Physics-Inspired Deep Learning and Transferable Models for Bridge Scour Prediction

TL;DR

Comparing SPINNs with traditional empirical models indicates substantial improvements in scour prediction accuracy, and can pave the way for further exploration of physics-inspired machine learning methods for scour prediction.

Abstract

This paper introduces scour physics-inspired neural networks (SPINNs), a hybrid physics-data-driven framework for bridge scour prediction using deep learning. SPINNs integrate physics-based, empirical equations into deep neural networks and are trained using site-specific historical scour monitoring data. Long-short Term Memory Network (LSTM) and Convolutional Neural Network (CNN) are considered as the base deep learning (DL) models. We also explore transferable/general models, trained by aggregating datasets from a cluster of bridges, versus the site/bridge-specific models. Despite variation in performance, SPINNs outperformed pure data-driven models in the majority of cases. In some bridge cases, SPINN reduced forecasting errors by up to 50 percent. The pure data-driven models showed better transferability compared to hybrid models. The transferable DL models particularly proved effective for bridges with limited data. In addition, the calibrated time-dependent empirical equations derived from SPINNs showed great potential for maximum scour depth estimation, providing more accurate predictions compared to commonly used HEC-18 model. Comparing SPINNs with traditional empirical models indicates substantial improvements in scour prediction accuracy. This study can pave the way for further exploration of physics-inspired machine learning methods for scour prediction.

Figures (17)

• Figure 1: SPINN framework, illustrating the integration of physics-based empirical equations as additional loss terms in the neural network training process.
• Figure 2: a) Profile of pier subjected to scour, showing sensors and variables [adapted from Hec-18:2012], b) Calculation of channel cross-sectional area and the relationship between velocity and discharge.
• Figure 3: Temporal features of live-bed local scour: a) Different phases of scour evolution, b) The ideal fitting line representing the upper bound of scour depth. Source: ngo2018guide.
• Figure 4: The training process of the time-dependent SPINN model, which incorporates scouring episode detection and applies the physical loss term only during scouring episodes.
• Figure 5: The architecture of LSTM base models for SPINNs, showing the LSTM Memory Unit gates and the input ($x_t$), hidden ($h_t$), and cell state ($c_t$) vectors.