Aplicação de algoritmos de controle e balanceamento de carga a um sistema perinatal

• Autor(a) principal: Oliveira, Ricardo Bruno Osés de
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFG
• dARK ID: ark:/38995/0013000005tjs
• Texto Completo: http://repositorio.bc.ufg.br/tede/handle/tede/11363
• topic: Balanceamento de carga; Markov; Perinatal; Maternidade; Comportamento das abelhas; Load balancing; Markov; Perinatal; Maternity; Honey bee behavior; ENGENHARIAS::ENGENHARIA ELETRICA

In this work, the process of admission and scheduling of pregnant women in a perinatal network composed of two maternities is considered. The main service sectors that make up each perinatal unit and their respective functions and resources used are described. In addition, a survey is made of the main challenges and difficulties faced by these health units in Brazil in recent decades. Thus, given the problems and challenges pointed out, different load balancing algorithms are used in the proposed perinatal network, to find the best task scheduling policy in the system that increases the efficiency of the network. A solution for a mixed integer linear programming problem is proposed using a load balancing algorithm based on metaheuristics of the behavior of honey bees. Besides, some algorithms are analyzed where each considered model has its own task routing strategy designed to reduce the average time of the pregnant women entering the perinatal system, balancing the workload among the perinatal care centers. Two classes of routing are used, non-deterministic and deterministic. In the deterministic class, three routing policies are analyzed that seek to decrease the average stay time, the average service time, or to improve the flow in the system. In addition, a dynamic control policy based on a queuing threshold is also analyzed, where a specific queue length is defined and identified by a threshold. The routing policy named Join-the-Shortest-Queue (JSQ) is also analyzed, where each pregnant woman who enters the system is directed to the maternity ward with the shortest queue. The results are presented and analyzed varying both the arrival rates of pregnant women and the rates of care in the main sectors in a maternity hospital. Also, a discrete event simulation model is made to analyze the waiting time in queues. Finally, using the formula of Erlang-B, the capacity of the perinatal units is calculated based on the time of permanence of the pregnant women in the system obtained through the load balancing algorithms. The results obtained confirm that routing and scheduling policies that consider the task arrival rate and the system queue length are more efficient as the arrival rate increases, therefore being applicable in healthcare systems with increasing demand and that with planning it is possible to obtain an accurate description of the number of occupied beds and the number of beds needed according to the demand required by the perinatal units making these units more efficient.