Handling Qualities Optimization for Rotorcraft Conceptual Design

Handling Qualities Optimization for Rotorcraft Conceptual Design

Ben Lawrence Colin R. Theodore Tom Berger San Jose State University, Wayne Johnson U.S. Army Aviation Development NASA Ames Research Center, National Aeronautics and Space Directorate Moffett Field, CA Administration Moffett Field, CA NASA Ames Research Center M offett Field, CA Abstract Over the past decade, NASA, under a succession of rotary-wing programs has been moving towards coupling multiple discipline analyses in a rigorous consistent manner to evaluate rotorcraft conceptual designs. Handling qualities is one of the component analyses to be included in a future NASA Multidisciplinary Analysis and Optimization framework for conceptual design of VTOL aircraft.

Similarly, the future vision for the capability of the Concept Design and Assessment Technology Area (CD&A-TA) of the U.S Army Aviation Development Directorate also includes a handling qualities component. SIMPLI-FLYD is a tool jointly developed by NASA and the U.S. Army to perform modeling and analysis for the assessment of flight dynamics and control aspects of the handling qualities of rotorcraft conceptual designs. An exploration of handling qualities analysis has been carried out using SIMPLI-FLYD in illustrative scenarios of a tiltrotor in forward flight and single – main rotor helicopter at hover. Using SIMPLI-FLYD and the conceptual design tool NDARC integrated into a single process, the effects of variations of design parameters such as tail or rotor size were evaluated in the form of margins to fixed- and rotary-wing handling qualities metrics as well as the vehicle empty weight. The handling qualities design margins are shown to vary across the flight envelope due to both changing flight dynamic and control characteristics and changing handling qualities specification requirements. The current SIMPLI-FLYD capability and future developments are discussed in the context of an overall rotorcraft conceptual design process.

Nomenclature 6 - DoF Six - degree - of - freedom OLOP Open Loop Onset Point Angular velocity about body, X, Y, Z - axes RCAH Rate - Command/Attitude - Hold (control system 𝑝 , 𝑞 , 𝑟 response type) nd Non - dimensional units RCD H Rate - Command/Direction - Hold (control system Velocity along body X, Y, Z - axes 𝑢 , 𝑣 , 𝑤 response type) ADD Aviation Development Directorate (U.S. Army) RCH H Rate - Command/Height - Hold (control system DM Design Margin response type) HQ Handling Qualities VTOL Vertical Take - Off and Landing Multidisciplinary Design Analysis and st MDAO 𝛽 , 𝛽 1 order longitudinal and lateral rotor flapping 1 𝑐 1 𝑠 1 1 Optimization states (subscript is rotor index) NDARC NASA Design and Analysis of Rotorcraft 𝜙 , 𝜃 , 𝜓 Euler angle orientation of body axes w.r.t to inertial frame Introduction  tools and analyses are needed now to incorporate the growing The process of designing a rotorcraft has remained a largely serial number of design constraints and complex system trades needed process, such that competing design objectives are not evaluated to assess future rotorcraft designs [1] .

in a formal, automated fashion. Instead, the outcome of separate optimization processes representing different disciplines (e.g., Handling qualities analyses have been historically neglected in rotor aerodynamics, propulsion, etc.) are exchanged and the aircraft conceptual design process [2] and [3] . In fixed-wing discussed by subject-matter experts from the design team and an design, requirements for good stability, control and handling iterative cycle between design groups ensues until certain qualities are addressed through the use of tail volume coefficients, objectives — usually empty weight and speed — are met. For new location of the center of gravity and relatively simple static VTOL aircraft manufacturers, especially those engaged with non- analyses [3] . Rotorcraft, in particular high performance designs, traditional configurations, in-house discipline tools appropriate cannot rely on such methods, since most bare-airframe designs are for rotary wing vehicles are rarely available. Formal optimization unstable typically and require a stabilizing control system to make Presented at the Rotorcraft Virtual Engineering Conference, Liverpool, UK, Nov 8-10, 2016. This is a work of the U.S.

Government and is not subject to copyright protection.

them adequately flyable. In the context of rotorcraft, Padfield optimization methods that couples these analyses via an (Ref [4] ) notes that 25-50% of flight testing time in an aircraft OpenMDAO framework [1] .

development program might be spent on fixing handling qualities Handling qualities (HQ) is one of the component analyses to be problems. Furthermore, Padfield suggests that handling qualities were not given their proper place in the early design trade-space, included in a future NASA framework for conceptual design of VTOL aircraft. Similarly, the future vision for the MDAO and were often left until flight test to discover and “put right” .

During the early days of helicopter development, Padfield notes (Multidisciplinary Analysis and Optimization) capability of the that handling qualities were extremely difficult to predict and Concept Design and Assessment Technology Area (CD&A-TA) of the U.S Army Aviation Development Directorate also includes were justifiably treated as an outcome of the series of complex design decisions relating to, for example, overall performance, a handling qualities component. In response, a new tool “ SIMPLI- FLYD ” [7] (Simplified Flight Dynamics for Conceptual Design) vehicle layout and structural integrity, and fixing vibration problems. was developed in a NASA and U.S. Army collaboration. SIMPLI- FLYD was developed to utilize the output of the NDARC [8] Ref. [5] emphasizes flying qualities as the vehicle stability, sizing code and integrates the CONDUIT [9] control analysis and control and maneuvering characteristics and handling qualities as optimization tool. The SIMPLI-FLYD toolset is designed to the combination of flying qualities and the broader aspects of the perform flight dynamics, control and handling qualities modeling mission task, visual cues and atmospheric environment. I n many and analysis of rotorcraft conceptual designs, and also provides a instances handling qualities is used informally as the vehicle capability to “fly” the concept designs in an X-Plane based real- oriented flying qualities, as ref [4] notes, there appears to be no time simulation.

universal acceptance on the distinction . In this paper, handling The objective of this paper is to demonstrate and explore SIMPLI- qualities is used in the “colloquial” sense in that it refers to the FLYD and NDARC integrated in a coupled process to investigate vehicle flight dynamic stability and control aspects. As such, the how handling qualities analyses participate in rotorcraft term “handling qualities requirements” refer to the stability and conceptual design. To develop understanding of handling control characteristics that have been determined to lead to good qualities in conceptual design, evaluations of the SIMPLI-FLYD piloted handling/flying qualities.

toolset in “typical” design scenarios were required. The paper will The lack of detailed flight dynamics modeling at the earliest present results of studies using NDARC and SIMPLI-FLYD of stages of design disregards a potentially significant contributor to the pitch axis handling qualities for conceptual design models of size, weight, and performance estimates for some design a tiltrotor aircraft in forward flight, and for the yaw axis handling activities. Omission of flight dynamics modeling during qualities of a single-main rotor helicopter in hover. Also conceptual design also defers flight dynamics, rotor response lags, presented are the results of a stability and control derivative and control authority considerations to later in the design process, sensitivity study developed to address a number issues related to which have led to problems during flight test. The flight dynamics the primary NDARC/SIMPLI-FLYD coupling task. The paper and control of an air vehicle are fundamentally a function of its will conclude with a summary of the lessons learned from the inherent control power and damping characteristics and are analysis so far and outline of planned future developments.

typically augmented by the feed-forward and feed-back loops SIMPLI-FLYD Overview programmed into the flight control system. Predicting these characteristics of a yet-to-be-built air vehicle at the conceptual Figure 1 shows the primary components within the SIMPLI- design phase may offer paths to avoid handling qualities issues FLYD process as well as the key interfaces to external later in the design lifecycle or to minimize over-design when components and processes. The dashed blue box indicates the faced with uncertainty in the handling qualities of a design. tools and activities encompassed in an overall conceptual design process involving NDARC. The green box is the SIMPLI-FLYD Over the past decade, NASA, under a succession of rotary-wing functions. Stage (1) is the primary conceptual design sizing programs has emphasized the importance of physics-based activity using NDARC. In a future context, this process might be modeling and interdisciplinary optimization, [1] . NASA has been represented by a variety of other analyses encompassed in a moving towards coupling multiple discipline analyses in a MDAO environment. Stage (1) is the source of input for SIMPLI- rigorous consistent manner to evaluate rotorcraft conceptual FLYD that encompasses stages (2) through (4) where simplified designs. NASA has developed a state-of-the-art aircraft sizing linear flight dynamics models are calculated, integrated with code, NDARC [6] , and is focusing on a more global vehicle control laws, and then analyzed and optimized by CONDUIT. The approach to the optimization of VTOL configurations. This global output stages (2) through (4) are set(s) of stability, control and approach requires the inclusion the analysis of aerodynamics, handling qualities parameters (stage 5), and an X-Plane acoustics, propulsion, handling qualities, and structures, among compatible real-time simulation model for use in an X-Plane others, to capture the critical interdisciplinary aspect of rotary simulation station (stage 6).

wing vehicle design. NASA ’ s ultimate goal is to develop formal Figure 1 SIMPLI-FLYD Architecture for including stability and control analysis into conceptual design Reference [7] reports a full description of the SIMPLI-FLYD toolset and its sub-functions; some of the key aspects are highlighted in the following section.

(no flapping terms in linear model). Other configurations that Flight Dynamics Modeling feature two main rotors for example, such as a tiltrotor or a The flight dynamics modeling uses a modular approach to tandem, contain 13-states, with 4 rotor states.

representing a vehicle with various combinations of rotors, wings, other surfaces and auxiliary propulsion, similar to NDARC. The For the control derivatives, a simplification was imposed for the imported data from NDARC consists of geometric, aerodynamic, CONDUIT point analysis flight dynamic models such that any and configuration data about the vehicle and pre-calculated trim vehicle had a fixed set of four “controls” for the primary roll, data for the flight conditions to be assessed. The flight dynamic pitch, thrust and yaw response axes. The effects of multiple or calculations then loop over the flight conditions and components redundant control effectors such as combinations of rotor controls calculating linear stability and control derivatives for each and wing or aerodynamic surface controls are combined via an component (a component is a force and moment generating NDARC defined “mixing matrix” in advance of analysis at stages element: rotors, wings, aerodynamic surfaces and fuselage). For (3) and (4) (the separate control derivatives are retained for use in the rotors, this process uses a blade element model which is the real-time model). The actuator characteristics are configurable initialized at the NDARC calculated trim state using numerical for each analysis point model to allow representation of different perturbation to calculate the stability derivatives. For the other actuator classes (i.e. swashplate vs. aerodynamic surfaces) components, a simplified calculation of the linear derivatives is required for particular flight conditions/configurations.

performed using a mix of analytical and empirical models. The total vehicle linear models are then computed through the Control System Modeling, Analysis and Optimization summation of the state- space ‘A’ and ‘B’ matrix terms from the The control system applied to the vehicle model at stage 3 in various components. The linear models can be optionally 6- Figure 1 is based on an explicit model-following architecture degree-of-freedom (6-DoF) rigid body states only or can include that consists of independent feed-forward and feedback paths, first-order flapping equations, with one longitudinal and one shown in Figure 2 . The control laws use a generic architecture lateral per “main” rotor, following the “hybrid” model with varying modes appropriate for use at different flight formulation in Tischler [10] . As such, a single main rotor conditions as per Table 1.

configuration has 11 states: 9 rigid body states (𝑢, 𝑣, 𝑤, 𝑝, 𝑞, 𝑟, 𝜑, 𝜃, 𝜓) and 2 rotor states (𝛽 , 𝛽 ) as the tail 1𝑐 1𝑠 1 1 rotor derivatives are always reduced to their 6-DoF contribution boundaries are drawn from the rotorcraft specifications in ADS- 33E [11] and the fixed-wing specifications of MIL-STD-1797B [12] . For the full list of the specifications currently used in SIMPLI-FLYD see Ref [7] .

Once the control system optimization is complete, the block diagram parameters (feedback gains, feed-forward gains, inverse model parameters, etc.) and the HQ specification results of the control system optimization, given individually for each axis, are saved for output and further use.

In addition to the individual specification parameter values and Figure 2 Explicit Model Following Architecture.

percent over/under margins, the CONDUIT analysis provides a Table 1. Control system response types for various axis and flight single HQ requirements “design margin” (DM) for each of the modes primary control axis analyzed (roll, pitch, yaw, vertical) and for both the feedback and feed-forward control paths. The DMs are Rotor - Borne Wing - Borne the percent over/under design for the worst or limiting specification for each feedback/forward/axis combination.

Hover Forward - Flight Forward - Flight Roll RCAH RCAH RCAH Method of Analysis of SIMPLI-FLYD/NDARC Angle - o f - Attack - Coupled Process Pitch RCAH RCAH Command In this paper, the results of two different analyses will be Sideslip - Yaw RCDH Sideslip - Command presented. The motivation for carrying out each of the analyses Command and the methods applied is first presented. The primary objective Thrust RCHH Open - loop Open - loop was to develop and assess NDARC and SIMPLI-FLYD coupled RCAH = Rate-Command/Attitude-Hold into a combined analysis. The goal of the NDARC/SIMPLI- RCDH = Rate-Command/Direction-Hold FLYD coupled analysis was to run sweeps of design parameter RCHH = Rate-Command/Height-Hold variations relevant to the handling qualities of the design scenarios selected (pitch axis forward flight or yaw axis hover The setup and optimization of the control laws is fully automated HQs) and examine the impact on both the SIMPLI-FLYD HQ within CONDUIT® and the overall SIMPLI-FLYD process output and NDARC design sizing. The main task for the (based on certain user configurations). The control system is development of the NDARC/SIMPLI-FLYD coupled analyses optimized for each axis where key metrics are used to assess the was the creation of Python- based “wrapper” scripts which level of over- or under-design in the control system, for both the handled the input data and variable initialization, called the feedback and the feed-forward paths. In the case of feedback, the NDARC and SIMPLI-FLYD (via Matlab) codes, and handled the metrics are for the control system's (combined with the vehicle) data interface and collection in a common environment. The stabilizing performance robustness and ability to reject NDA RC run and utility functions were drawn from the “rcotools” disturbances. Starting with a baseline required value for each library, a Python-based toolset being developed by NASA for the specification (defining 0% over-design), the requirements are integration of NDARC into the OpenMDAO environment.

progressively increased (more over-design) until a feasible design can no longer be achieved. If the baseline design cannot be met, Another aspect to preparing the SIMPLI-FLYD/NDARC coupled the requirements are decreased (under-design) until a feasible analyses was to down-select a subset of the many possible design solution is achieved. After the feedback path is optimized, the variables available. The down-selection of design parameters was feed-forward path is optimized using specifications such as primarily necessary to reduce the computational task and avoid piloted bandwidth, quickness, and control power. The the “Curse of Dimensionality” [13] where for k-dimensions k optimization is carried out in a phased process; CONDUIT tunes (parameters) n runs or calculations are required for a full factorial the gains to first meet all of the “ Hard Constraints ” (stability sweep of all possibilities. For example, a NDARC/SIMPLI- specifications), then the soft constraints (the handling qualities FLYD coupled analysis run for a single flight condition (all axes) specifications) are optimized. Finally, CONDUIT tunes the gains currently takes typically 15-20 minutes on a desktop PC, and to reduce the Summed Objective (“cost of feedback” or computation times would rapidly increase as the number of design performance specifications) to find the design that meets the parameter dimensions increase. Therefore it was not a realistic requirements with the minimum cost (e.g. such as actuator task to run sweeps of all potential inputs and evaluate the output.

requirements).

Although for the design scenarios being examined an experienced flight dynamics engineer could likely choose the most relevant The handling qualities specifications used to drive the control design variables it was desirable to devise a method that might system optimization are divided by aircraft type, flight regime, verify the down-selection process. A technique was devised to control axis, and feedback or feed-forward. Specification compute the sensitivity of the stability and control derivatives that which is a form of non-dimensional (to fairly compare different SIMPLI-FLYD calculates to the design parameters it imports types of derivatives) and reference area “ weighted ” version of the from NDARC. The design parameters that most influenced the sensitivity value (to allow equal comparison between key stability and control derivatives known to influence the components). The color in these plots does not represent any value handling qualities scenarios selected, could be then identified for and is merely intended to help to differentiate between each row investigation in the full NDARC/SIMPLI-FLYD analyses. of the plot data. For reasons of clarity, only the key derivatives known to influence those dynamics relevant to the pitch or yaw Both the derivative sensitivity study and the NDARC/SIMPLI- axis HQ Design Margins are presented.

FLYD coupled process used two example vehicle test cases; a tiltrotor aircraft similar in size and design to XV-15, with the The main premise of the analysis is to focus on the relative values focus being the forward flight pitch axis, and a single-main rotor rather than the absolute values computed. The figure show the helicopter (similar in size and weight to a UH-60a), with a focus sensitivities for Rotor1 (Rotor 2 results are identical in this on the hover yaw axis characteristics. forward flight condition and are thus omitted), the wing, fuselage and, “tail” components ( 1 being the horizontal stabilizer and 2 and 3 the two end plates of the H-tail configuration).

Results Intuitively, the design parameters that influence the most are the Stability and Control Derivative Sensitivity Analysis horizontal tail parameters such as the tail area, X-location (loc(1)), The results of sensitivity analysis of the bare-airframe stability lift curve slope (dclca), elevator size (Scont_S) and control flap and control derivatives to a subset of the design parameters lift effectiveness (Lf). The only other parameter that approaches imported by SIMPLI-FLYD are presented first. The results using the same level of sensitivity is the wing X-location (loc(1) on the the NDARC XV-15-like tiltrotor as the source of input to wing subplot).

SIMPLI-FLYD are shown in Figure 3 . Each sub-plot is a 3-axis bar chart – one axis is for the stability and control derivatives Figure 4 shows the same analysis for the single-main rotor names, namely key pitch moment (M) and vertical force (Z) helicopter configuration at hover for the key yaw axis bare- derivatives with respect to control inputs, elevator and cyclic and airframe stability and control derivatives. At hover, the key longitudinal axis states: pitch rate, (q) and vertical velocity, derivatives are most sensitive to the design parameters defining (w) (which reflects the angle of attack response). The derivative the tail rotor (rotor 2) that the derivatives are most sensitive to, naming convention is the force/moment separated by an though the main rotor (rotor 1) equivalent hinge offset parameter underscore from the state/control e.g. M_elevator or Z_w. The (“e”) approaches a similar order of influence. The most sensitive second axis is for the components design parameters being parameter of the tail rotor is the radius, followed by the tail rotor perturbed (each plot is for a single component) – only those that longitudinal location (hub_loc(1)), and blade chord (chord) .The had a non-zero effect are shown. For the purposes of this paper, it results are mostly intuitive although the inclusion of some of the is not necessary to specifically identify all the listed design lateral-directional coupling terms show relationships that are less parameters perturbed, however parameters of particular intuitive, such as the effect of the tail rotor radius and Lock significance are identified in the following section. than The third number on the overall vehicle roll due to yaw derivative, L .

r axis is the value of the stability and control derivative “sensitivity” Rotors Wing Fuselage Horizontal Tail Vertical tails Figure 3 Sensitivity of component pitch stability and control derivatives to design parameters imported from NDARC model (tiltrotor, airplane mode, 160 kts) Main Rotor Tail Rotor Figure 4 Sensitivity of component lateral-directional stability and control derivatives to design parameters imported from NDARC model (single-main rotor, hover) flap actuator rate-limit. Three flight conditions – 160 kts, 230 kts NDARC/SIMPLI-FLYD Coupled Analysis and 300 kts are evaluated for airplane mode. All the values varied The results of applying variations to design parameters in the full for the design parameters are listed in the following bullets coupled NDARC/SIMPLI-FLYD analyses are presented in the (*initial nominal value): following section. The sensitivity analysis of the previous section guided the selection of a subset of design variables to use. The  Horizontal tail Area: 25.125, 40.2, 50.25*, 60.3, 75.375 [ft ] same two NDARC models, an XV-15-like tiltrotor, and a single  Horizontal tail flap ratio: 0.0647, 0.1656, 0.2587*, 0.36, 0.45 main rotor helicopter were used as the test example vehicles.

[nd] These results build upon the work in [7] in two aspects: 1) for each  Horizontal tail flap actuator rate limit: 10, 20*, 30 [deg/s] change of variables, NDARC now performs a sizing task (the As described earlier, the analysis computes a design margin for previous analysis varied the design which affected the weight but both the feedback (stabilization) and feed-forward (response) did not re-size), and 2), the analyses use multidimensional components of the pitch axis HQs. The figures use 3-D Matlab parameter changes to design variables rather than single “slice” plots that provide a color weighted “cloud” of data for the parameter sweeps and thus are able to demonstrate coupling 3 design parameters. The color indicates the design margin value, effects of design parameters on the HQs.

ranging from red for -200% under-design margin, to deep blue for The NDARC sizing task determines the dimensions, power, and the +200% over-design margin. The plots are a space-efficient weight of a rotorcraft that can perform a nominal design mission method to presenting data for up to 3/4-dimensions. The printed [8] . The aircraft size is characterized by parameters such as design versions are somewhat limited in that the slices can only be gross weight, weight empty, rotor radius, and engine power viewed from a fixed perspective whereas in the Matlab software available. NDARC calculates the size and weight of certain the user can manipulate the viewing angle to inspect the data from specified aircraft components while factoring in the weight of any vantage. Nevertheless, the main intention is to present the other components with a predetermined size (such as the broader trends, which in these cases are adequately covered in a parameters being varied in this analysis). The exceptions to this more compact manner by this form of data presentation.

currently are the actuator parameters which are uncoupled to The first result to highlight is the trend with flight speed, where NDARC and only currently affect the handling qualities and not the sensitivity of the design margins to variations in the design the design weight and power. In all the results in this paper, the parameters reduces with greater airspeed. Here, the higher change of the moments of the inertia of the vehicle with respect dynamic pressure confers increased bare-airframe stability, to the design variations is not directly modeled. Instead, inertia damping and control power improving the ability of the aircraft to changes are currently represented via the weight change and the meet the stabilization (feedback) and control response (feed- use of fixed radii of gyration which SIMPLI-FLYD inherits from forward) HQ specifications respectively. As the speed reduces, NDARC. After NDARC has completed its task, the aircraft design and particularly at 160 kts (which is approximately 1.5x the stall data is input to SIMPLI-FLYD. The final outputs of the SIMPLI- speed in airplane mode) the boundaries between over-design and FLYD analysis are the HQ design margins, as calculated by under design in the pitch HQs become more apparent. Intuitively, CONDUIT. Conceptually, it appeared reasonable to use the the worst pitch axis feed-forward HQs (under design) are for the CONDUIT HQ design margins as the overall HQ analysis metrics smallest tail, flap and lowest actuator rate limit. The trend for the for two reasons: feedback specifications is similar with the exception being that 1) The design margins offered a mechanism by which multiple the rate limit is generally unable to impact the under-design feedback and feed-forward HQ specifications are reduced to margin cases, such as those at the lower tail areas at 160 kts (some two HQ design margins per axis – these respectively coupling effect is seen at the highest tail flap ratio and highest rate represent the aircraft’s overall flight dynamic limit). Note that throughout these analyses, to save computational stabilization/disturbance rejection and control response time by preventing the CONDUIT optimization continuing to characteristics for each axis.

very high positive design margins, a limit was applied that 2) The design margins are a non-dimensional metric (% prevented further optimization if a +60% design margin was over/under design) based on the worst case HQ specification.

reached.

Both of these factors offered a mechanism that avoided a situation In Figure 5 (a) shows that another region of reduced design margin where each analysis case presented a large array of multiple emerges indicating that the tail can be “too big” from a HQ specifications with differing units and meanings. Instead, the perspective. In these cases, the large tail with small control design margins present more holistic metrics of the HQ surface, at low rate limit, has under-design for the feed-forward “goodness” that are more convenient for integration with an pitch HQs. The aircraft is likely over-damped or too stable and the overall MDAO process and for presentation to the design aircraft is unable to meet certain control response specifications.

engineer.

In the tiltrotor examples, varying the tail flap actuator rate limit Figure 5 (a) to (c) show the tiltrotor pitch axis handling qualities had different effects on the outcome of the feed-forward and (HQ) design margins (DM) for a sweep of three design feedback design margins. For the feed-forward design margins, parameters: horizontal tail area, tail flap control area ratio and tail rate limit had a graduated effect where increased rate-limit was able to affect the design margin achievable for varying tail area The relationship of the empty weight with respect to the three and flap area ratios. The feed-forward specifications have a design variables is shown in Figure 5 (d). The tail area dominant number of time-domain specifications which are sensitive to non- factor affecting the aircraft weight. The tail flap area ratio has a linearity such as rate-limiting. Conversely, the feedback design relatively weak effect. The actuator characteristics have no effect margins had weak sensitivity to rate limiting (for the range of as NDARC does not know anything about those. Currently the values evaluated), probably because the feedback specifications actuator properties are only an input to SIMPLI-FLYD and thus are predominately linear system analyses and the non-linear effect can only influence the HQ DMs. T he “cost” of changing the of rate-limiting essentially plays no part in determining the margin actuator performance on the design, such as in terms of weight, is to these specifications. The OLOP (Open Loop Onset Point) not yet modelled and is a planned future development.

specification [14] is included to capture the effect of rate-limiting Figure 6 is a second, comparative set of sweeps for the tiltrotor at but closer inspection of these cases discovered that the OLOP the 160 kts airplane condition. In these cases however, the specification was not determining the under-design, and in fact actuator rate limit is fixed at the nominal 20 deg/s and the third eigenvalue stability and crossover frequency were the limiting axis (vertical) of design parameter variation is now the horizontal specifications. OLOP was primarily developed to predict tail longitudinal position, expressed in non-dimensional terms, handling qualities issues due to pilot input induced rate limiting X/L (X-location divided by the reference length of rotor radius), and there are clear guidelines to use the maximum pilot input as for values of 1.4317, 1.7896*, 2.1475 (*nominal). Sensitivity of the disturbance input for the OLOP test. However, OLOP has also the design margins in (a) is observed for all three variables.

been included as a feed-back HQ specification in SIMPLI- Vehicle empty weight in (b) is mostly affected by the tail area and FLYD – an extrapolation of its original design. As such, the X-location. Comparing the feedback and feed-forward design concept of the disturbance input is no longer a pilot input but some margins it can be seen a trade-off exists between an optimal external disturbance (gust etc.) for a which nominal value of 5ft/s configuration for stability (feedback) and response (feed- in the vertical velocity (w) was selected. This value is relatively forward). The feed-forward tends to prefer a larger control surface small and thus OLOP limits are not approached even for the on a smaller tail for the intermediate tail location and the feedback smallest tails which should require the greatest feedback prefers a larger tail, at the largest X-location and is only weakly stabilization due to the reduction in bare-airframe stability. Hence sensitive to the flap size (the very smallest tail is improved by OLOP, the only feedback specification that would be sensitive to having a bigger fraction of control surface).

actuator rate limiting, does not act as a limiting specification and no relationship is observed with rate limit for the overall feedback design margin.

(a) Pitch axis HQ design margins v design parameters, 300kts (b) Pitch axis design HQ margins v design parameters, 230kts (c) Pitch axis HQ design margins v design parameters, 160kts (d) Empty weight v design parameters Figure 5 Tiltrotor Feed-forward/Feedback Pitch Axis HQ design margins and empty weight for variations in tail area, tail flap area ratio, tail flap actuator rate limit, at various speeds (airplane mode, sea level) (a) Pitch axis HQ design margins v design parameters, 160kts (b) Empty weight v design parameters Figure 6 Tiltrotor Feed-forward/Feedback Pitch Axis HQ design margins and empty weight for variations in tail area, tail flap area ratio, tail non- dimensional longitudinal position (X/L), at 160kts (airplane mode, sea level) The results of the NDARC/SIMPLI-FLYD coupling applied to vehicle in this flight condition (such as the empennage becoming the single main rotor example, with the focus on the yaw axis effective in forward flight) but also due to different HQ handling qualities are shown in Figure 7 . The design margins are specification requirements being applied for this forward flight shown for a sweep of the tail rotor size (a coupled increment in condition, which also has a secondary effect in that the CONDUIT rotor radius at constant solidity and tip speed), non-dimensional control system gains are optimized differently. The design margin longitudinal tail rotor location (X/L) and tail rotor actuator concept is useful here, as many of HQ specifications are not bandwidth. The range of parameters was chosen to investigate common between hover and forward flight, and the concept of the whether an optimum size/position existed for rotors smaller than margins provides a constant metric of HQ performance across the the nominal 6ft radius. Hence, the rotor size variations ranged changing requirements.

from 6 ft to 3.6 ft radius (a 40% reduction) and a longitudinal The outcome of only achieving a positive design margin in a small position, X/L, from 1.3 to 1.82 (a 50% increase from nominal).

portion of the parameter space led to a second sweep for the single Note that there is a region of the plot for radii below 5ft and the main rotor configuration, the results are shown in Figure 8 . The smallest X-locations that is empty. Here, NDARC was unable to plots show the yaw axis design margins but for a slightly different converge on a sizing or trim solution and thus these cases were set of design parameter variations. The variation in the tail rotor discarded as invalid.

size is retained but this time larger tail rotors are examined.

For the feedback design margin, there is very little sensitivity to However, to ensure design geometry “consistency”, instead of the design parameter variations, and almost all cases reached the specifying the location of the tail rotor directly, a tail rotor upper limit of +60% design margin. The feed-forward margins clearance (modifying its longitudinal position with respect to the exhibited much greater sensitivity to the design parameter main rotor) was varied. Specifying the clearance was necessary to variations. The design margin improved with increased radius and ensure that the larger tail rotors did not impinge the main rotor X-location but only reached +10% positive design margin case disk – and was a more convenient and efficient method than with largest rotor, longest tail rotor location, and greatest actuator manually recalculating the rotor positions, especially when the bandwidth value. The design margins at forward speed are shown NDARC sizing automatically configures the main rotor disk size.

in the adjacent Figure 7 (b). Essentially the same trends as hover Finally, and simply for comparative purposes, instead of actuator are reflected but the region of positive design margin is enlarged. bandwidth, the tail rotor actuator rate limit was varied.

This can be partly attributed to different performance of the (a) yaw axis HQ design margins v design parameters, 0kts (b) yaw axis HQ design margins v design parameters, 80kts Figure 7 Single main rotor helicopter feed-forward/feedback yaw axis HQ design margins v tail rotor radius, tail rotor non-dimensional longitudinal position (X/L), tail rotor collective actuator bandwidth at various speeds, sea level (a) yaw axis HQ design margins v design parameters, 0kts (b) yaw axis HQ design margins v design parameters, 80kts Figure 8 Single main rotor helicopter feed-forward/feedback yaw axis HQ design margins v tail rotor radius, tail rotor non-dimensional longitudinal clearance, tail rotor collective actuator rate limit at various speeds, sea level (a) tail rotor radius, longitudinal position (X/L), actuator bandwidth (b) tail rotor radius, longitudinal clearance, actuator rate limit Figure 9 Single main rotor helicopter empty weights v design parameters for two different parameter sweeps At hover in Figure 8 (a), a larger region of cases now equal or to analyze the assumptions and constraints made in the conceptual exceed the 0% design margin in feed-forward (feedback is at the design phase.

+60% limit in all cases). Actuator rate limiting is an influential The handling qualities characteristics and subsequent design factor, with its increase enabling a greater proportion of the rotor margins vary across the flight envelope due to both the changing size/clearance cases to meet the 0% or better yaw axis design performance of the vehicle and the specifications applied. For the margin. The change to the forward flight speed condition (b) again tiltrotor, using the pitch axis in airplane mode example, greater enlarges the region of cases with positive design margins.

criticality emerges mostly at low speed, but not exclusively, with The empty weight for the two sweep sets for the single main rotor design margin degradations occurring at the higher speed for other configuration is compared in Figure 9 . Figure 9 (a) shows that for reasons. This raises questions about how many handling qualities the tail rotor size reduction the vehicle empty weight actually flight condition analyses should be included. Clearly, if the increases. The reduced size tail rotors are increasingly inefficient vehicle can hover and fly at some forward speed a minimum of and require greater power, which increases weight for a number two should be assessed, but more could be incorporated.

of the other components of the vehicle. In the second set of sweep NDARC’s sizing typically uses 3 -5 critical flight conditions and cases in Figure 9 (b) a minimum weight region is observed a similar approach could be taken for the handling qualities between the smallest and largest tail rotors and toward the larger analysis with the approach of some baseline characteristics at a tail rotor clearance positions. couple of nominal flight conditions being “ensured”.

Alternatively, an approach that perhaps brackets the operational envelope with min/max airspeed, min/max altitude plus certain Discussion of SIMPLI-FLYD in conceptual design configuration changes (e.g. rotor tilt) might be required or and future developments considered prudent. The results thus far tend toward a The current paradigm for the use of SIMPLI-FLYD in an overall recommendation for the latter but it may be dependent on the level vehicle conceptual design process is that the computed HQ design of effort required/appropriate for handling qualities in conceptual margins would set the constraints for an optimization i.e. some design.

minimum is required, 0% or perhaps a positive margin while other Currently, there is no requirement for how much computational design objectives are maximized or minimized, such as the empty weight. Defining appropriate design margin levels more time should be allowed for the handling qualities analysis in conceptual design but if the objective is to explore very large conclusively for conceptual design will require the analysis of parameter spaces rapidly, the 15-20 minute time per flight SIMPLI-FLYD in an analysis with many more multidisciplinary design constraints. Additionally, an assessment of the HQ condition of the current approach is likely a limiting factor. For comparison, NDARC typically completes its sizing task in times constraints choices made in SIMPLI-FLYD will require retrospective analysis from later in the design lifecycle – i.e. using of the order of seconds. Possible solutions might be a faster running version of SIMPLI-FLYD by streamlining its tasks or higher fidelity tools, models and data available in detailed design incorporating an analogous but alternative approach to the full Thus far, the only NDARC/SIMPLI-FLYD sub-stage sensitivity CONDUIT optimization. Alternatively, a framework that study that has been carried out is the sensitivity of the stability and analyzes the handling qualities in a less tightly-coupled approach control derivatives to the SIMPLI-FLYD internal parameters. The might allow for the current computational times. method offers useful insight into which design parameters are the most important to determining the key stability derivatives for the During the development of derivative sensitivity analysis, it handling qualities. Admittedly in the cases examined, the aircraft became apparent that the process of importing the NDARC designs are well understood (a main wing- aft tail “airplane” and parameters to SIMPLI-FLYD for the calculation of the stability single main rotor/ tail rotor configuration), so any experienced and control derivatives is one of a series of steps in a process of flight dynamics engineer would have been able to predict the translation and reduction of parameters, as outlined in Figure 10 .

majority of the outcomes of the sensitivity sweeps. There are, At the beginning is an aircraft design in NDARC, defined by however, a number of limitations of the approach which should hundreds, if not thousands, of parameters defining the geometry, be highlighted. The results are likely configuration-specific, and aerodynamics, weights and power. Of these parameters, on the they are a further simplification, as coupling effects between order of 150-250 are imported by SIMPLI-FLYD to define the design parameters on the models are neglected as per classical flight dynamics models with a few additional parameters to define linear theory. Furthermore, the technique does not consider the the control actuator characteristics. The flight dynamics (if effect of performing a re-optimization of the CONDUIT control restricted to rigid-body, 6-DoF) are essentially described by 36 system gains after a design change. The re-optimization would stability derivatives and a minimum of 24 control derivatives.

attempt to rebalance the gains to maximize HQ performance and These, in conjunction with the optimized control system gain thus any determined sensitivity is not at the “optimal control parameters, determine the approximately 10-20 HQ (feedback d esign” point, which may lead to further inaccuracies when and feedforward) specifications per axis. The most limiting compared to the full process.

specifications then determine the overall HQ design margins which number up to eight, two parameters per axis, for feedback Nevertheless, the key advantage is that the calculations for the and feedforward. Examining the steps of the overall stability and control derivative and the other sub-stage sensitivity NDARC/SIMPLI-FLYD coupled process it appeared there might analyses are much faster to calculate. In fact, the whole chain of be utility in studying the sensitivity between the parameters of all sensitivities are faster to calculate than the full process, and for of the constituent steps. The basis of such an approach would be many more parameters, due to the fact that no re-optimization of to allow a piece-wise “chain of sensitivity” to be identified instead the control system would be carried out. It is the CONDUIT of treating the whole process as a “black box” . For example, a user optimization that is the major computational cost to the current may trace the sensitivity of a subset of specifications to a subset process. If a simplified surrogate analysis could be developed, a of derivatives which in turn are only sensitive to a particular framework incorporating frequent calls of the fast, simplified subset of input parameters. analysis alongside the more computationally expensive full CONDUIT optimization might lead to an overall more computationally efficient approach for conceptual design.

NDARC 100’s - 1000’ s CONDUIT Optimization Fligh t Dynamics modeling parameters SIMPLI-FLYD approx. 150-250 parameters Derivatives HQ specs approx. 60 approx. 40 parameters parameters Stability and HQ DMs control approx. 4-8 derivative parameters sensitivity Figure 10 Schematic showing the reduction of the number of parameters from NDARC model to SIMPLI-FLYD handling qualities design margins There is clearly a tradeoff between computational resources and qualities design margins. Cost models for the actuators in terms the number of analyses and the rigor they contain. This tradeoff, of weight, size, power and cost as function of their performance coupled with the number of potential variables that could be characteristics must be incorporated into this analysis so their involved also highlights challenges in how handling qualities selection can be properly accounted for if the handling qualities should be managed in an optimization. A human in the loop would are to be considered in a conceptual design.

seemingly be overwhelmed by the number of potential variables that could be used to influence the handling qualities although Summary and Conclusions other design constraints (with greater priority) may rapidly reduce This paper has reported the continued exploration of the recently the number of variables that the handling qualities aspects may developed SIMPLI-FLYD toolset. The use of Python-based reasonably be allowed to influence. Solutions to this challenge scripting has enabled the integration of NDARC and SIMPLI- have not yet been identified but techniques like the stability and FLYD in fully automated analyses for design parameter variations control sensitivity analysis could also form the basis of a tool to that has accelerated the learning of how a handling qualities guide an engineer using SIMPLI-FLYD, either to help users that analysis interacts with conceptual design models. Processes have do not possess the relevant handling qualities knowledge or to been demonstrated that can calculate design margins with respect inform when a configuration is non-classical and where usual to handling qualities specification criteria while also evaluating rules-of-thumb cannot be relied upon.

the vehicle design weight and other design metrics. Also, secondary techniques like the bare-airframe stability and control Another important aspect of the current SIMPLI-FLYD capability derivative sensitivity analysis offer insight to the inner-workings to highlight is the method of representing inertia changes via the of a complex process. They may also offer pathways to mitigate weight change combined with radii of gyration. This method is the computational cost of running full CONDUIT optimization so likely to be only satisfactory for gross changes in design weight, frequently if faster run times become a requirement.

and as long as the vehicle configuration does not vary drastically, and probably does not possess the level of sensitivity/fidelity for Through using the tools in a coupled approach to examine the design changes such as those being applied in these examples.

different vehicle types while varying a mix of design parameters A future development that will improve this is the anticipated and flight conditions, and evaluating different handling qualities integration with the U.S. Army ADD developed “ALPINE” problems, the following items are highlighted: (Automated Layout with a Python Integrated NDARC  The calculated handling qualities design margins vary across Environment) tool ref [15] . ALPINE provides a capability to the flight envelope due to both changing flight dynamic and generate a 3-D geometry in OpenVSP (Open Vehicle Sketch Pad control characteristics and the handling qualities [16] ) from NDARC output and thus enables a calculation of the requirements specifications applied.

mass and inertia properties using Open VSP’s mass properties  The current SIMPLI-FLYD analysis process for a single functions which are more sensitive to arbitrary configuration flight condition, compared to other conceptual design tools, changes.

is relatively computationally expensive. This cost is either Indeed, a 3-D geometry engine approach to managing the design likely to impose constraints on how to deploy the tool in an configuration such as that provided by OpenVSP would be also overall design process or will require further evaluation of advantageous when manipulating a design’s geometry (manually what HQ aspects should be incorporated in attempt to gain or automatically). This became apparent during the work in this computational efficiencies.

paper, as even for these relatively simple cases, such as when  The challenge of ensuring a consistent vehicle geometry moving a tail, or changing its size, so that the fuselage adjusts to when performing the analysis in this paper have highlighted support the tail, ensuring geometry consistency is difficult to the advantages for representing and maintaining a consistent manage (NDARC has features that addresses some aspects but is geometry while adjusting a design such through using a 3-D not comprehensive). Geometry consistency is not only important geometry engine such as OpenVSP.

to ensure that the design is valid structurally (i.e., components are  The actuator performance characteristics influence the attached), but also for capturing design cross-couplings from a handling qualities design margins strongly and therefore handling qualities perspective. For example, moving the including their cost (in terms of weight, power requirements horizontal tail for better handling qualities/weight savings may etc.) is critical if handling qualities are to be included in an impact the vertical tail location, depending on the configuration, overall design optimization.

and thus may affect the lateral-directional characteristics. This experience in ensuring geometry consistency correlates with the Acknowledgements considerations of Ref [17] , which also places great value in The authors would particularly like to acknowledge Larry Meyn, integrating a 3-D geometry engine at the heart of any future NASA Ames Research Center, for his support in the development conceptual design environments.

of the Python scripts and the provision of the “ rcotools ” library Finally, the results in this paper reinforce the conclusions of prior that supported their development. Andrew Gallaher, U.S. Army results using SIMPLI-FLYD Ref [7] that the actuator ADD, Moffett Field is acknowledged for his OpenVSP insight characteristics play an important role in determining the handling and Python development advice.

[15] Perry, T. " ALPINE: Automated Layout with a Python References Integrated NDARC Environment, OpenVSP Workshop [1] Gorton, S.A., Lopez, I., and Theodore, C.R, “NASA 2016, NASA Ames Research Center, Moffett Field, CA, Technology for Next Generation Vertical Lift Vehicles”, th th USA, 25 Aug 2016 [PDF File] AIAA SciTech, 56 AIAA/ASCE/AHS/ASC Structures, https://nari.arc.nasa.gov/sites/default/files/attachments/34A Structural Dynamics and Materials Conf., Kissimmee, FL, th LPINE%20%28002%29_Perry.pdf USA, 5-9 January, 2015.

[16] Gloudemans, J. R., Davis, P. C., and Gelhausen, P. A., “A [2] Morris, C. C., Sultan, C., Allison, D. L., Schetz, J. A., & rapid geometry modeler for conceptual aircraft”, 34th Kapania, R. K., “Towards Flying Qualities Constraints in the Aerospace Sciences Meeting and Exhibit, AIAA-1996-52, Multidisciplinary Design Optimization of a Supersonic Jan. 15-18 1996.

Tailless Aircraft.”, 12th AIAA Aviation Technology, [17] Johnson, W., and Sinsay, J.D., “Rotorcraft Conceptua l Integration, and Operations (ATIO) Conference and 14th nd Design Environment”, 2 International Forum on Rotorcraft AIAA/ISSM, Indianapolis, IN, USA, 17-19 Sep 2012.

Multidisciplinary Technology, Seoul, Korea, Oct 19-20, [3] Raymer, D.P., “ Aircraft Design: A Conceptual Approach ” , th 2009.

American Institute of Aeronautics and Astronautics Inc, 5 Ed., 2006.

[4] Padfield, G. D. “Rotorcra ft Handling Qualities Engineering; managing the tension between safety and performance”.

Journal of the American Helicopter Society, Vol. 58 (no. 1).

pp 1 – 28, January, 2013.

[5] Andrews, Harold, Technical Evaluation Report on the Flight Mechanics Panel Symposium on Flying Qualities, AGARD- AR-311, April 1992.

[6] Johnson, W., “NDARC — NASA Design and Analysis of Rotorcraft, Theoretical Basis and Architecture”, American Helicopter Society Aeromechanics Specialists’ Conference Proceedings, San Francisco, CA, January 20-22, 2010.

[7] Lawrence, B., Berger, T., Theodore, C.R., Tischler, M.B., Tobias, E.L., Elmore, J., and Gallaher, A., “ Integrating Flight Dynamics & Control Analysis and Simulation in Rotorcraft nd Conceptual Design ”, 72 American Helicopter Society Annual Forum, West Palm Beach, FL, USA, May 17-19, 2016.

[8] Johnson, W. “NDARC, NASA Design and Analysis of Rotorcraft.” NASA TP 2009 -215402, 2009.

[9] Tischler, M. B., Colbourne, J., Morel, M., Biezad, D., Cheung, K., Levine, W., and Moldoveanu, V., “A Multidisciplinary Flight Control Development Environment and Its Application to a Helicopter,” IEEE Control Systems Magazine, Vol. 19, No. 4, pg. 22-33, August, 1999.

[10] Tischler, M. B., Remple, R. K., Aircraft and Rotorcraft System Identification: Engineering Methods and Flight Test Examples , 2nd Edition, AIAA, 2012, pp 332-333.

[11] Anon., "Handling Qualities Requirements for Military Rotorcraft", Aeronautical Design Standard-33 (ADS-33E- PRF), US Army Aviation and Missile Command, March 21, 2000.

[12] Anon., “Flying Qualities of Piloted Aircraft,” MIL -STD- 1797B, Department of Defense Interface Standard, February, 2006.

[13] Forrester, I.J., Sóbester, A. Keane, A.J., Engineering Design via Surrogate Modelling: A Practical Guide , John Wiley & Sons, 2008.

[14] Duda, H., “Prediction of Pilot -in-the-Loop Oscillations due to Rate Saturation”, Journal of Guidance, Navigation, and Control, Vol. 20, No. 3, May-June 1997.