Rancang Bangun Sistem Deteksi Real-Time Kesalahan Angular Misalignment pada Motor Induksi Menggunakan Sensor Arus Berbasis Deep Learning

Yaqin, Muhammad Ainul (2023) Rancang Bangun Sistem Deteksi Real-Time Kesalahan Angular Misalignment pada Motor Induksi Menggunakan Sensor Arus Berbasis Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311940000070-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (2MB) | Request a copy

Abstract

Sistem deteksi real-time kesalahan angular misalignment pada motor induksi diperlukan untuk sistem deteksi yang lebih mudah dan teramati secara kontinyu dan langsung sehingga ketika diimplementasikan pada industri dapat mencegah kerusakan mekanik motor induksi yang lebih parah hingga bisa mengganggu sistem produksi, kualitas produk, dan peningkatan biaya perbaikan. Pada penelitian ini dilakukan pengembangan berupa dibuatnya sistem deteksi kesalahan angular misalignment secara real-time. Pada penelitian ini digunakannya teknik windowing time series secara real-time sebelum dilakukan recurrence ploting. Kondisi yang dideteksi meliputi kondisi normal, misalignment 1°, dan 2°. Untuk masing-masing kondisi diambil 100 data dan dalam melakukan training perbandingan data training, validation, dan testing sebesar 8:1:1. Hasil training menunjukkan validasi akurasi sebesar 96,67%. Kesalahan misalignment anguler dapat dideteksi dengan model CNN secara real-time, yang diusulkan dengan input berupa recurrence plot dari hasil windowing time series sinyal arus stator secara real time untuk bisa diproses setiap 4 detik dan hanya memiliki delay pemrosesan sekitar 0,03 detik saja. Dari hasil sistem deteksi secara real-time yang diimplementasikan pada hardware didapatkan akurasi sebesar 96%, presisi minimal 94%, sensitivitas minimal 94%, spesifisitas minimal sebesar 97% dan NPV minimal 97%.
===================================================================================================================================
A real-time detection system for angular misalignment errors in induction motors is needed for a detection system that is easier and can be observed continuously and directly so that when implemented in industry it can prevent more severe mechanical damage to induction motors that can disrupt production systems, product quality, and increase repair costs. In this research, development was carried out in the form of making a real-time angular misalignment error detection system. In this study, windowing time series techniques were used in real time before recurrence plotting was carried out. Conditions detected include normal conditions, 1°, and 2° misalignment. For each condition, 100 data were taken and in conducting training, the comparison of training, validation, and testing data was 8:1:1. The training results show an accuracy validation of 96.67%. Angular misalignment errors can be detected with the CNN model in real-time, which is proposed with input in the form of a recurrence plot of the windowed time series results of the stator current signal in real-time to be processed every 4 seconds and only has a processing delay of around 0.03 seconds. From the results of a real-time detection system implemented on hardware, an accuracy of 96% is obtained, a minimum precision of 94%, a minimum sensitivity of 94%, a minimum specificity of 97%, and a minimum NPV of 97%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CNN, misalignment anguler, motor induksi, real-time, recurrence plot; angular misalignment, CNN , induction motor, real-time, recurrence plot
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK4055 Electric motor
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Ainul Yaqin
Date Deposited:	14 Sep 2023 06:00
Last Modified:	14 Sep 2023 06:00
URI:	http://repository.its.ac.id/id/eprint/102022

Actions (login required)

View Item