A new method to measure damage progression in Homogeneous Low-Crested Structures with a low-cost 3D depth-sensor

Abstract

Mechanical profilers are commonly used to measure damage to rubble mound breakwaters in small-scale physical tests. This study developed a non-intrusive methodology using low-cost 3D depth-sensors to measure damage progression in Homogeneous Low-Crested Structures (HLCS) with and without water in the wave flume. Although light refraction causes distortion in the presence of water, distorted scans are corrected with this new method using a Neural Network (NN) model. The new methodology is adequate to obtain undistorted profiles of emerged or submerged breakwaters and may be described in two steps: (1) a digital profiler algorithm, to describe the breakwater models in series of profiles in order to measure damage, and (2) a trained NN model, to correct the distortions caused by light refraction for the scans taken with water in the wave flume. The NN model to correct profiles requires two sets of scans, one in empty conditions and the other in submerged conditions; the two scans (dry-wet) can be easily obtained without disturbing the usual test programs. This is done by taking the scans at the beginning (dry and wet) and at the end (wet and dry) of each test series. By comparing profiles, the breakwater damage can be analysed after obtaining the series of undistorted profiles from the trained NN model and the distorted scans. To validate the method, a series of small-scale physical tests with Cubipod HLCS were carried out from no damage to severe damage, and then the estimations given by the trained NN models were compared with blind observations, taken from additional pairs of dry and wet scans. The good agreement between the observations and the estimations (R² > 0.985, r > 0.993) showed that low-cost 3D depth-sensors may be used as non-intrusive methods for breakwater profiling, even when scanning is done with water in the wave flume or basin.