Model validation metrics for CFD numerical simulation under aleatory and epistemic uncertainty

With the emergence of new computer technologies, Computational Fluid Dynamics (CFD) numerical simulation has been extensively implemented in many areas such as aerospace, national defense, ship hydrodynamics, wind power, and water conservancy. CFD numerical simulation provides effective decision-making and validation methods for complex fluid analysis of equipment, model parameter evaluation, and aerodynamic optimization design. Existing CFD simulations are conducted under the premises of deterministic mathematical models, fixed physical parameters, and fixed boundary conditions. Due to the complexity of physics and cognitive biases of human beings, there are, however, many potential uncertain factors with different representation forms in CFD, such as model parameter uncertainty, numerical dispersion, and model form uncertainty, resulting in a great challenge to the credibility of CFD simulation results. This article elaborates on the uncertain factors encountered in CFD, and provides a comprehensive discussion on the mainstream model validation metrics. We focus on the model validation metrics under epistemic uncertainty, including several metrics under the interval theory and probability-box theory. The NACA0012 airfoil flow problem is leveraged to demonstrate the effectiveness of the model validation metrics under various uncertainties in CFD. © 2022 AAAS Press of Chinese Society of Aeronautics and Astronautics. All rights reserved.