Damage detection for offshore structures using long and short-term memory networks and random decrement technique

Abstract

A damage detection method is presented which combines the random decrement technique (RDT) with long and short-term memory (LSTM) networks. The method uses the measured vibration response of offshore structures subjected to random excitation and is able to locate and assess the damage with accuracy, even in noisy conditions. The applicability of the proposed RDT-LSTM method is verified through a numerical example and laboratory tests. The numerical example consists of a jacket platform subjected to random wave excitation. The simulated damage cases encompass single and multiple damage locations not only on whole segments but also on local elements (one-fifth of the whole segment) of the numerical structure, with minor (1%–5%) severity, and different noise levels. RDT is applied first to process the noisy random data, and then the damage detection is carried out using LSTM. After the numerical example, the proposed method is applied to laboratory tests of a jacket platform model under random loading produced by a shaking table. Minor and major damages and their combination at different locations are discussed. Both the numerical simulation and laboratory test show that the proposed RDT-LSTM method has an outstanding performance in structural damage detection.