Sequential space conversion method with multi-armed bandit algorithm for time-variant reliability and dynamic project management

Time-variant reliability analysis plays a crucial role in assessing uncertainties in engineering structures throughout their lifespan. Given the inherent time uncertainties and their associated cost correlations during the construction phase, project management can be formulated as a special case of time-variant reliability analysis, where these uncertainties are rigorously modeled. Here, when the traditional approaches such as the critical path method provide a deterministic solution, the Project Evaluation and Review Technique (PERT) and Monte-Carlo method consider activity durations as uncertainty and solve the problem using the central limit theorem. However, these methods overlook cost uncertainty, often leading to overly optimistic evaluations. In this study, a Multi-Armed Bandit (MAB) approach is combined with the Sequential Space Conversion method to efficiently handle uncertainty quantification of time-variant systems. Then, a time-cost correlation approach is suggested for the dynamic modeling of the planning problem, providing a comprehensive simulation framework for project management. The results demonstrate that the proposed method achieves a superior balance between accuracy and computational efficiency in time-variant reliability assessments. Moreover, the time-cost correlation approach provides more realistic projections, revealing discrepancies of up to 90 % compared to models that consider only time uncertainty in project evaluation.