The purpose of this project is to assess the feasibility of a Kalman Filter approach for fault detection in a highly unstable system, specifically the heart pump currently under development at RIT. Simulations and experimental work were completed to determine the effects of possible position sensor fault conditions on the system; that information was then used in conjunction with a pair of Kalman filters to create a method of detecting faults and providing fault-tolerant operation. The heart pump system was modeled using Simulink and then the fault diagnosis and tolerance system was added to the model and tested via simulation in SIMULINK TM. The simulations showed the filters were able to calculate and remove bias caused by any type of position sensor error, provided the estimated plant model is nearly identical to the actual plant model. Sensitivity analysis showed that the fault detection/fault-tolerance method is extremely sensitive to discrepancies between the estimated plant model and actual pump behavior. Because of this, it is considered unfeasible for implementation on a real system. Experimental results confirmed these findings, demonstrating the drawbacks of model-based fault detection and tolerance methods.
Library of Congress Subject Headings
Fault location (Engineering); Fault tolerance (Engineering); Kalman filtering; Cardiovascular instruments, Implanted--Evaluation; Blood--Circulation, Artificial; Biomedical engineering
Department, Program, or Center
Mechanical Engineering (KGCOE)
Gillespie, Erin, "Feasibility assessment of a Kalman filter approach to fault detection and fault-tolerance in a highly unstable system: The RIT heart pump" (2009). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus