Parameter Estimation by Nonlinear Smoothing for Fault Monitoring on Rocket Engines

Abstract

Abstract Nonlinear parameter estimation algorithms based upon the extended Kalman filter are designed for the problem of estimating some of the parameters that describe the dynamics of a reusable space propulsion system. In this paper, the results of applying these algorithms to simulated data and to test firing data from the Space Shuttle Main Engine (SSME) are presented. The parameters that are estimated are among those likely to change when some common engine degradations occur. The SSME operating at its 100% Rated Power Level is used as the baseline system. All estimator designs are based upon a reduced-order dynamic model of the SSME. The simulated data used to drive the algorithms is generated by a high fidelity transient simulation of the SSME with small magnitude random “dither” signals applied to the inputs and with substantial random noise added to the measured outputs. The results for a healthy SSME indicate good parameter estimation accuracy. However, the algorithms fail to track the correct parameter values when parameter changes representing engine degradations are introduced if the change results in a significant change in the values of dynamic variables that are not included in the reduced order model used to design the estimator. Preliminary results of using hot fire data from the SSME are also presented. These results indicate that the estimator produces valid results when it is driven with real data, but that it must incorporate logic to change the estimator design as the commanded power level of the engine changes.