A negative selection algorithm applied to helicopter actuator fault detection

In order to detect the failure of an actuator of a helicopter swashplate, a simple detection method based on a negative selection algorithm (NSA) is developed. This technique, inspired by existing rules found in the biological immunity, rely on a set of elemental detectors which have been trained in order to recognize only the non-self (i.e. faulty) behaviors of the system, without needing an explicit model of the system dynamics. A simple model for the actuator failure is developed and coupled with a helicopter flight dynamics model, and some trajectories for the algorithm training and validation are simulated. A procedure for improving the efficiency of the algorithm by introducing weighting coefficients to the detection function is proposed and successfully tested. Moreover, the performance of the algorithm is slightly improved by observing the covariance of the flight dynamics states, instead of the variables themselves.