Advanced Signal Processing Techniques for Vibration Analysis of Rotating Plant Machinery

The occurrence of vibration problems in rotating plant machinery has significant economic implications for aircraft ground-test facilities, and technological advances in predictive maintenance and health monitoring are needed. The increased availability of high-performance low-cost computer equipment combined with advanced high-speed signal processing techniques represent a relatively inexpensive way to increase machine health monitoring and predictive maintenance capabilties. Problems of particular difficulty are the separation of mixed vibratory signals which may overlap in both the time and frequency domains and fault signature identification. The proposed work recommends development of a digital signal processing scheme for fault detection and signature identification using cepstral decomposition of response signals. It is also proposed to develop a scheme for source point identification in a multi-source environment using a neural network technique. The work tasks include coding of the signal processing algorithm and feasibility demonstrations using simulated data. Simplified proof-of-concept experiments area also planned.