Sistem Penentuan Kecepatan Konveyor Berdasarkan Objek Yang Terdeteksi Menggunakan Convolutional Neural Network Pada Sorting Machine

Frizzy, Jonathan Oktaviano (2026) Sistem Penentuan Kecepatan Konveyor Berdasarkan Objek Yang Terdeteksi Menggunakan Convolutional Neural Network Pada Sorting Machine. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221060-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Sorting Machine merupakan sistem pemilahan sampah otomatis menggunakan pengolahan citra. Pengujian awal menunjukkan konveyor dengan kecepatan konstan menghasilkan tingkat keberhasilan perpindahan objek 72,7% dari 150 percobaan, dengan kegagalan 27,3% yang terjadi konsisten pada empat jenis objek, yaitu botol plastik 600 ml, botol HDPE, kertas A4, dan kaleng aluminium. Objek keluar jalur sebelum mencapai titik pemilahan karena perbedaan luas permukaan antarobjek tidak terakomodasi oleh kecepatan tunggal. Penelitian ini menentukan setpoint kecepatan konveyor menggunakan informasi visual objek melalui Convolutional Neural Network (CNN), yaitu klasifikasi jenis objek menggunakan YOLOv8n dan perhitungan luas permukaan dua dimensi melalui segmentasi menggunakan LightUNet. Jenis objek dan luas permukaan hasil segmentasi menjadi parameter penentu setpoint RPM yang dipetakan secara empiris dalam Lookup Table (LUT). Setpoint tersebut direalisasikan oleh kontroler Proportional-Integral-Derivative (PID) closed-loop dengan sensor RPM sebagai feedback agar kecepatan aktual motor konvergen pada setpoint. Pendekatan ini tidak menambah sensor maupun mengubah mekanik mesin. Hasil pengujian menunjukkan YOLOv8n mencapai mAP@0,5 sebesar 0,995 dan LightUNet mencapai mean IoU 0,989 dengan error perhitungan luas permukaan 1,85 sampai 2,35%. Pemetaan LUT menghasilkan setpoint 154 RPM untuk HDPE, 138 RPM untuk Plastik, 104 RPM untuk Kaleng, dan 89 RPM untuk Kertas, yang direalisasikan skema closed-loop dengan penurunan rata-rata error RPM sebesar 46,3% (40,3% hingga 52,8% pada empat setpoint operasional) dibandingkan open-loop, sehingga tingkat keberhasilan perpindahan objek meningkat dari 72,7% menjadi 94,0% dan kegagalan menurun dari 27,3% menjadi 6,0% pada pengujian end-to-end.
=======================================================================================================================================
The Sorting Machine is an automated waste-sorting system based on image processing. Initial testing showed that a conveyor operating at constant speed achieved an object-transfer success rate of only 72.7% across 150 trials, with 27.3% failures occurring consistently on four object types, namely a 600 ml plastic bottle, an HDPE bottle, A4 paper, and an aluminium can. Objects left the track before reaching the sorting point because the differences in surface area between objects were not accommodated by a single speed. This study determines the conveyor speed setpoint using visual object information through a Convolutional Neural Network (CNN), namely object-type classification using YOLOv8n and two-dimensional surface-area computation through segmentation using LightUNet. The object type and the computed surface area serve as the parameters determining the RPM setpoint, which is mapped empirically in a Lookup Table (LUT). This setpoint is realized by a closed-loop Proportional-Integral-Derivative (PID) controller with an RPM sensor as feedback, so that the actual motor speed converges to the setpoint. The approach adds no sensors and does not modify the machine mechanics. Testing showed that YOLOv8n achieved an mAP@0.5 of 0.995 and LightUNet achieved a mean IoU of 0.989 with a surface-area computation error of 1.85 to 2.35%. The LUT mapping produced setpoints of 154 RPM for HDPE, 138 RPM for Plastik, 104 RPM for Kaleng, and 89 RPM for Kertas, realized by the closed-loop scheme with an average RPM error reduction of 46.3% (40.3% to 52.8% across four operational setpoints) compared with the open-loop scheme, increasing the object-transfer success rate from 72.7% to 94.0% and reducing failures from 27.3% to 6.0% in end-to-end testing.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Convolutional Neural Network, Lookup Table, PID closed-loop, Sorting Machine, setpoint RPM. Convolutional Neural Network, Lookup Table, PID closed-loop, Sorting Machine, setpoint RPM.
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > TJ Mechanical engineering and machinery > TJ1398 Conveyors T Technology > TJ Mechanical engineering and machinery > TJ223 PID controllers T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Jonathan Oktaviano Frizzy
Date Deposited:	03 Aug 2026 08:21
Last Modified:	03 Aug 2026 08:21
URI:	http://repository.its.ac.id/id/eprint/142441

Actions (login required)

View Item