Bayesian optimization (BO) performs efficient global sequential sampling guided by an acquisition function (AF) based on the prior belief in the objective distribution modeled by a probabilistic surrogate model, which is widely employed in expensive black-box problems including airfoil design to shorten the design cycles and cut energy consumptions. Intuitively, multimodal airfoil design objectives require variable sampling preferences of the AF for exploration and exploitation across optimization stages, and mismatching between them causes either a waste of function evaluations or traps of local optima. Therefore, we proposed a portfolio-based airfoil Bayesian optimization framework for adaptive sampling, where the most suitable constituent in a group of AFs is exclusively authorized for the current stage under the instruction of a meta-criterion and the optimization process is thus sped up. Further in multi-objective design scenarios, full utilization of the possible correlation between airfoil design objectives has theoretical potential for improvements on both design efficiency and optimized performances. Therefore, we also proposed a multi-objective Bayesian optimization framework for objective correlation considerations in airfoil design, where through accurate capture and effective delivery and utilization of this information in the surrogate model and the AF, well-distributed Pareto-frontier airfoils can be obtained with limited computational resources.