In the current PHM research, most study focuses on the AI-based approach. Although worth pursuing, they are often impractical due to large fault data requirement. While there are ISO standards, they are not capable of detecting early fault nor classifying the fault types. In order to overcome these, a practical solution is given for fault diagnosis of rotating machinery, which consists of anomaly detection and distinction of shaft and bearing faults. The idea is based on the established findings on features for each fault. The performance is validated by several test cases to enable universal applications: Bently Nevada's RK4 test rig with imbedded faults, bearings underwent accelerated life test, and large shaft with crack under high speeds and torques.