Matthew J. LeVine

Development of Generic Vehicles for Fleet-Level Analysis of Noise and Emissions Tradeoffs

A fleet-level analysis of technology impacts on environmental metrics via generic vehicle design is proposed. Vehicles are grouped into “vehicle classes” distinguished by the vehicle-level environmental metrics, which include fuel burn (as a surrogate of CO2 emissions), NOx emissions, and SEL noise contours. These groupings are compared against traditional seat class groupings. Target metrics are established for a subset of 94 airports by designing a series of tests of sequentially increasing complexity, with ideal generic vehicle designs minimizing the error distributions across these airports when vehicle-level performance is aggregated to fleet-levels. Latin hypercube design of experiments are employed to explore the Environmental Design Space (EDS), and Stochastic Multicriteria Acceptability Analysis (SMAA) is used to evaluate potential generic vehicle designs against each other and identify the best designs for simultaneously matching aggregate fuel burn, NOx emissions, and DNL contours. In general, the average generic vehicles provide greater accuracy for each of these metrics across across the 94 airports for a representative six weeks of operations at these airports derived from various sources such as the Bureau of Transportation Statistics (BTS), as compared to the traditional approach of choosing a representative vehicle per class. The average generic vehicle approach works well for both vehicle-class and seat-class groupings, but the former leads to slightly tighter error distributions for all of the metrics.

Average Generic Vehicle Method for Fleet-Level Analysis of Noise and Emission Tradeoffs

A method is proposed for modeling average generic vehicles for fleet-level analysis of technology impacts on environmental metrics. Vehicles are grouped into classes distinguished by aircraft-level metrics, which include fuel burn (a surrogate of carbon dioxide emissions), nitrogen oxide species (NOx) emissions, and sound exposure level noise contours. Target metrics are established for a subset of 94 airports by designing a series of tests of sequentially increasing complexity, with ideal average generic vehicle models minimizing the error distributions across these airports when aircraft-level performance is aggregated to fleet levels. A Latin hypercube design of experiments is employed to explore the aircraft-level design space and construct surrogate models for each metric. Monte Carlo samplings of these surrogate models are paired with desirability functions to rapidly identify locations in the design space that simultaneously match aggregate fuel burn, NOx emissions, and day/night-level contours. In...

Methodology for Calibration of ANGIM Subjected to Atmospheric Uncertainties

The Airport Noise Grid Integration Method (ANGIM) was developed to enable rapid airport and fleet-level noise contour analysis, but the savings in computational speed over higher-fidelity noise models such as the Aviation Environmental Design Tool (AEDT) are accomplished by requiring a few simplifying assumptions. These assumptions limit the robustness of ANGIM under varying atmospheric and operational conditions. The underlying concept of ANGIM is rooted in the fact that Day-Night Average Level (DNL) noise grids calculated by a detailed noise tool are simply logarithmic additions of all the Sound Exposure Level (SEL) events occurring during a given flight schedule. ANGIM enables rapid calculation of airport-level DNL grids by pre-calculating SEL grids for individual aircraft at standard day sea-level atmospheric conditions and maintaining a database of these grids. To enable calculation of DNL grids in ANGIM under atmospheric uncertainty would require generating and storing grids for every aircraft under every possible atmospheric condition, which would be both time and cost prohibitive. This study outlines a methodology for calibration of standard day sea-level SEL grids for varying atmospheric conditions using neural nets. This calibration method would improve the robustness of ANGIM under varying atmospheric conditions with a small increase in computation time, but this increase in computation time is still much faster than a detailed model and expands the applicability of ANGIM as a screening tool.

Noise-Sensitivity to Vehicle-Level Design Variables

i Space-filling design of experiments are performed on the Environmental Design Space (EDS) architecture to enable an understanding of the sensitivity of various noise-metrics to vehicle-level design variables. These include aerodynamic, propulsion, and airframe design variables. Half-normal probability plots are used to show that airframe design variables dominate approach noise, while design variables related to bypass ratio have the greatest influence on departure noise. These results are consistent for both NPD-curves and sound- exposure-level (SEL) contour areas, but not for the certification effective-perceived-noise- level (EPNL) metrics or Maximum A-weighted Sound Level (LAmax) based spectra, which demonstrate some unique sensitivities. Interdependencies of noise, fuel burn, and NOx emissions metrics are also explored.

Methodology for Runway-Level DNL Contour Calibration in ANGIM to Capture Impacts of Deviation from Standard Day Sea-Level Atmosphere

The Airport Noise Grid Integration Method (ANGIM) was developed to enable rapid airport and fleet-level noise contour analysis, but the savings in computational speed over higher-fidelity noise models such as the Aviation Environmental Design Tool (AEDT) are accomplished by requiring a few simplifying assumptions. These assumptions limit the robustness of ANGIM under varying atmospheric and operational conditions. The underlying concept of ANGIM is rooted in the fact that Day-Night Average Level (DNL) noise grids calculated by a detailed noise tool are simply logarithmic additions of all the Sound Exposure Level (SEL) events occurring during a given flight schedule. ANGIM enables rapid calculation of airport-level DNL grids by pre-calculating SEL grids for individual aircraft at standard day sea-level atmospheric conditions and maintaining a repository of these grids. To enable calculation of DNL grids in ANGIM under atmospheric uncertainty would require generating and storing grids for every aircraft under every possible atmospheric condition, which would be both time and cost prohibitive. Previous studies demonstrated a methodology for calibration of standard day sea-level SEL grids for varying atmospheric conditions using neural nets. Implementing that method, however, would still require vehicle-level calibration for every vehicle in the fleet, which could greatly increase computational time. This study instead explores introducing this calibration after the runway-level DNL contours are calculated, which would be much less computationally demanding than calibrating each aircraft SEL grid. To create this runway-level calibration requires running mixture DOEs for a set of representative vehicles at various atmospheric conditions, and then constructing a neural net calibration as a function of the mixture, the volume of operations, and the atmospheric conditions. A notional study of an airport schedule with three unique aircraft is explored to test the feasibility of calibrating runway-level DNL grids for varying atmospheric conditions. Accuracy of the calibration method with respect to the actual results are compared through contour area calculations as well as visual overlays of the calibrated runway level contours and the actual runway level contours.

Stochastic Multicriteria Acceptability Analysis for Nominal is Best MCDA in the context of Generic Vehicles

Stochastic Multi-criteria Acceptability Analysis (SMAA) is used to assess potential generic vehicle designs that simultaneously match target fuel burn, NOx emissions, and DNL noise contours. While SMAA is typically used for criteria with minimum is best or maximum is best formulation, the nature of the generic vehicle problem is a nominal is best formulation. Absolute value of error distributions transform the nominal is best problem into a minimum is best formulation, which provides the additional benefit of exposing the existence of model biases. Design rules for the criteria value functions with respect to the variance of the criteria are explored to avoid favoring accuracy in one metric versus another.