In this paper, the damage identification of offshore floating platform mooring system is investigated. Firstly, based on the change of the axial stiffness of the mooring line, different damage severities of the mooring system are simulated and the static analysis of the platform and mooring system are calculated under a series of sea states. The Radial Basis Function (RBF) neural network is applied for damage identification to deal with the complex behavior of floater and mooring system. The numerical results show that RBF neural network has a good performance on damage identification of mooring lines.
Mooring line is the key component of offshore floating system as providing the station keeping function. During the service life, damages of the mooring system are unavoidable as a result of the action of various loads including operational and environmental forces. The structural health monitoring (SHM) system is very necessary to ensure the safety of the structures, lower the maintenance cost and prolong the service lives. A SHM system is defined as the process of implementing a damage detection strategy for engineering infrastructure related to aerospace, civil and mechanical engineering (Farrar and Sohn, 2000).
For damage detection of offshore structures, Mangal et al. (2001) studied the law of natural frequency and its time-domain response changing with different member damage through numerical simulation calculation and physical experiment analysis of jacket platform. Ho and Ewin (2000) compared the effectiveness of the damage diagnosis method using different mode information by numerical calculation of a group of finite element models. However, because of the complexity of the interaction between ocean environment and floating system, the mooring lines may have nonlinear characteristics in the dynamic responses, which increases the difficulty of damage identification using the traditional methods.
The emergence of artificial neural networks has greatly improved this situation. The neural network has a good nonlinear mapping ability, and converts the inverse problems such as damage identification and positioning of the engineering structure into the positive problem. The earliest use of neural networks for structural damage identification was the Venkatasubramanian and Chan of Purdue University in the United States. In 1989, they first used neural networks for damage identification of large structures (Venkatasubramanian and Chan, 1989).
