The present study addresses the problem of estimating time-varying time constants associated with thermocouple sensors by a set of basis functions. By expanding each time-varying time constant onto a finite set of basis sequences, the time-varying identification problem reduces to a parameter estimation problem of a time-invariant system. The proposed algorithm, to be called as orthogonal least-squares with basis function expansion algorithm, combines the orthogonal least-squares algorithm with an error reduction ratio test to include significant basis functions into the model, which results in a parsimonious model structure. The performance of the method was compared with a linear Kalman filter. Simulations on engine data have demonstrated that the proposed method performs satisfactorily and is better than the Kalman filter. The new technique has been applied in a Stirling cycle compressor. The sinusoidal variations in time constant are tracked properly using the new technique, but the linear Kalman filter fails to do so. Both model validation and thermodynamic laws confirm that the new technique gives unbiased estimates and that the assumed thermocouple model is adequate.
A Novel Adaptive Fading Kalman Filter-Based Approach to Time-Varying Brain Spectral/Connectivity Analyses of Event-Related EEG Signals
