https://journals.bilpubgroup.com/index.php/ese <p>ISSN: 2661-3247(Online)</p> <p>Email: ese@bilpublishing.com</p> <p><a href="https://journals.bilpubgroup.com/index.php/ese/about/submissions#onlineSubmissions" target="_black"><button class="cmp_button">Online Submissions</button></a></p> BILINGUAL PUBLISHING GROUP en-US Electrical Science & Engineering 2661-3247 https://journals.bilpubgroup.com/index.php/ese/article/view/7532 <p>Motor design involves multiple complex parameters, and traditional methods rely on experience and experimentation, which are inefficient and difficult to optimize. With the rapid development of emerging industries such as electric vehicles and intelligent manufacturing, the performance requirements for motors are constantly increasing. Optimizing design under multi-objective and multi-constraint conditions has become a key challenge. To address this, this paper proposes a motor design performance optimization algorithm based on Graph Neural Networks (GNN) representation and adaptive weighting. GNN, as a deep learning model capable of handling complex structured data, can model the multi-parameter relationships in motor design and automatically extract key features through its feature propagation mechanism, thus overcoming the difficulty of capturing parameter dependencies with traditional methods. At the same time, Mixed-Integer Linear Programming (MILP) provides a powerful global optimization tool that can find the global optimal solution when dealing with complex decision variables and constraints, overcoming the shortcomings of traditional optimization algorithms in terms of global convergence. Moreover, the adaptive weighting mechanism allows the optimization algorithm to dynamically adjust the weights according to the influence of parameters on motor performance, ensuring the accuracy and adaptability of optimization results in different scenarios. Through the organic combination of these three methods, this paper aims to solve the problems of low efficiency, poor global convergence, and inability to dynamically adjust the importance of design parameters in traditional motor design optimization. By introducing advanced machine learning models and optimization algorithms, this work constructs an efficient motor design performance optimization framework.</p> Guojun Zhang Weidong Huang Tong Zhou Copyright © 2024 Guojun Zhang, Weidong Huang, Tong Zhou https://creativecommons.org/licenses/by-nc/4.0 2024-10-26 2024-10-26 6 1 1 13 10.30564/ese.v6i1.7532