Fleet optimization considering overcapacity and load sharing restrictions using genetic algorithms and ant colony optimization. Issue 1 (13th February 2020)

• Record Type: Journal Article
• Authors: Kristjanpoller, Fredy; Michell, Kevin; Kristjanpoller, Werner; Crespo, Adolfo
• Is Part Of: AI EDAM. Volume 34:Issue 1(2020)
• Page Start: 104
• Page End: 113
• Publication Date: 2020-02-13
• Subjects: Ant colony optimization, -- availability analysis, -- fleet design, -- life cycle cost, -- overcapacity; Engineering design -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; 620.00420285
• DOI: 10.1017/S0890060419000428
• Languages: English
• ISSNs: 0890-0604

Abstract

Abstract: This paper presents a fleet model explained through a complex configuration of load sharing that considers overcapacity and is based on a life cycle cost (LCC) approach for cost-related decision-making. By analyzing the variables needed to optimize the fleet size, which must be evaluated in combination with the event space method (ESM), the solution to this problem would normally require high computing performance and long computing times. Considering this, the combined use of an integer genetic algorithm (GA) and the ant colony optimization (ACO) method was proposed in order to determine the optimal solution. In order to analyze and highlight the added value of this proposal, several empirical simulations were performed. The results showed the potential strengths of the proposal related to its flexibility and capacity in solving large problems with a near optimal solution for large fleet size and potential real-world applications. Even larger problems can be solved this way than by using the complete enumeration approach and a non-family fleet approach. Thus, this allows for a more real solution to fleet design that also considers overcapacity, availability, and an LCC approach. The simulations showed that the model can be solved in much less time compared with the base model and allows for the resolution of a fleet of at least 64 trucks using GA and 130 using ACO, respectively. Thus, the proposed framework can solve real-world problems, such as the fleet design of mining companies, by offering a more realistic approach.