On-Line Detection and Identification of Faults and Abnormalities in Sensors for Ultra Precision Process Monitoring

In Ultra precision machining best results are obtained with on-line monitoring and adaptive control of various process parameters during machining. For a reliable on-line process monitoring and error compensation system, it is necessary to have accurate sensor readings. However, sometimes sensors may become faulty and due to failure it gives erroneous or constant values throughout the process. The problem of sensor validation is therefore a critical part of effective process monitoring. The objective of this study is to develop a sensor fault detection module which will be useful for different error compensation /diagnostic techniques needed for the ultra precision machine. A procedure based on Principal Component Analysis (PCA) is developed, which enables to perform detection and identification of sensor failures. PCA is a data driven modelling that transforms a set of correlated variables into a smaller set of new variables (principal components) that are uncorrelated and retain most of the original information. This new index is proposed in order to detect simple and multiple faults affecting the process and diagnose abnormalities in the original system in a robust way. The PCA model maps the sensor variables into a lower dimensional space and tracks their behaviour using Hot teling T² and Q statistics.