COMPUTER ASSISTED INCIPIENT FAULT DETECTION ON ROTATING AND RECIPROCATING MACHINES.

The monitoring and analysis of machine vibration signals for the purpose of establishing changes in condition are becoming more and more widely used. In particular in the the case of high speed rotating machines, the monitoring of vibration spectra has proved very successful for predicting incipient failure at a very early stage, thereby not only avoiding catastrophic failure, with the extra expense this would involve, but also allowing advance planning of repair work to minimize loss of production, often a more important economic factor in the continuous process industries. The main emphasis in this article is placed on the detection of significant change in the vibration signals, ie the detection of incipient faults. The basic methods can be implemented on relatively simple instrument systems based around an FFT analyzer and a desktop calculator but can also be extended to more powerful mini-computer systems when more data storage and processing are required (eg trend analysis) or when the intermittent spectrum monitoring considered here is to be combined with continuous monitoring of some simpler parameters over a large number of monitoring points (to provide rapid shutdown in the event of sudden failure).