In modern manufacturing environments, machinery failures are predictable. Vibration monitoring is the most widely
used technique. A performance study on the critical bearing vibration of a lathe was carried out. Estimation of bearing vibration
was done by measuring bearing vibration along horizontal and vertical directions, flank wear (maximum and average) of the
tool, machining time and tool tip temperature. From the experimental analysis it was observed that the vibration velocity depends
on cutting conditions. For reliable estimate of bearing vibration, methods like multiple regression analysis and Group Method of
Data Handling (GMDH) have been applied. It was observed that the multiple regression estimates give better results for higher
speeds. For GMDH, three criterions were considered. Regularity and combined criterion gave better representation. The
comparison studies on estimates of bearing vibration by multiple regression analysis and GMDH were made.
