Discussion on: "FDI using multiple parity vectors for redundant inertial sensors"

The paper by Cheol-Kwan Yang and Duk-Sun Shim represents an interesting contribution to the important area of the Fault Detection and Isolation (FDI) of dynamicsystems, in particular related to the problem of the Hardware Redundant FDI with Multiple Parity Vectors. More in detail, the work under discussion is focused on the Fault Detection, Isolation and Accommodation (FDIA) of redundant sensors when applied to Inertial Navigation Systems (INS). Cheol-Kwan Yang and Duk-Sun Shim have suggested the Averaged Parity Vector (APV) technique in order to improve the probability of correct isolation, as well as to reduce the rates of false alarms and wrong isolation. The proposed scheme exploits an accommodation threshold based on the error covariance of an estimated variable, which is related to the navigation accuracy of the INS. Moreover, as this accommodation threshold is used also for the fault detection, the APV algorithm is able only to detect non-tolerable faults. Thus, if the conditional probability of correct isolation is given, the minimum number of measurement data samples is determined with respect to the considered fault size. On the other hand, the fault accommodation provided by the APV method is based on the computation of averaged fault indication functions. Hence, if the maximum averaged fault indication function is greater than the accommodation threshold, the corresponding sensor will be excluded. In order to obtain better navigation accuracy, the use of the accommodation threshold is suggested, rather than the common thresholds determined from a given false alarm rate. Finally, computer simulations and experimental results using real data are proposed and compared with traditional FDI schemes for assessing the validity of the proposed APV technique.