Diagnosis and prognosis of bearings using data mining and numerical visualization techniques

Traditionally, condition-based monitoring techniques have been used to diagnose failure in rotary machinery by application of low-level signal processing and trend analysis techniques. Such techniques consider small windows of data from large data sets to give preliminary information of developing fault(s) or failure precursor(s). However, these techniques only provide information of a minute portion of a large data set, which limits the accuracy of predicting the remaining useful life of the system. Diagnosis and prognosis (DAP) techniques should be able to identify the origin of the fault(s), estimate the rate of its progression and determine the remaining useful life of the system. This research demonstrates the use of data mining and numerical visualization techniques for diagnosis and prognosis of bearing vibration data. By using these techniques a comprehensive understanding of large vibration data sets can be attained. This approach uses intelligent agents to isolate particular bearing vibration characteristics using statistical analysis and signal processing for data compression. The results of the compressed data can be visualized in 3-D plots and used to track the origination and evolution of failure in the bearing vibration data. The Bearing Test Bed is used for applying measurable static and dynamic stresses on the bearing and collecting vibration signatures from the stressed bearings.