Open Collections

Esmaeilidouki, Aida

University of British Columbia

Production Planning and Control (PPC) is defined as a predetermined process to incorporate human resources, raw material, machines, and other system-level constituents into a manufacturing process. Nowadays, while many companies tend to utilize efficient tools in their manufacturing systems to succeed in the competitive markets, potential to choose the optimal PPC methods continues to be a challenging task due to the presence of multiple concurrent criteria such as social, economic, technical, and environmental. The present thesis introduces a hybrid framework of different modeling methods to investigate PPC scenarios in manufacturing systems. Namely, the approach, shown through a case study, provides insight into the integration of simulation tools, Machine Learning (ML) and various Multi-Criteria Decision Making (MCDM) techniques in PPC. At the first step, a factory-level simulation model in AnyLogic is employed to examine the time and cost of the entire manufacturing process. Additionally, Environmental Impacts (EnvI) imposed on the system is determined via SimaPro, a Life Cycle Assessment (LCA) tool. At the second step, the data generated by AnyLogic and SimaPro are used by Random Forest (RF), M5', Gaussian Process (GP), Group Method of Data Handling (GMDH), and Multi-Layer Perceptron (MLP) as different ML models for training under different PPC scenarios. Finally, using the ML models output, two MCDM methods (built upon regret and rejoicing as well as an integrated version of the Best Worth Method (BWM) and the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)) are used to rank the optimal PPC scenarios. The case study results indicate that the proposed framework could be highly practical for selecting an optimal PPC scenario in manufacturing systems under highly complex data structures and conflicting criteria, while lowering the cost of data trials using ML models.

Text

2021-08-31

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/79552

Mechanical Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/