Machine-learning surrogate modeling approaches used in electromechanical systems and electrical components

Abstract

This bachelor’s thesis focuses on machine learning surrogate modeling approach used in electromechanical systems and electrical components. Modern electromechanical systems essentially include mechanical, electronic, and computational technologies which could create complex tasks systems. These complexity systems need precise coordination in components and data processing, especially in unknown conditions. As a result, surrogate modeling is particularly important in these conditions. These model systems with data-based can simulate and predict the behavior of electromechanical systems in advance, which is important in system design, control strategy formulation and performance optimization. This thesis will focus on surrogate modeling methods of finite element models with neural networks. The thesis will also explore how to use data learning to optimize surrogate models, which included all aspects of data collection, processing, and model training. In addition, it also discusses the advantages and limitations of machine learning surrogate modeling approach and will present its application in electromechanical systems.

Finally, the thesis discusses the challenges of current methods and possible directions for future development. Surrogate models reduce the system's dependence on physical experiments and complex mathematical modeling, which is an important method for design and analysis of electromechanical systems and electrical components. It will make a important position in intelligence and automation in the future.