Files
Abstract
Generating diverse and adequate datasets for machine learning and data analysis in manufacturing is a daunting task. Synthetic data generation offers a practical solution by enabling the creation of expansive and varied datasets.
This thesis reviews the current literature on synthetic data generation for assembly lines, and presents a step-by-step guide for generating synthetic data suitable for machine learning and data analysis in manufacturing. The study illustrates a framework for utilizing synthetic data generation techniques using a case study of binary data, comparing the performance of models trained on real data versus hybrid data. Results show that synthetic data generation significantly enhances model performance, particularly when real data is limited or biased.
Overall, this work demonstrates the potential of synthetic data generation techniques for improving the availability and diversity of manufacturing data for model development and analysis. The presented framework serves as a guideline for researchers and practitioners interested in applying synthetic data to manufacturing applications.
This thesis reviews the current literature on synthetic data generation for assembly lines, and presents a step-by-step guide for generating synthetic data suitable for machine learning and data analysis in manufacturing. The study illustrates a framework for utilizing synthetic data generation techniques using a case study of binary data, comparing the performance of models trained on real data versus hybrid data. Results show that synthetic data generation significantly enhances model performance, particularly when real data is limited or biased.
Overall, this work demonstrates the potential of synthetic data generation techniques for improving the availability and diversity of manufacturing data for model development and analysis. The presented framework serves as a guideline for researchers and practitioners interested in applying synthetic data to manufacturing applications.