A reduced order modeling methodology for the parametric estimation and optimization of aviation noise

Abstract

The successful mitigation of aviation noise is one of the key enablers of sustainable aviation growth. Technological improvements for noise reduction at the source have been countered by increasing number of operations at most airports. There are several consequences of aviation noise including direct health effects, effects on human and non-human environments, and economic costs. Several mitigation strategies exist including reduction of noise at source, land-use planning and management, noise abatement operational procedures, and operating restrictions. Most noise management programs at airports use a combination of such mitigation measures. To assess the efficacy of noise mitigation measures, a robust modeling and simulation capability is required. Due to the large number of factors which can influence aviation noise metrics, current state-of-the-art tools rely on physics-based and semi-empirical models. These models help in accurately predicting noise metrics in a wide range of scenarios; however, they are computationally expensive to evaluate. Therefore, current noise mitigation studies are limited to singular applications such as annual average day noise quantification. Many-query applications such as parametric trade-off analyses and optimization remain elusive with the current generation of tools and methods.

There are several efforts documented in literature which attempt to speed up the process using surrogate models. Techniques include the use of pre-computed noise grids with calibration models for non-standard conditions. These techniques are typically predicated on simplifying assumptions which greatly limit the applicability of such models. Simplifying assumptions are needed to downsize the number influencing factors to be modeled and make the problem tractable. Existing efforts also suffer due to the inclusion of categorical variables for operational profiles which are not conducive to surrogate modeling.

In this research, a methodology is developed to address the inherent complexities of the noise quantification process, and thus enable rapid noise modeling capabilities which can facilitate parametric trade-off analysis and optimization efforts. To achieve this objective, a research plan is developed and executed to address two major gaps in literature. First, a parametric representation of operational profiles is proposed to replace existing categorical descriptions. A technique is developed to allow real-world flight data to be efficiently mapped onto this parametric definition. A trajectory clustering method is used to group similar flights and representative flights are parametrized using an inverse-map of an aircraft performance model. Next, a field surrogate modeling method is developed based on Model Order Reduction techniques to reduce the high dimensionality of computed noise metric results. This greatly reduces the complexity of data to be modeled, and thus enables rapid noise quantification. With these two gaps addressed, the overall methodology is developed for rapid noise quantification and optimization. This methodology is demonstrated on a case study where a large number of real-world flight trajectories are efficiently modeled for their noise results. As each such flight trajectory has a unique representation, and typically lacks thrust information, such noise modeling is not computationally feasible with existing methods and tools. The developed parametric representations and field surrogate modeling capabilities enable such an application.