Multiobjective Location Model Design Based on Government Subsidy in the Recycling of CDW

Abstract

With the generation of a large amount of construction and demolition waste (CDW), many scholars have recently paid more attention to the recycling of CDW. In this paper, we design a classification recycling method based on the degree of CDW availability in the recycling of CDW. Considering the important role of the government in reverse logistics, a model of reverse logistics network based on the trade-off between cost and recycling rate is put forward, which is subject to government subsidy. The model includes the location of classification processing center and ensures the best route of transportation. Then, the improved particle swarm optimization algorithm is applied to solve the model to get Pareto frontier by transforming it into a multiobjective integer-programming problem. As a case study, the results of the statistical modeling used in this study indicate the feasibility of the model. Finally, according to the sensitivity analysis of government's subsidy, we evaluate the effectiveness of this program and its applicability.

Improved Simulated Annealing Based Network Model for E-Recycling Reverse Logistics Decisions under Uncertainty

“…Metropolis et al first introduced SA in 1953 [46] and it was later popularized by Kirkpatrick et al [47]. SA has been applied to many hard combinatorial optimization problems, including the design of reverse logistics network and related closed-loop supply chain network [29,48]. The SA's neighborhood search method is a random fashion that lets every factor in the candidate solution have the same probability to be changed [49], which increases the probability of local optima and computing time [50].…”

“…All the nodes in RLN have the probability of increasing storage capacity, so the initial solution is obtained by increasing the capacity of all the nodes averagely, named as equal division method. Note that the increasing quantity is discrete, so the average may not be an integer, and then the integer less and most near to the average will be taken as the increasing capacity of the nodes and the remainder will be assigned randomly [29].…”

“…As a consequence, in the proposed solution approach, the improved SA algorithm is applied to swiftly search the feasible domain of the variable , denoted by in the pseudocode of the improved SA in Algorithm 1. This is an extension of the pseudocode presented in [29], which is related to the standard SA. In Algorithm 1, ( ) represents the process of calling the algorithm of RLN simulation model, whose value is under the specific ; = { } is the set of quantitative importance of each node in RLN;…”

“…represents the total number of iterations that the perturbation (neighborhood search) should repeat at a particular temperature; 0 and are the initial and final temperature, respectively; is the Boltzmann constant used in computing the probability of accepting a worse solution; and is the coefficient of the cooling schedule [29].…”

“…The optimization part selects a suitable solution based on the output of the evaluation part and generates a new candidate solution. The optimization part is usually according to the solution techniques, which include the branch-and-cut [24,25], branches-and-bound [25], and Benders' decomposition [14,26] in the standard mathematical optimization techniques of MILP and simulated annealing (SA) algorithm [20,27], genetic algorithm [18,28], and swarm optimization algorithm [29] from the heuristic methods. As the RLN design problem belongs to a class of combinatorial optimization problems that require computational resources at an exponentially growing rate when the number of decision variables increases, hence, heuristic methods are usually adopted when the number of decision variables is large [29][30][31].…”