Proactive rehabilitation planning for concrete road networks considering multiple hazards and multiple assets: A Kriging-based multi-objective efficient global optimization approach
Enhancing the resilience of concrete road networks (CRNs) against multiple hazards is critical, as networks are increasingly threatened by independent hazards such as earthquakes and floods. Existing rehabilitation optimization methods lack frameworks that simultaneously address conflicting resilience objectives across hazards and rely on repeated, computationally expensive traffic simulations. This study develops a surrogate-assisted multi-objective optimization framework to minimize Average Disruption Cost (ADC) from earthquakes and floods through optimized rehabilitation policies. The framework integrates a Kriging-based surrogate model with Multi-Objective Efficient Global Optimization (MO-EGO). Critical assets are identified, and rehabilitation policies are generated to jointly reduce disruption costs under multiple hazards. Physical damage assessments and dynamic traffic simulations are used to quantify disruption costs for rehabilitation policies selected using Latin Hypercube Sampling (LHS). The resulting data are then used to develop Kriging surrogate models that approximate computationally expensive simulations during optimization. The framework is applied to the road network in Arden-Arcade, Sacramento, California. The resulting Pareto front demonstrates clear trade-offs between seismic and flood resilience, with different assets identified as critical for each hazard. A balanced rehabilitation policy is identified that effectively reduces disruption costs from both hazards, demonstrating the advantages of multi-hazard optimization over single-hazard approaches. Compared with a traditional simulation-based Multi-Objective Genetic Algorithm (MOGA), the Kriging-based MO-EGO framework reduced computational time by approximately 91.6%, although solution quality was moderately reduced due to limitations in the earthquake surrogate model. The proposed framework enables efficient, cost-effective multi-hazard rehabilitation planning for large-scale CRNs.
Authors
- Sumaya Sharveen (ORCID: https://orcid.org/0009-0006-9624-5839)
- Mohsen Shahandashti (ORCID: https://orcid.org/0000-0002-2373-7596)
- Sushil Bhatta (ORCID: https://orcid.org/0009-0002-0459-5817)
- Abhijit Roy (ORCID: https://orcid.org/0009-0002-9603-8144)
- Sandesh Adhikari (ORCID: https://orcid.org/0009-0004-1962-3061)
Institutions
- The University of Texas at Arlington (US)
Publication Details
- Journal
- Resilient Cities and Structures
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1016/j.rcns.2026.10.001
- Primary Topic
- Infrastructure Resilience and Vulnerability Analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- U.S. Department of Transportation