AUTOMATED-SYSTEM MONITORING AND DIAGNOSIS VIA SINGULAR-VALUE DECOMPOSITION

We present a new methodology for use in automated monitoring of a syst em to detect abnormal or degraded system behavior. The methodology use s the singular value decomposition technique of matrix algebra to disc over relationships between the elements of data observed from a normal ly operating system. It then tests these relationships against newly a cquired data to detect system malfunctions. Based on a theory for syst em identification in the presence of nuisance parameters, distribution theory developed specifically for this application allows the system monitor to properly set threshold levels so as to control the false al arm rate. The algorithm is numerically stable and relatively easy to a pply, and was tested successfully on data from a motorized globe valve in a nuclear power plant. (C) 1997 Elsevier Science Ltd.