Induced draft (ID) fan is a crucial member to carry our hot flue gases and always subjected to high temperature and pressure. Flue gases carry higher temperature than atmospheric so, ID fan always works under elevated temperature 24x7. Supporting members such as by bearing and coupling also have to work under significant temperature more than atmospheric condition which leads create unbalance, misalignment or unexpected failures. Predictive maintenance has its own advantages to avoid this, Vibration analysis is a best tool for this. Vibration analysis clearly indicate the unbalance and misalignment fault in induce draft fan and coupling respectively as earliest. Vibration data has been collected on machine in paper mill and statistical features has been extracted. Machine learning techniques has its own significance for checking fault accuracies. Proposed study suggested unsupervised machine learning approach to separate 2 different faults like unbalance and misalignments faults in ID fan.