Akademik Veri Yönetim Sistemi
STRUCTURES, cilt.53, ss.806-821, 2023 (SCI-Expanded)
This study proposes an improved version of teaching learning-based optimization (TLBO) with a modified dynamic-opposition learning (MDOL) strategy, called modified dynamic oppositional learning TLBO (MDOLTLBO) to narrow the gap between research and practice for generalized time-cost optimization. Instead of opposite numbers, this study employs modified dynamic opposite points for the population initialization and generation jumping. This novel feature of the proposed algorithm is intended to diminish the randomness of the initial population and minimize the search effort required to achieve the optimal set of time-cost alternatives in the search space. An iterative-based varying weighting factor for MDOL is introduced to adjust the range of the search space with the aim of balancing the dynamicity and diversity of the algorithm. Hence, the MDOLTLBO algorithm can identify diversified and high-quality solutions based on non-dominated solutions. Four different case studies of real construction projects, ranging from 29 to 290 activities, were investigated to demonstrate the applicability of the proposed model. The experimental findings verify that the proposed algorithm is efficient, effective, and highly satisfying. Hence, the primary contribution of this study is MDOL approach that can successfully achieve high-quality solutions and good convergence speed for large-scale discrete time-cost optimization problems with generalized logical relationships.