| Summary: | Bibliography: p. 181-183. === The analysis of the vibrations produced by roller bearings is one of the most widely used techniques in condition determination of rolling element bearings. This project forms part of an overall plan to gain experience in condition monitoring and produce a computer aided vibration monitoring system that would initially be applied to rolling element bearings, and then later to other machine components. The particular goal of this project is to study signal processing techniques that will be of use in this system. The general signal processing problems are as follows. The vibration of an undamaged bearing is characterised by a Gaussian distribution and a white power spectral density. Once a bearing is damaged the nature of the vibration changes often with spikes or impulses present in the vibration signal. By detecting these impulses a measure of the condition of the bearing may be obtained. The primary goal in machine condition determination then becomes the detection of these impulses in the presence of noise and contaminating. signals and to discriminate between those caused by the component in question and those from other sources. A wide range of signal processing techniques were reviewed and some of these tested on vibrations recorded on the Mechanical engineering departments bearing test rig. It was found that the time domain statistics (RMS, kurtosis, crest factor) were the simplest to use, but could be unreliable. On the other hand, frequency domain analysis techniques, such as the power spectrum were more reliable, but more difficult to apply. By making use of a variety of these techniques and applying them in a systematic manner, it is possible to make an assessment of bearing condition under a wide variety of operating conditions. A small number of the signal processing techniques were programmed for a DSP processor. It was found that all of the techniques, with the exception of the bispectrum could be programmed for the DSP chip. It was found however that the available DSP card did not have sufficient memory to allow analysis and preprocessing routines to be combined. In addition to this the analogue to digital conversion system would benefit from a buffered IO system. The project should continue, with the DSP card being upgraded and all the necessary signal processing routines programmed. The project can then move to the next phase which would be inclusion of display and interface software and Artificial Intelligence analysis aids.
|
|---|
