This work presents two different machine learning (ML) algorithms. In the first, a random forest (RF)-based multifidelity ML algorithm to predict the high-fidelity Reynolds-averaged Navier-Stokes flow field is proposed. In the second, the gradient-enhanced multifidelity neural networks (GEMFNN) algorithm is proposed. The RF ML algorithm is used to increase the fidelity of a low-fidelity potential flow field.

Two cases are studied, a backward-facing step and a subsonic flow around the NACA 0012 airfoil. The proposed GEMFNN algorithm is applied to analytical functions and a transonic airfoil optimization case. It is compared to other ML algorithms, namely, neural networks (NN), gradient enhanced NN, and multifidelity NN. This work demonstrates the benefits of using gradient and multifidelity information in NN for high-dimensional problems. The scope of this work is to show the computational cost benefits of combining multifidelity data with ML algorithms to make high-fidelity predictions.