Enhancing the Performance of Direction Sensetive Filters for Multiple Failures

For improving reliability of control systems and from safety considerations, it is important to detect and identify faults in a control system. Faults occuring in a linear dynamic system can be modeled in several ways. This paper concentrates on additive type of failures. For detecting and identifying additive type of failures, Beard-Jones (BJ) filters can be used. These filters are observers designed in such a way that when a failure occurs, the error in the measured and estimated output grows in a particular direction in the output error space. The problem associated with this kind of filters is that the number of different failures it can identify is limited by the number of outputs. To overcome this problem, a bank of BJ filters along with likelihood ratio tests is porposed. Each filter in the bank will be designed for different groups of failures. The properties of these filters when coupled with likelihood ratio test give some intersting properties to the ratio test results.