Machine Learning-Based Inverse Modeling of Manufacturing Processes

In modern manufacturing, identifying the optimal input process parameters to achieve desired output characteristics is crucial for cost reduction and waste minimization. While artificial neural networks cannot fully solve the inverse modeling problem, this study proposes using machine learning techniques that combine neural networks with evolutionary algorithm-based models, such as particle swarm optimization, to address this challenge. This thesis presents eight publications, including one published in a high-impact journal - IEEE Transactions on Industrial Informatics, resulting from this research, that aim to tackle and resolve the complex inverse modeling problem in manufacturing using two manufacturing processes as case studies.