Ahtmct-This paper describes the application of Fisher Discriminant Analysis and the Hotelling T2 statistic to the detection and classification of major failures that can occur in underwater vehicles. Simulation results are presented that demonstrate that rapid detection and reliable e~assifi-cation can be obtained with these methods. FAULT DETECTION A. OvermewThe first step in reconfigurable control is to detect the need to change the controller. The method for fault detection that is used in this research is the HotellinF: T2 statis-I. INTRODUCTION This paper describes our research results in detecting and classifying major faults in underwater vehicles that effect the maneuvering capabilities of the vehicle. Examples of these typcs of faults include jams in control surfaces or loss of the speed sensor. The methods described in this paper can be applied to other processes as well, particularly those with slow dynamics. This work is part of our ongoing research in reconfigurablc control.The fault detection step is performed using the Hotclling T2 statistic and Principal Component Analysis (PCA).PCA is used to identify the most significant variables and to reduce the dimensionality of the data for computational efficiency. The T2 statistic is computed based on the comparison of a set of measured variables ovcr a block of time with basis (training) data for those variables that show the expected behavior of the system during normal (no fault) operation. At present, two approaches to fault classification are being evaluated. The first approach is Fisher Discriminant Analysis (FDA). Measurcd data arc compared with basis data that represent each of the fault classes being considered. It is assumed that thew sets of fault basis data, as well as the basis data for normal operation, will he generated on-line. The second approach to fault classification that is being evaluated is Quantification of Contributing Variables (QCV). This method attempts to identifv the physical variable causing the fault, and the method does not require basis data to he generated for the various fault classes. Only the FDA approach is presented in this paper; results using QCV are given in [I]. Section I1 in this paper describes the T 2 statistic and principal component analysis as we are applying it in this research. Section 111 describes the use of Fisher Discriminant Analysis (FDA) for classifying faults. Results of simulation experiments are presented in Section IV, and conclusions are presented in Section V. This work w a~ performed under a research contract with the N a d Surface Warfare Center, Cardemrock Division. -tic [Z], [3], [4]. Process variables are measured during the current operation of the system, and those variables are statistically compared with basis data for the same variables representing the expected behavior of the system. In this research, the basis data are generated by simulation of the system dynamics under the appropriate operating conditions. In actual implementation, the basis data will be generated on-line thro...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.