A study on liquid-cooled jackets for synchronous reluctrance motors and application of machine learning in cooling strategies

Name: Hoang Khan Nguyen

LCSH: Motors

LCSH: Electric driving -- Automatic control

LCSH: Machine learning

Abstract: The prevalence of electric vehicles (EVs) is a near future that we can envision. As the driving force of EVs, the motor needs attention in its structure and safety systems, which require research. A motor with a well-functioning cooling system can operate in more adverse environments. In the system, there needs to be a balance between design and control systems, as an unstable control system, no matter how intelligent the design, will result in lower efficiency and vice versa. Therefore, the research objective includes assessing the factors affecting the cooling process, proposing a rational design, and applying an intelligent control method for the coolant flow. This study investigates and evaluates the factors affecting the cooling process through experimentation and simulation. The experimental process serves two primary purposes: collecting input parameters for the simulation step and validating the simulation results to ensure confidence. The simulation process is based on the step running technique, offering the advantages of fast computation and reduced computer memory usage. Machine learning (ML) is an intelligent water pump control method with 1000 cases of training data, including three key parameters: motor torque, flow rate, and inlet temperature. The results show that a water-cooling jacket can significantly reduce the temperature of the motor; the heat generation of the motor and the flow rate of water affect the maximum temperature and temperature fluctuation This comprehensive validation method incorporates three types of validation: time-dependent validation, stable state validation of 3 key parameters, and temperature distribution validation. The topology of the water-cooling jacket (WJ) influences temperature difference, pressure drop, and temperature distribution. เท term of control, ML models can directly determine the optimal flow rate and save pump capacity. An equation derived from a machine learning (ML) model functions as a motor cooling system to optimize memory and facilitate integration with other systems, even large systems like EVs. The accuracy of the equation was certificated by calibrating it with the ML model and checking its loss against experimental data, achieving R2 scores of 0.9901 and 0.9994, respectively. These results inferred that the error of the equation will increase as the boundary conditions deviate further from the actual experimental data. This equation is the initial step for generating a deep neural network (DNN) model to derive equations for various types of motors if the database for the training step is sufficient

King Mongkut's Institute of Technology Ladkrabang. KMITL Lifelong Learning Center

Address: BANGKOK

Email: lifelong@kmitl.ac.th

Name: Chinda Charoenphonphanich

Role: Thesis Advisor

Email : chinda.ch@kmitl.ac.th

Name: Manop Masomtob

Role: Thesis Co-Advisor

Name: Shuichiro Hirai

Role: Thesis Co-Advisor

Created: 2024

Modified: 2025-07-26

Issued: 2025-07-26

วิทยานิพนธ์/Thesis

application/pdf

eng

DegreeName: Master of Engineering

Level: Master's Degree

Descipline: Automotive and Advanced Transportation Engineering

Grantor: King Mongkut's Institute of Technology Ladkrabang

©copyrights King Mongkut's Institute of Technology Ladkrabang

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	Hoang Khan Nguyen.pdf	4.81 MB

ใช้เวลา

0.030899 วินาที