Job shop estocástico con minimización del valor esperado del maximum lateness

The drawbacks that programming in job -shop environment imply, refer to a notorious difficulty for its resolution due to its NP-hard nature. However, the research has grown in the late years because of its constant use in manufacturing industries. According to studies, most of the research has approached the job shop scheduling through a deterministic approach. Nevertheless, real industrial environments are subject to random events as: machinery faults, maintenance duration, processing duration, enlistment times, availability times, among many others. In this project, a stochastic job shop that minimizes the expected maximum lateness is addressed. The problem consider sequence dependent setup times, and the stochastic events are machine breakdowns. To solve the problem a simheuristic approach is proposed. The simheuristic Hybridizes a tabu search algorithm with a Monte Carlo simulation. The problem was solved in three phases: Firstly, a mixed integer linear programming model was designed for the deterministic counterpart of the JSSP studied. Secondly, the meta-heuristic tabu search was designed to solving large instances of the deterministic problem. Thirdly, the simheuristic was designed and implemented to minimize the expected maximum lateness value, considering stochastic machine breakdowns. For the simheuristic designing, stochastic variables were generated: times between failures and repair times, following exponential and log-normal distributions. To generate their respective parameters [expected value (μ) and standard deviation (σ)], the mean time to repair was found (MTTR Mean Time to Repair), out of the total mean time between breakdowns. Four different variation coefficient values were proposed (0%, 5%, 10% and 15%), them being: 0% for the deterministic case and 5%, 10% and 15% for stochastic events, to calculate the (σ) in log-normal distribution. On the other hand, a simulation was performed to calculate the expected objective function. The simheuristic was firstly parametrized through an experimental design considering different tabu list sizes and number of iterations without improvement. With the generated parametrization, another computational experiment was executed for a total of 554 instances of different sizes. First, the performance of the simheuristic, for small instances, was evaluated in comparison with the simulation of optimal solutions obtained with the mathematical model. Results show that the simheuristic improves the results of simulations of the model in a 37% for 4x4 instances and in an 11% for 6x6 instances, demonstrating that the simheuristic is better than a deterministic mathematical model simulated. Additionally, the simheuristic performance was evaluated, for large instances, in comparison with the simulation of EDD dispatching rule sequences. Results show that the average improvement is 28% in log-normal distribution and 10% for exponential distribution.

• Autores: Forero Ortiz, Gabriel Fernando; Ocampo Monsalve, María José; Rivera Torres, Andrea
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2020
• Institución: Pontificia Universidad Javeriana
• Repositorio: Repositorio Universidad Javeriana
• Idioma: spa
• OAI Identifier: oai:repository.javeriana.edu.co:10554/53042
• Acceso en línea: http://hdl.handle.net/10554/53042
• Palabra clave: Tienda jop estocástico; Tardanza máxima; Averías; Sim heurístico; Búsqueda tabú; Stochastic jop shop; Maximum lateness; Breakdowns; Simheuristic; Tabu search; Ingeniería industrial - Tesis y disertaciones académicas; Análisis estocástico; Algoritmos de aproximación
• Rights: openAccess
• License: Atribución-NoComercial-SinDerivadas 4.0 Internacional