24小时热门版块排行榜

返回列表

中关村胖虎

铁虫 (小有名气)

应助: 0 (幼儿园)
金币: 349.4
帖子: 105
在线: 9.7小时
虫号: 27517193

[交流] 【2025伦敦玛丽女王大学-CSC公派】Amin PAYKANI老师招博士新生：燃烧方向

【项目简介】
研究方向：燃烧。
课题名称：Data-Driven Optimisation of Hairpin Winding and Oil Cooling in Traction Motors for Improved Thermal Management.
联系方式：a.paykani@qmul.ac.uk（感兴趣的同学请从速联系 Dr Amin Paykani）。
截止日期：2025年1月29日。
详情介绍请见官方链接，QMUL-SEMS-CSC

【福利】
提供csc奖学金外的额外资金，用于参与国际会议、走访合作大学或工业伙伴等。

Description
Increasing the power density of traction motors is a critical challenge for the next generation of electric vehicles. Combining hairpin windings with direct oil cooling has emerged as a popular solution, but optimising the design of such systems requires a deep understanding of fluid dynamics and heat transfer. The formation of the oil film on windings is influenced by various factors, including jet parameters and winding geometry, making the design process complex and computationally expensive when relying on traditional high-fidelity Computational Fluid Dynamics (CFD) simulations. This PhD project aims to develop a data-driven framework that integrates experiments, CFD, and Machine Learning (ML) to co-optimise the hairpin winding geometry and oil injector parameters for enhanced cooling performance. The research will proceed through the following steps:

Experimental Validation: The initial phase will involve conducting experiments to capture data on oil jet behavior and cooling performance in hairpin windings. This experimental data will be used to validate the CFD models, ensuring accuracy in simulating the complex interactions between oil jets and winding surfaces.

[list=2]

CFD Simulations: Following validation, a detailed 3D CFD setup will be used to simulate the system under various operating conditions. These simulations will form the foundation for generating the training dataset for the ML model. High-fidelity CFD results will be leveraged to understand the effects of different winding geometries and oil jet parameters on the formation of the oil layer and heat transfer efficiency.

[list=3]

ML-Based Surrogate Modeling: Using the experimental and CFD data, a hybrid surrogate model will be developed with the help of advanced ML techniques. Bayesian optimisation will be applied to select the optimal ML hyperparameters, ensuring the model accurately predicts system performance while reducing computational costs.

[list=4]

Design Optimisation: An optimisation technique will be employed to find the optimal combination of hairpin winding geometries and oil injector parameters, based on the ML surrogate model. This approach allows for efficient exploration of the design space, identifying configurations that maximise cooling effectiveness and power density.

回复此楼