Identifikasi cacat lintasan luar bantalan bola menggunakan Support Vector Machine (SVM) pada fan industri
Turbo: Jurnal Program Studi Teknik Mesin
Air regulation creates a comfortable and healthy environment for activities in the industry. A fan is a rotary machine that functions to regulate and circulate air in a room or work area. The bearing is an important component of a fan that is potentially damaged during operation. Damage to the bearing will interfere with perform the fan and can even disrupt a whole production process. A bearing condition monitoring method is needed that is effective and easy to use. Conventional methods such as
... nal methods such as spectrum analysis and sound analysis are not easy to use by operators in the field due to spectrum analysis requires spectrum reading experience while sound analysis is highly dependent on personal experience. This study proposes a vibration-based pattern recognition method that is Support Vector Machine (SVM) to detect damage to a bearing. This method effectively classifies bearing conditions and is easy to use. The study aims to obtain a method of detecting defects in single-row bearing outer paths on industrial fans using SVM. The study uses an industrial fan test rig with two bearing conditions, that is normal conditions (no defects) and external track defects with a depth of 1.4 mm defect and 0.4 mm width. Recording vibration signals using a data acquisition module with a sampling speed of 17066 Hz and a motor rotational speed of 2850 rpm. The SVM classifier is trained using 9 selected statistical parameters which are extracted from 700 sets of vibration signal recordings. The results showed the statistical parameters that were effectively used were Root Mean Square (RMS), Standard Deviation, Kurtosis, Variance, Entropy, Standard Error, Median, Signal-to-Noise and Distortion Ratio (SINAD), and Signal to Noise Ratio (SNR). The most optimal SVM model is obtained by applying combine Median-SINAD statistical parameters, with the same testing accuracy for the RBF, Polynomial and Linear kernels at 100%.