Multi-objective dynamic hybrid flow shop scheduling using machine learning approach

Abstract

Keeping pace with the rapidly changing production system is a challenge, where each day new technology is being invented and the market is becoming more competitive. One effective way to sustain is to have a production planning system that can react to sudden changes in the production phase and also is capable of finding an optimum solution among production challenges. This thesis aims to propose a semi-automated dynamic hybrid flow shop scheduling model that can provide an optimum production schedule considering capacity limitations, operators learning effect, machine break-down conditions, etc. In order to make the production scheduling semi-automated, a machine learning algorithm, Support Vector Machine (SVM) is used to formulate a job classification model that can classify jobs based on their priority level. Furthermore, a scheduling model is developed that utilizes each job’s corresponding priority level information. The model aims to address three objectives: minimization of make-span, minimization of tardiness and maximization of efficiency. In this work maximization of efficiency is calculated in terms of machine idle time. To make this model applicable to real-life production challenges, uncertainties related to processing time and machine break-down are considered. Finally, a meta-heuristic algorithm, Particle Swarm Optimization is used to find the optimum schedule. The job classification approach used in this thesis has not been explored by other researchers so far. This proposed method has proved its efficacy in depicting real-life production challenges and providing an optimum result.