Nonlinear prediction surfaces for estimating the structural response of naval vessels

Abstract

Structural health monitoring (SHM) of naval vessels is essential for assessing the performance of the structure and the fatigue damage accrued over the service life. The direct integration of available SHM data may be useful in reducing the epistemic uncertainties arising from inaccuracies in the modeling and the variations in the as-built structural configuration from the initial design. Based on SHM data, fatigue damage indices can be predicted by implementing cell based approaches, such as the lifetime weighted sea method, that discretizes the operational conditions of the vessel into cells with specific wave height, heading angle, and speed. The integration of SHM data into the fatigue assessment using lifetime weighted sea method requires a complete set of data that covers the whole operational spectrum. However, technical malfunctions or discrete monitoring practices generate incomplete data sets. This paper proposes nonlinear prediction surfaces to estimate the ship structural response in unobserved cells based on available cell data. Expected theoretical variations of the structural response to changes in wave height, heading angle, and vessel speed are integrated in the development of the prediction surface. The proposed methodology is illustrated on the SHM data from a high speed aluminum catamaran.