This paper performs a numerical and model test on a jacket-type platform, with two neural network models predicting wave height and wave direction from vibration acceleration data. The data set used for training were collected by performing numerical simulation (finite element model) of the platform with various wave height and wave direction. Cross-correlation function was adopted to extract the feature of the vibration signal. Adjustment has been made to the activation function, which is helpful to the performance of the model. Finally, the reliability of the proposed method is verified by model test results with the same platform.


The recent decades have seen the rapid development of the fixed offshore structure industry, of which the jacket-type platform is the most commonly used type of structure for various purposes of resource exploration (Wisch, 1998). Having reached a relatively mature level, the major theoretical issue dominating the field is optimizing production activities from structural health monitoring (SHM) perspective. SHM refers to an anomaly detection procedure for engineering infrastructure. This procedure typically involves periodic observation of the mechanical system's dynamic response measurement, damage-sensitive feature extraction, and performing statistical analysis to determine the current health status (Doebling et al., 1996). Vibration-based structural health monitoring is on the rationale that anomaly behavior could lead to structural stiffness change and, consequently, changes its dynamic properties. Such technology will allow the current time-based maintenance approaches to be evolved into condition-based maintenance philosophies (Farrar, C. R., and Worden, K., 2012).

Since the early 1970s, considerable efforts have been made in developing vibration-based SHM technologies for offshore structures. However, the evidence for anomaly identifying has been weakened by varying environmental conditions since the developed indicator is not only sensitive to damage but also potential sensitive to the natural environmental variables (E.J. Cross, 2012).

For the jacket-type platform, the most dominant factor will be the random wave load where the structure is exposed. The varying wave height and wave loading direction can exert a confounding influence on the measured structural responses and mask the extracting feature indicative of the damage. There are potential options already explored for counteracting the influence of damage-unrelated variables. For example, Alampalli (2000) models an indicator concerning the environmental factor. Li (2019) adopted the principle analysis to eliminate the wave-induced variation in structure response. These methods show more or less reliance on the acquisition of the information, which necessitates a need to identify the environmental information accurately and readily. However, obtaining this information is still challenging by limited facilities or the SHM system's adaptability with state-of-art field sensing technologies. This situation heightens a need to explore reliable and effective methods to identify environmental factors with measured data. Up to now, most studies in this field have only focused on establishing the correlation between temperature and measured data. Yet, few works have investigated the wave condition detection method for the jacket-type platform.

This content is only available via PDF.
You can access this article if you purchase or spend a download.