Establishment of a hybrid Fuzzy–Krill Herd approach for novelty detection applied to damage classification of offshore jacket-type structures.

Offshore jacket platforms have been widely developed and their damage detection is of great importance to ensure their safe operation. Structural health monitoring method (SHM) has been used to decline the effect of the damages caused by marine environmental conditions. Global SHM methods are considered as efficient methods which include both economic and safety issues, and use modal characters of the structure. Major parts of offshore structures which are not available for diagnosing change the monitoring process to a big challenge. Different procedures have been used for diagnosing process, where simulating the structure as a finite element (FE) model is the most common one. Comparison between defected structures FE model and the intact model of the structure is the fundamental concept of SHM diagnostic. Mathematical models are used to develop limited measured data for health monitoring. An important step in SHM process includes FE model updating which is applicable using similar algorithms with diagnosing process. Some algorithms, such as neural network, have been used to detect the damage patterns, where their developing complication is their most important challenge. In this study, a new Fuzzy–Krill Herd algorithm is used to perform structural health monitoring process in a fixed jacket platform and compared with common fuzzy–genetic method. For this purpose, an experimental fixed platform model (SPD9 located in Persian Gulf) is prepared to obtain modal parameters using vibration behavior. Structural FE model is built and numerical model updating is performed on the results of the physical model. Moreover, the effects of uncertainty in damages and health monitoring in experimental and numerical models are considered for detecting damages and monitoring the structure. Using a combination of smart algorithms and fuzzy logic, a complicated and efficient system for structural health monitoring process is introduced. The suggested Fuzzy–Krill Herd method is shown to be applicable and appropriate in SHM process and also for handling noisy data. [ABSTRACT FROM AUTHOR]