CitationDownload as .RIS
Emerald Group Publishing Limited
Copyright © 2000, MCB UP Limited
Optimization software helps design low-noise rotorcraft flight procedures
Optimization software helps design low-noise rotorcraft flight procedures
Keywords: Software, Optimization, Helicopters, Noise levels
NASA engineers used commercial optimization software to design quiet approach procedures for an experimental tiltrotor aircraft. These procedures are predicted to be at least 6dB, or about 50 per cent, quieter than the baseline approach profile. They were able to harness their existing rotorcraft noise model to the optimization code, iSIGHT from Engineous Software, Inc., Morrisville, North Carolina, and produce usable results within weeks. The most interesting result from the optimization activity showed that a landing pattern in which the aircraft performs the first segment of the approach at a 0-degree approach angle (i.e. level flight), provided better results than many of the previous patterns they had developed by manual methods. Further testing is required to validate all of these new landing patterns but NASA engineers are optimistic that it will help to demonstrate the viability of using tiltrotor aircraft to relieve congestion at crowded airports.
More and more US airports are approaching their capacity with respect to the maximum number of daily flights that can be handled by their existing runways. As a result, the civil aviation industry is considering the use of tiltrotor aircraft, which have the ability to take off and land vertically while flying like an airplane during cruise, to reduce the demand on runways. However, the noise generated by a large 40-passenger tiltrotor aircraft, which would be nominally the same size as the V-22 Osprey, could be a potential barrier to civil market penetration. The Civil Tiltrotor Development Advisory Committee, in a report to Congress in December 1995, found that a civil tiltrotor was technically feasible, but noted that noise was the most critical environmental issue standing in the way of scheduled passenger service.
Complex engineering challenge
But while they are noisier overall than conventional aircraft when in helicopter mode, tiltrotor aircraft also provide engineers with an added noise abatement feature. This is the extra variable provided by the nacelle tilt capability, which can significantly reduce the noise footprint on the ground. The challenge for engineers working to demonstrate the feasibility of tiltrotors in civil aviation is the complexity of optimizing the nacelle tilt in addition to the normal landing pattern variables, airspeed and approach angle, to develop flight procedures that are quiet, safe, and easy to fly. Yet, the ability to optimize these factors will have an enormous impact in setting the design requirements for a civil tiltrotor aircraft, potentially even determining the economic viability of this new concept in aviation.
Rotorcraft noise modeling
The first step for engineers at NASA Langley who are currently investigating tiltrotor civil aviation in cooperation with Wyle Laboratories, Arlington, Virginia, was developing the Rotorcraft Noise Model (RNM), which allows them to estimate the noise experienced at nearly any point on the ground based on a given landing pattern. RNM uses as input from one to ten sets of sound pressure hemispheres that surround the aircraft, with each hemisphere centered at any specified location on the aircraft. One set of hemispheres provides broadband data in the form of one-third octave band sound levels. The other set provides narrow band data in the form of pure-tone sound pressure levels and phase. For example, a set of hemispheres could be centered on each of the two proprotors for a tiltrotor aircraft, or a single set could represent the entire aircraft, as would be acquired during actual measurements on an aircraft in flight.
Such a set of hemispheres were obtained by experimental measurements on the NASA/Army/Bell Helicopter XV-15, which is a smaller, experimental tiltrotor aircraft which has been a productive research aircraft for over 20 years. The flight track module in RNM determines the aircraft's position, airspeed, crab angle, angle of attack, roll angle, approach angle and nacelle angle at half second intervals. This information is passed to a propagation module, which uses these arrays to construct the sound spectra at each receiver position as a function of time and to calculate the integrated noise metrics, such as SEL and EPNL. The output of the program is noise contours on the ground, which are output graphically or in tabular format.
After developing the model, NASA engineers validated it by flying the XV-15 over an array of microphones that were deployed to directly measure the noise footprint produced during realistic approaches and departures. The results of these experiments were compared to the predictions of the RNM code. Generally, the noise predictions were within two SELdB, which is considered to be excellent correlation. Higher sound level discrepancies in a few areas were primarily attributed to the fact that the actual trajectory did not precisely match the parameters entered into the model.
Problems with conventional approaches
Being able to model the noise produced by a given landing pattern was a major step but still left NASA engineers far from their goal. They needed to know how much noise would be produced by the quietest, but still easily flyable, landing pattern. The conventional method of addressing this problem would have been to turn the analysis code over to an engineer and ask him or her to work with it until they thought they had a good answer. The problem with this approach was that there were an enormous number of combinations of the major design variables that would need to be evaluated in order to address this problem on a comprehensive basis. It could easily take an engineer several months or more to explore this design space. Even then, there would be no guarantee that the results were truly optimized. And down the road, as the problem definition continued to evolve, such as by substituting a less noisy aircraft or an area whose population was concentrated on one side of the landing zone, it would be necessary to start again from scratch.
That is why NASA has typically addressed problems such as this in the past by patching custom optimization routines on to analysis software. A programmer would typically spend several months developing a master program that combined the optimization routine and the analysis code. Besides being time-consuming this approach has the major disadvantage that it requires changes to the inner workings of the analysis code. The combined analysis-optimization code needs to be revalidated after the programming is completed. NASA engineers had already spent a large amount of time and money validating their rotorcraft noise model and were not anxious to repeat the process. So they decided to look at commercial optimization tools. According to Mike Marcolini, Langley's Rotorcraft Program Manager, "iSIGHT caught our attention because of its ability to interface with RNM without writing any code, by simply linking to our input and output files".
Creating the interface
"The process of creating an interface between iSIGHT and the RNM code was as simple as graphically selecting the right parameters in the input and output files of the model", said Sharon Padula, Senior Research Engineer for NASA. The variables that the engineers allowed iSIGHT to control were the airspeed of the aircraft during the three different segments of the approach, the altitude of the aircraft at the beginning of each of these three segments and the nacelle angle during each segment. Initially, the software was directed to minimize noise at three receiver locations on the ground plane. Constraints were applied to prevent the optimization software from selecting unsafe flight patterns, such as those in which the speed was too high or the approach angle too steep. In only one week, Padula had linked iSIGHT and RNM, developed a basic set of constraints and run several optimization routines.
Engineers configured iSIGHT to use a gradient-based optimization method to seek out a combination of design variables that would provide the minimum noise. It is worth noting that the software package provides an optimization adviser that guides the user towards the best of the several available techniques for a particular problem. The selected optimization method was combined with a response surface approximation method. Adding the approximation method is advantageous because:
it reduces the number of calls to the RNM code;
it reveals global opportunities for noise reduction; and
it maps out portions of the design space that need to be avoided.
Optimizing the landing pattern
At this point, the iSIGHT software was already capable of optimizing the noise level for any given set of receiver locations and constraints. It was then turned over to the team of engineers charged with the task of designing low-noise flight procedures. Determining exactly which flight paths were feasible for reasons other than noise is a challenging problem that occupied the engineering team for several months. But, whatever set of constraints were proposed, the team was able to determine the optimum landing pattern and corresponding noise level in a matter of a few hours. Only a few minutes were required on the part of an engineer to set up the problem and the rest of the time the software worked without any attention.
Being able to quickly optimize the noise level reduced what would have otherwise been a virtually impossible problem into one that could be handled by a small team in a few weeks. Throughout the process, the engineering team agreed that the optimization results routinely pointed them in the right direction. The team originally focused on a two-segment approach with a shallower angle used for the first step, as opposed to the baseline approach, which had a constant six-degree approach angle. The optimization software consistently moved in the direction of minimizing the steepness of the first step while maximizing the steepness of the second step. The team finally settled on a small set of approach profiles in which the angle of the first step is shallower (zero to three degrees) and the second step is usually nine degrees. These landing patterns hopefully provide acceptable flying conditions while still providing an overall noise value 50 per cent lower than the baseline approach profile, as measured on the ground below the flight path. These results need to be validated with further testing but they have, in the minds of the engineering team, already demonstrated the value of computerized engineering methods in simplifying otherwise difficult analysis tasks.
For more information, please contact Engineous Software, 1800 Perimeter Park West, Suite 275, Morrisville, NC 27560. Tel:'+1 919-319-7666; Web site: www.engineous.com