Hadi Karimi, Abbas Seifi,
Volume 23, Issue 4 (11-2012)
Abstract
The analytic center cutting plane method (ACCPM) is one of successful methods to solve nondifferentiable optimization problems. In this paper ACCPM is used for the first time in the vehicle routing problem with time windows (VRPTW) to accelerate lagrangian relaxation procedure for the problem. At first the basic cutting plane algorithm and its relationship with column generation method is clarified then the new method based on ACCPM is proposed as a stabilization technique of column generation (lagrangian relaxation). Both approaches are tested on a benchmark instance to demonstrate the advantages of proposed method in terms of computational time and lower bounds quality.
Dr. Zahra Esfandiari, Prof. Mahdi Bashiri, Prof. Reza Tavakkoli-Moghaddam,
Volume 31, Issue 1 (3-2020)
Abstract
One of the major risks that can affect supply chain design and management is the risk of facility disruption due to natural hazards, economic crises, terrorist attacks, etc. Static resiliency of the network is one of the features that is considered when designing networks to manage disruptions, which increases the network reliability. This feature refers to the ability of the network to maintain its operation and connection in the lack of some members of the chain. Facility hardening is one of the strategies used for this purpose. In this paper, different reliable capacitated fixed-charge location allocation models are developed for hedging network from failure. In these proposed models, hardening, resilience, and hardening and resilience abilities are considered respectively. These problems are formulated as a nonlinear programming models and their equivalent linear form are presented. The sensitivity analysis confirms that the proposed models construct more effective and reliable network comparing to the previous networks. A Lagrangian decomposition algorithm (LDA) is developed to solve the linear models. Computational results show that the LDA is efficient in computational time and quality of generated solutions for instances with different sizes. Moreover, the superiority of the proposed model is confirmed comparing to the classical model.
