Machines Fault Detection

Diagnostics technologies are used to increase rotating machines efficiency in energy systems by detecting upcoming faults. Small rotating machines usually do not have on board diagnostic units. Portable diagnostic units are expensive and require very detailed information about monitored machinery, from the diameter of rolling elements in the bearings, to the number of rotor bars. Therefore, there is an area of opportunity to develop a low cost diagnostic unit that does not require detailed machine information. Modern smartphones seems suitable for this task because they have built-in acoustic and vibration data acquisition and considerable computing capacity. However, they have hardware limitations compared to state-of-the-art diagnostic units such as data sampling rate and sensors sensitivity.

A set of induction motors are tested in both, healthy and faulty conditions (unbalanced rotor, damaged bearings and broken rotor bars) to analyze vibration and acoustic signals recorded with a smartphone. Then, recorded data is analyzed to identify healthy and faulty emissions signatures. A total of roughly 85 minutes of acoustic emissions and around 125 minutes of vibration data are recorded along all different operating conditions. Results show that it is possible to estimate machine rotational speed and detect faults with the smartphone recordings. Acoustic emissions’ faulty signature is located between 4 KHz – 8 KHz in the form of high magnitude frequency clusters and speed can be estimated using mechanical rotational frequency harmonics present between 100 Hz- 1 KHz. Similarly Vibration faulty signature is located along the frequency spectrum in the form of high magnitude peaks and rotational speed can be estimated by using peak vibration frequency. Finally a fully functional Android application was develop based on test results to automatically detect motor speed and health status. Validation testing showed 90% accuracy in fault detection.