Fault detection and classification are crucial steps in the implementation of reconfigurable control. That work has involved the application of the Hotelling T² statistic to the detection of major faults in underwater vehicles, such as stern plane and rudder jams and loss of the speed sensor. Principal Component Analysis (PCA) is used to reduce the dimensionality of the data and improve the reliability of the T² statistic. Measured data are collected over blocks of time, and a summary value of the T² statistic is computed at the end of each block to accomplish the detection. Two methods are being studied for fault classification, namely, Fisher Discriminant Analysis (FDA) and Quantification of Contributing Variables (QCV). Simulation results indicate that these methods are capable of providing rapid and reliable detection and classification for these types of faults.
The most recent work has focused on the development of the reconfigurable control strategy to be followed once the fault detection and classification tasks have been performed. Reconfiguration to mitigate the effects of a stern plane jam in underwater vehicles has been accomplished. The results are very good. Not only can catastrophic behavior be prevented when a stern plane jam occurs, but it has been shown that desired depth and course trajectories can be closely followed as well.