| Summary: | Hull deformation is an important factor that affects the attitude accuracy of shipborne weaponry and equipment, and hull deformation should be estimated and compensated for a large ship to establish a high-precision attitude reference. The traditional inertial measurement matching methods for hull deformation estimation largely depend on prior deformation information. However, the real-time deformation data characteristics are commonly different from the characteristics of historical deformation data, and a measurement matching method based on an inaccurate deformation model will result in a larger deformation estimation error. To overcome this difficulty, a hull deformation measurement method based on attitude quaternion matching and an online neural network is proposed. First, a deformation measurement equation based on quaternion matching is established with the angular velocity information measured by inertial sensors. Then, an online neural network is used to fit the deformation angle in the measurement equation to avoid establishing the deformation model. Considering the real-time training of an online neural network, the connection weight coefficients are estimated using a nonlinear filter. The simulation results demonstrate that the proposed method can accurately estimate the deformation without prior information of the deformation.
|
|---|
