Sensor and Actuator Fault Detection and Isolation Using Extended Constrained Kalman Filters

Barron Associates, Inc. has developed a generic fault detection and isolation methodology that can be applied to any dynamic system.  This model-based approach is based on the use of extended constrained Kalman filters (ECKF) to detect and isolate sensor, actuator, and plant faults by exploiting analytic redundancy that exists among various subsets of actuator input and sensor output data. A computationally efficient statistical change detection technique is applied to a modified sequence of Kalman filter innovations to detect faults. 
Hardware redundancy relies on multiple sensors that nominally measure the same variable and fault decisions are usually limited to majority voting logic.  Obviously, the increase in sensors leads to an increase in cost, weight, and complexity.  Analytic redundancy, by contrast, eliminates the need for additional instrumentation hardware.  Implementation of analytic redundancy involves constructing observers for estimating the internal states of the plant and using the resulting state estimates to synthesize predicted sensor outputs.

Features and Benefits

Barron Associates' model-based health monitoring algorithms, including the ECKF technique, offer two important features not readily-available in expert systems or signal processing-based approaches:  virtual sensor signals and remaining useful life prediction.  In the event of a sensor failure, virtual sensor values can be provided to the control system until the failed sensor can be repaired or replaced.  The virtual sensor value is an optimal sensor state estimate that is calculated as part of the diagnostic process, so this potentially mission-saving feature is provided at no extra computational cost or algorithm complexity.  Remaining useful life estimation is also straightforward because the trend in detection statistics tracks the failure progression.  This valuable feature is also provided with very low additional complexity and computational burden, which minimizes the footprint of deployed systems.  Accurate fault isolation is especially crucial in safety- and mission-critical applications and few approaches simultaneously detect and isolate both sensor and actuator faults.    In summary, the ECKF methodology provides:

  • Analytic redundancy that eliminates costly hardware redundancy
  • Fast, accurate detection and isolation of both sensor and actuator faults
  • Computational efficiency that facilitates real-time implementation on low-cost hardware even for high-dimensional systems
  • Virtual sensor values in the event of a failed sensor
  • Reliable prognostics without requiring costly and/or impractical failure progression models

Recent Applications and Customers

Barron Associates has applied its ECKF technique to aircraft flight control systems for NASA Langley Research Center as part of the NASA Aviation Safety Program (AvSP) and the NASA Integrated Vehicle Health Management (IVHM) initiative.  Barron Associates has also worked with Lockheed Martin Aeronautics to provide diagnostic information to a reconfigurable flight control algorithm (see videos below).  These designs have been demonstrate in real time in NASA Integrated Flight Deck pilot-in-the-loop flight simulator. 

Barron Associates also developed an Integrated Diagnostic and Maintenance System for marine diesel engines for the Office of Naval Research (ONR), investigated a highly-dynamic diagnostic approach for the Naval Air Warfare Command (NAVAIR), autonomous health monitoring methods for hybrid propulsion vehicles for the U.S. Army TARDEC. 

Publications

E. Larson, B. Parker, and B. Clark, "Model-based sensor and actuator fault detection and isolation," in Proceedings of 2002 American Control Conference, May 2002, pp. 4215 - 4219.

N. Richards, T. Summers, J. Monaco, J. Burkholder, and D. Ward, "Retrofit fdi and adaptive outer-loop control for inflight failure accommodation and upset recovery," in Proceedings of the AIAA Guidance, Navigation, and Control Conference, No. AIAA-2005-5935, 2005.

T. Summers, J. Burkholder, J. Wadley, and D. Hopper,  "Integrated fdi enhancements for inflight failure accommodation and upset recovery," in Proceedings of the AIAA Guidance, Navigation, and Control Conference, No. AIAA-2006-6550, 2006.

Contact Info

Jason Burkholder:  burkholder@barron-associates.com