PDE-Constrained Inverse Analysis Using Bayesian Optimization for Finding Hidden Corrosion Given Partial Surface Information

To ensure infrastructure resilience, regular assessment of infrastructure damage must be carried out. To assess hidden corrosion in reinforced concrete structure, inverse analysis can be employed. Inverse analysis combines partial field data and computational analysis, including machine learning. In this paper, Bayesian optimization was used for inverse analysis to find hidden corrosion in a reinforced concrete corrosion. A simple case study was presented where the hidden corrosion location was known. The partial data on the surface of the concrete was known and used for the inverse analysis. The Bayesian optimization was successfully used to find the location of the hidden corrosion. The Max Expected Improvement acquisition function performs better than Max Probability Improvement one.