Machine learning approach to predict flow fields induced by internal solitary waves acting on mid-water structures based on particle image velocimetry experiments

Abstract

Mid-water structures (MWSs) have important applications in the ocean to minimize the effects of environmental factors in the surface ocean and to avoid the effects of high pressure on the seabed. However, their operational positions are vulnerable to internal solitary waves (ISWs), which poses a significant challenge to the safe functioning of MWSs. Consequently, there is a pressing need to explore the flow field characteristics surrounding MWSs under the action of ISWs to establish a reliable foundation for structural design. In this study, experimental investigations were conducted to analyze the flow fields induced by ISWs acting on MWSs, employing the particle image velocimetry (PIV) method. The intricate details of the flow fields around the structure were thoroughly analyzed, and the impact of the structure's submerged depth on the ISW-induced flow fields was discussed. A machine learning approach is utilized to develop an Artificial Neural Network (ANN) model aimed at predicting the flow fields of ISWs. The experimental results are compared and analyzed alongside the predicted outcomes, focusing on the anomaly region, the tail field region, and the occluded region. The results show that the presence of the structure complicates the flow characteristics in the tail flow regions, and the flow patterns are constantly changing. The characteristics of the ISW flow fields around the structure vary with the different submerged depths of the MWS. Positioning the MWS above the wave trough results in an ascending trend followed by a descending trend in the strength of the surrounding flow field with increasing submerged depth. The ANN predictions for different types of regions of the flow fields show satisfactory accuracy. The model closely matches the PIV results in predicting the extremes and positions of horizontal velocities in the tail flow fields at different stages of ISW trough propagation. The ANN model can be recommended for predicting the flow fields of ISWs acting on MWSs. This study offers comprehensive insights into the characteristics of the flow fields induced by ISWs acting on MWSs, emphasizing the valuable contribution of machine learning methods to the field of fluid dynamics.