Machine learning driven interpretation of computational fluid dynamics simulations to develop student intuition

Abstract

Employers need engineers capable of leveraging CFD simulations to make intelligent design decisions, but undergraduate computational fluid dynamics (CFD) courses are not adequately preparing students for this type of work. CFD courses commonly familiarize students with topics, such as method derivation, domain creation, boundary conditions, mesh convergence, turbulence models, numerical convergence, and error analysis. This approach is an effective way to teach novices how CFD software works and how to prepare CFD analyses. However, it neglects development of higher level CFD skills and intuition important to engineering analysis and design, deferring this task to future study and training. This paper introduces the “Machine Learning Driven Interpretation of Fluid Dynamics Simulations to Develop Student Intuition” (MIFoS) software, a program designed to help CFD novices develop the high‐level skills and intuition that employers need in their engineers. A data‐driven approach was used to create the MIFoS software, which allows the submission of arbitrary geometries, automates an external flow simulation, and returns expert‐level graphical interpretation of simulation data. MIFoS's automated CFD simulation and feedback space allows novices to experiment with expert‐level suggestions on their own designs, enabling the skill and intuition development typically gained through years of study, practice, and expert guidance.

Case study‐based learning using a computational tool to improve the understanding of the jet engine cycle for aerospace engineering degree students

“…Gamez-Montero et al [9] combined computer-based learning with flipped teaching strategies to improve the understanding of turbomachinery working principles, showing an improvement in students' satisfaction. Patterson [23] used machine-learning techniques in computational fluid dynamics environment, with the aim of helping students to understand how to optimize an aerodynamic profile to reduce drag force. Minichiello et al [22] presented a mobile-based particle image velocimetry tool, concluding that it helped to increase students' engagement.…”

Experiencia sobre el empleo del estudio de caso y herramientas de simulación para mejorar la comprensión del ciclo termodinámico de motores a reacción

“…Gámez-Montero et al (2020) combinaron el aprendizaje asistido por ordenador con estrategias de clase inversa para mejorar la comprensión de los principios de funcionamiento de las turbomáquinas, mostrando una mejora en la satisfacción de los estudiantes. Patterson (2020) utilizó técnicas de aprendizaje automático en un entorno de mecánica de fluidos computacional (CFD), con el objetivo de ayudar a los estudiantes a comprender cómo optimizar un perfil aerodinámico para reducir la resistencia aerodinámica. Minichiello et al (2020) presentaron una herramienta de velocimetría de imágenes de partículas basada en dispositivos móviles, y concluyó que contribuía a mejorar la participación de los estudiantes.…”