1. ULUSLARARASI MÜHENDİSLİK BİLİMLERİ VE MULTİDİSİPLİNER YAKLAŞIMLAR KONGRESİ, İstanbul, Türkiye, 23 - 24 Şubat 2021, ss.80-86
Increasing the performance of induction motors, which are widely used in the industry, depends on the improvements to be made in some design steps. One of the most important of these design
steps is to create the stator winding structure and decide on the coil pitches. With appropriate stator coil
pitches structures, some single-row harmonics can be suppressed in the induction motors and motor
performance can be improved by keeping motor losses at a certain level. In this study, K – Nearest
Neighbor (KNN) model, which is a machine learning method, was designed to classify and predict the
stator coil pitches and tested with five-fold cross-validation. As a result of this experimental study, a
significant success was achieved by estimating both 100o
and 120o
coil pitches with 99.7% accuracy and
all winding steps with 92.3% average accuracy. It has been revealed that KNN algorithms are a machine
learning method that can be used effectively in a way that can contribute to electric machine design.