Distributed Fusion Parameters Extraction for Integrated System Health Management to Space Avionics

Abstract

This paper addresses a parameter extraction problem for space avionics integrated system health management. A distributed fusion-based parameter extraction scheme is developed by fusing the preprocessed results of the space avionics multi-sensor module. Because of the complex environment in space avionics, the feature parameters of electronic circuit in avionics systems are often corrupted by temporally correlated noise. For these correlated noise parameter systems, a distributed fusion parameters extraction algorithm is presented, in which the traditional method of dealing with correlated noise by augmenting the systems state vector is avoided. The proposed distributed fusion parameters extraction algorithm guarantees optimality in the sense of being equivalent to the optimal centralized estimated fusion. Numerical examples are provided to demonstrate the performance of the new algorithm.Nomenclaturefilter gain at time k K i k = Kalman filter gain of ith sensor at time k L k = measurement prediction error covariance matrix at time k L i k = measurement prediction error covariance matrix of ith sensor at time k N k;τ = correlations of prediction error of θ k and θ k-τ using measurements up to k-τ − 1 N i k;τ = added recursive terms at ith sensor p, q = orders of autoregressive moving average model P kjk = estimation error covariance matrix of θ kjk P kjk−1 = prediction error covariance matrix of θ kjk−1 P i kjk = estimation error covariance matrix of θ i kjk P i kjk−1 = prediction error covariance matrix of ith sensor at time k q = correlated step of noises R k = correlation matrix of w k R i k = correlation matrix of measurement noise w i k S k-τ= added terms for expression clarity T k;τ = covariances of prediction error of θ k using measurements up to k-τ T i k;τ = added recursive terms at ith sensor V j;t = correlation matrix between v j and v t V t = covariance matrix of noises v t v k = process noise at time k W k = stacked measurement noises in fusion center at time k w i k = measurement noise of ith sensor at time k yt = system output y k = stacked measurements received from all sensors at time k y i k = measurement of ith sensor at time k Δy k = prediction error of measurement at time k δ t−j = Dirac delta function θ k = extracted parameter at time k θ kjk = optimal estimate at time k using all measurements up to time k θ kjk−1 = one-step prediction of extracted parameter θ t = extracted parameters θ i kjk = optimal estimate of ith sensor at time k θ i kjk−1 = one-step prediction of ith sensor at time k Φ k = state transformation matrix at time k