PEMODELAN DAN PREDIKSI CONDUCTED EMISSION PADA LED DRIVER MENGGUNAKAN MACHINE LEARNING DENGAN ALGORITMA JARINGAN SYARAF TIRUAN (ARTIFICIAL NEURAL NETWORK)

Authors

  • Nadya Sumardani PCR

Abstract

Electronic devices are often subject to unwanted emi interference which results in performance degradation or damage. Some solutions that have been used to reduce EMI problems, including spread spectrum techniques, shielding, converter design and filter design. In this study a prediction approach using machine learning is carried out which has advantages such as shortening the prediction time and helping to develop mitigation techniques. This study measured the EMI power level on the LED driver, the data obtained was then predicted to use machine learning using the ANN (Artificial Neural Network) algorithm method. Modeling and prediction are done by classification in two ways the learning process is supervised learning and unsupervised learning. The results of modeling and prediction carried out is that machine learning can predict conducted emission accurately with accuracy on the average periodic signal that is 82.24% and 80.299% for non-periodic signal predictions.  Keywords: Conducted Emission modeling prediction, Machine Learning, CISPR 22, LED Driver, ANN algorithm

Published

2020-09-09

Issue

Section

Artikel