Year: 2023
Type: dissertation
Abstract: Electrical motors are essential components of many types of mechanical systems and are used in a wide range of applications, from appliances to industrial equipment.The utilization of electric motors and the associated systems is projected to account for approximately 43-46% of worldwide energy usage.Designing electrical machines, such as generators and motors, is a complex task that involves balancing a number of competing objectives, such as efficiency, cost, and reliability.Traditional optimization methods, such as gradient-based optimization and evolutionary algorithms, have been widely used to design electrical motors.However, these methods can be computationally expensive and may struggle to find the global optimum solution in complex design spaces.Reinforcement learning (RL) is a rapidly growing field in machine learning, which has been proven to be an effective tool for optimizing complex systems.The approach of RL is based on an agent learning to interact with its environment to maximize the cumulative reward signal.RL has been successfully used in various fields like robotics, game playing, natural language processing and other fields.This study examines the utilization of RL in the field of electrical motor design.A new approach based on RL is introduced to optimize the design of electrical motors, and its efficacy is demonstrated through the evaluation of various test cases.The thesis makes a noteworthy contribution by introducing a unique reward function tailored for the RL method.A comparison is made between the outcomes achieved by the proposed approach and those attained using traditional optimization methods.The potential advantages and difficulties associated with employing RL for electrical motor design are analyzed.The findings indicate that RL offers substantial enhancements in efficiency when searching for the optimal motor design within specified user requirements and manufacturing limitations, while also providing greater ease of use compared to traditional optimization techniques.Furthermore, potential directions for future research are outlined and discussed.
Source: Aaltodoc (Aalto University)
Author: Sarcheshmehpour, Yasmin
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7035870648
