A multi sensor data fusion approach to fualt diagnosis of CSTR

Multi sensor data fusion has played a significant role in diverse areas. Various multi sensor data fusion methods have been extensively investigated by researchers. In this work, measured data derived from 5 installed sensors are addressed as individual evidences to infer process situation for a series of defined fault occurrences in a CSTR process plant. A multi sensor data fusion approach is developed based on the Dempster-Shafer evidence theory to fuse the individual evidences registered by the installed sensors for fault detection applications. An important issue relates to the mechanism this theory is employed to generate mass functions on the basis of the recorded information from sensors. Feature matrix is utilized to extract preliminary probability values and a qualitative method is then used to select mass functions. The developed technique has been successfully evaluated on the CSTR process plant.