Applying Bayesian networks in making intelligent applications for static and dynamic unbalance diagnosis

Abstract

One of the problems often encountered in vibration analysis is unbalanced or imbalanced, namely the occurrence of a shift in the center of mass from the center of rotation to cause high vibrations. Unbalance itself is divided into two, namely static and dynamic unbalance. Identification of the right type of unbalance must be done because each type of unbalance requires different handling. Therefore, this study aims to design a system to identify the type of unbalance based on the required parameters. The system design determines the input and then builds an algorithm by combining vibration analysis methods and Bayesian networks (BN). Systems and applications are built using MATLAB. After the application is finished, testing is carried out using vibration measurement data obtained from a demo machine that has previously been conditioned for damage. The BN method has been successfully applied to the unbalance diagnosis system. When there is evidence of large amplitude in 1X the frequency spectrum and the value of the static phase range, the percentage of static unbalance from 26.8% increases to 75%. The system can predict all testing data quickly and precisely for the six experiments.