Operational Experience of Advanced HUMS Data Analysis Techniques

This paper describes the development of an advanced HUMS anomaly detection system, produced in support of a UK CAA research program to demonstrate the intelligent analysis of helicopter HUMS Vibration Health Monitoring data. The primary goals of the work are to improve HUMS fault detection capability and enhance system usability. A data driven modeling approach utilizes existing HUMS data to define models of ‘normal’ behavior, which can then be used to detect ‘abnormal’ or ‘anomalous’ behavior that may be associated with a defect. The anomaly modeling capability has been implemented in a web-based system and trialed on Bristow Helicopter’s North Sea Super Puma fleet. The trial confirmed that the system represents a significant advance in HUMS data analysis1, resulting in enhanced fault detection performance. The system has both corroborated existing HUMS indications and highlighted anomalous data that were not seen by the existing HUMS.