Abstract:Considering rocket double-redundant strapdown control system consisting of six twodimension-of-freedoms gyros, advantages of integrated hardware redundancy and analytical redundancy, an advanced method for gyro fault diagnosis is designed; by using neural network observer, not only the problem which the fault diagnosis method based on hardware redundancy could not solve is able to be solved, but also the weakness of traditional analytical redundancy method could be overcome; by gyros’ adaptive thresholds, the robustness of fault diagnosis is improved; and it achieves the recovery of a fault gyros’ data.