exceed normal thresholds, which can affect product quality and disrupt the distribution process. This study develops a classification system to assist the monitoring process automatically using the K-Nearest Neighbors (K-NN) algorithm. The system reads temperature and humidity data from sensors, processes it through Node-RED, stores the data in a MySQL database, and displays the classification results directly through a website interface. The fertilizer condition is categorized into three classes: NonCaking, Potential Caking, and Caking. Model testing was performed using 5-Fold Cross Validation, yielding an accuracy of 93.79% and consistent F1-scores across all classes. Additionally, the system was tested in the field 10 times and showed classification results that matched the actual physical condition of the fertilizer in the hopper. The system is also capable of automatically issuing notifications when a caking condition is detected.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	KNN, Pupuk Urea, Hopper, Node-RED, Klasifikasi, Penggumpalan, Suhu,Kelembapan
Subjects:	Q Science > Q Science (General) > Q180.55.M38 Mathematical models Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) Q Science > QA Mathematics > QA76.9D338 Data integration
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Muhammad Rakha Brillianto
Date Deposited:	04 Aug 2025 09:44
Last Modified:	04 Aug 2025 09:44
URI:	http://repository.its.ac.id/id/eprint/125292

Actions (login required)

View Item