DIAGNOSIS KETIDAKLURUSAN (MISALIGNMENT) POROS MENGGUNAKAN METODE MULTICLASS SUPPORT VECTOR MACHINE (SVM)

Abstract

Misalignment is a condition where the centerlines of two coupled shafts do not coincide. Misalignment is the commonly fault in rotating machinery. Detection and diagnosis of shaft misalignment is crucial to achieve its optimal performance. The purpose of research is to diagnose shaft misalignment using multiclass support vector machine (SVM). The time-domain vibration signals of a shaft alignment rig with normal, parallel misalignment and angular misalignment of shaft conditions were obtained from vibration measurement signals. The accelerometer was used to measure vibration with a sampling frequency of 20 kHz at the constant speed operation of 1000 rpm. The features of median, RMS, crest factor, variance, kurtosis, shape factor, impulse factor, skewness, range, standard deviation and maximum were extracted from the vibration signal. The Principal Component Analysis (PCA) was applied for reduce the number of variables for data input to principal components with lower dimension. The multiclass SVM with One Against One (OAO) methodand linear kernel were used for classification. The results show that SVM for diagnosis of shaft misalignment show a good performance with an accuracy of 100%.