HYBRID COMPUTED ORDER TRACKING

Abstract

Vibration analysis is an integral part of modern condition monitoring and fault diagnosis systems for rotating machinery. Orders (cycles per revolution) are used as a frequency base for this analysis, thus making speed-related vibrations easier to detect. Fundamental to the performance of such systems is the accuracy and reliability of the required synchronously sampled vibration data. In this paper, the accuracy of three different synchronous sampling schemes are studied: a traditional hardware solution, computed order tracking and a hybrid of the two. Run-ups and run-downs are of particular interest in condition monitoring systems as they highlight many shaft defects. Also, because of the sometimes rapid shaft speed changes, this is just where the traditional approaches to producing synchronous sampling are prone to producing erroneous results. The three methods are assessed on data produced from a simulation of the rundown of a gas turbine shaft, typical to those found in the power industry. The use of this simulation allows the true accuracy of the techniques to be accessed, and inadequacies of traditional methods are clearly highlighted. The different sampling schemes rely on various interpolation algorithms. The accuracy and reliability of these algorithms is fundamental to the performance of the different sampling schemes, and hence a survey of the state-of-the-art interpolation algorithms is presented. This ensures that the most appropriate algorithms are identified, and as a result the novel computed order tracking technique introduced in this paper is shown to produce superior results.