Semi-empirical Bayesian helicopter noise source model

Flight plans are commonly developed for helicopters that perform low-noise operations. There are military cases where these flight plans need to be developed within a matter of hours. These mission planners rely on computer models to accurately and efficiently estimate the noise that will be developed for certain maneuvers. These models depend on accurate estimations of the noise generated by the helicopters æat the sourceÆ. However, developing helicopter noise source models for use in flight path planning is currently limited due to enormous computational time requirements utilizing high fidelity simulations, or by collecting expensive noise measurement data. Development of a physics-based semi-empirical model that can be trained using experimental data will allow mission planners to optimize a large number of flight plans. The proposed approach combines a low-order aerodynamic algorithm with an efficient implementation of the Ffowcs-Williams Hawkings methodology to produce helicopter noise sources for arbitrary flight path conditions. The aerodynamic algorithm implements a modified strip theory/vortex lattice approach to capture the effects of all the relative blade velocity effects and inflow conditions, while a prescribed vortex wake approach will capture any potential blade vortex interactions. The parameters of this model will be set based on historical modelling approaches and can be trained by a Bayesian parameter estimation approach to minimize the noise error between a measured and modeled flight condition. The Bayesian parameter estimation module will couple a parameter subset selection algorithm to identify the minimum number of parameters necessary to efficiently tune the predictive noise estimate. Additionally, a Gaussian process algorithm will tie the model parameters to non-dimensional flight condition hyper-parameters that will model the function that describes how the parameters change with flight conditions. The proposed approach will have the added benefit of quantifying the uncertainty at various stages and allow operators to be confident of the noise sources developed.