A Neural Network Surrogate Model for Structural Health Monitoring of Miter Gates in Navigation Locks

Abstract

Structural health monitoring (SHM) of miter gates of navigation locks is crucial for facilitating cargo ship navigation. Closure of these inland waterway structures causes considerable economical loss to the marine cargo and associated industries. In practice, strain gauges are often mounted in many of these miter gates for data collection, and various inverse finite element techniques are used to convert the strain gauges data to damage-sensitive features. Arguably, these models are computationally expensive and sometimes they are not suitable for real-time health monitoring or for monitoring confounding environmental effects. In this work, a Multi-Layer Artificial Neural Network (MANN) is designed to serve as a “run time” surrogate model that links data (from the strain gages) to damage classification (gaps in the miter gate contact). Three cases of complexity, combining hydrostatic and thermal loading scenarios with varying gap scenarios, are considered to design the MANN. A confusion matrix is used to evaluate the performance of the networks and derive probabilities. Results show the potential of MANNs as a reliable surrogate model for computationally expensive inverse finite element modeling in damage classification for this application.