ABSTRACT
Determining the optimal maintenance strategies, as for when and how to perform maintenance for deteriorating reinforced concrete buildings, has attracted a significant amount of attention in the past decades. Previous work has built a decision support system (DSS) to examine the trade-off relationship among maintenance cost, structural safety and usability, considering various sources of uncertainty in the analysis of seismic hazard and deterioration of structures. The existing DSS uses Multiobjective Particle Swarm Optimization (MOPSO) to search for the optimal maintenance strategies, whose performance is evaluated using Monte Carlo simulation. To alleviate the heavy computation load required by the existing DSS, the present study develops parallel computing models: Master-slave, Island, and Ring. The models are implemented and tested in a computer cluster, composed of 40 processor cores. A four-story school building is used as the test case. In this article, we compare the speedup (scalability) of the parallel computing models and discuss how the increased computational efficiency would lead to much better solution quality and therefore improves the decision support system.
