Nonlinear Fault Detection, Isolation and Recovery Techniques for Unmanned Systems
Abstract
In order to avoid adverse consequences due to failures, it is desirable to have an advanced failure detection and isolation (FDI) system that detects and identifies anomalies early to minimize the damage, and that can remedy as many failures as possible. In complex systems, fault diagnosis is typically accomplished using a hierarchical approach. In our proposed autonomous unmanned vehicle (UAV) system, fault diagnosis, isolation and recovery (FDIR) is accomplished by using a hierarchical and decentralized approach. At this level of the hierarchy the model based or analytical redundancy based approach to FDIR would require a mathematical model of the process or sub-process under consideration. Based on this knowledge quantities called residuals will be generated. The residuals should be small or close to zero when there are no failures in the system. On the other hand, they should become nonzero and grow large if there are malfunctions in the system. This will accomplish the failure detection. The next important task will be the design of a fault isolation module that would isolate the faulty components or subsystems. There are two major approaches to the design and implementation of recovery procedures. One is to synthesize the procedures for every possible failure mode at the design stage. Once the diagnostic and recovery system is activated, it monitors the system and if it detects a failure, then the system will initiate the appropriate recovery procedure. In the other approach, suitable recovery procedures are generated "on-line" upon the detection of failures. In this report, we will examine the advantages and drawbacks of the above approaches in our framework.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 30, 2007
- Accession Number
- ADA475438
Entities
People
- K. Khorasani
Organizations
- Concordia University