3.1 Scenario development

The simulated scenario comprises a tanker, receiver and a hose-and-drogue refuelling system. The tanker aircraft model is referred to as Future Military Transport Aircraft (FMTA). The helicopter model is representing a generic heavy lift cargo helicopter, and is referred to as Generic Helicopter Model (GHM). The hose, drogue and probe elements are modelled as a coupled multi-body system used in Ref. Reference Fezans and Jann16. This refuelling model is not yet included in the real-time HAAR simulation but utilised to determine manoeuver relevant positions. Hence, the positions of the hose and drogue are fixed relative to the tanker, which is why both components are visualised without any natural dynamics.

Prior to the development of the simulation scenario shown in Fig. 1, current operational standards were analysed, and workshops were conducted with pilots who had previous experience performing HAAR [Reference Schmidt and Jones6]. In the simulation the HAAR flight condition for FMTA and GHM was set to a velocity of 115Kn CAS (= 65.71m/s true airspeed) at 7,000ft (2,134m) altitude AMSL. The FMTA’s high-lift flaps were in flap position 3 (i.e. 30 deg deflection). To evaluate the FMTA’s flight state, it is trimmed in forces by changing the angle-of-attack and the propeller’s pitch angle. This led to an angle-of-attack of 11.7 deg. The propellers were modelled as actuator disks, to offer a quasi-steady state Computational Fluid Dynamics (CFD) solution of the rotating propellers. In combination with the propeller’s speed of 860rpm and the pitch angle the propeller’s downstream can be modelled depending on the propeller diameter. For this flight condition, a CFD solution was determined in Ref. Reference Löchert, Schmidt, Jann and Jones19 and implemented within the simulation scenario as a time invariant wake.

The tanker’s wake flow was simulated using a scale resolved CFD method the improved delayed detached eddy simulation (IDDES). The wake flow of the FMTA and its potential impact to HAAR is described in Ref. Reference Löchert, Schmidt, Jann and Jones19. Figure 3 shows the wake flow topology of the FMTA under the HAAR flight condition. The figure shows the vorticity orthogonal to the y-z plane and referred to $\pi $ on slices through constant x-values from $x = 20{\rm{m}}$ to $x = 100{\rm{m}}$ with $\Delta x = 10{\rm{m}}$ . For the HAAR simulation scenario the tip vortex and the flap vortex highlighted in Fig. 3 are of crucial importance.

Figure 3. FMTA’s wake flow topology under HAAR flight condition [Reference Löchert, Schmidt, Jann and Jones19].

In order to approximate the position of the low-speed drogue relative to the tanker, a refuelling system model (in Matlab Simulink) is applied for both the 80ft and the 120ft hose length. A more detailed description of the multi-body refuelling system including the parameter sets for both hose lengths can be found in Ref. Reference Löchert, Schmidt, Jann and Jones19. The influence of the tanker wake on the hose and drogue is also considered for the estimation of the drogue’s position.

At the given conditions, the simulation results in a drogue position [−24.2m, −3.0m, 6.3m], relative to the wing pod’s hose outlet for the 80ft case. The drogue’s relative position in the 120ft case is determined with [−35.0m, −3.0m, 10.8m]. Since the coordinates are inherent to the tanker’s body-fixed frame both hose lengths lead to a position behind, below and on the left of the wing pod. Figure 4 shows the hose’s shapes and the drogue’s positions with respect to the first joint at the hose outlet. Since the refuelling system model had not yet been implemented into the real-time helicopter simulation for these initial evaluations, the hose’s and drogue’s positions were fixed relative to the FMTA during piloted simulation.

Figure 4. Relative position of low-speed drogue (115Kn CAS; 7,000ft AMSL).

Within the AVES test facility, the tanker FMTA was simulated and displayed in the external visual environment. The aircraft motion was simulated using the Vehicle Control traffic server described in Ref. Reference Maibach, Jones and Strbac20. The tanker was configured with a constant airspeed and heading. Through discussions with pilots and through NATO guidance documents [Reference Schmidt and Jones6, 18], it is known that cases during turns and manoeuvering flight are of interest for simulation, particularly in failure or extreme situations. These cases will be explored at a later date within the project F(AI)²R but were beyond the scope of these initial evaluations. Figure 5 shows the view from the AVES cockpit during the approach to formation.

Figure 5. Cockpit view of the HAAR scenario.

In Fig. 6 the receiver helicopter is visualised maintaining both pre-contact positions for the 80ft and the 120ft hose within the tanker’s wake flow. As shown, for the 80ft hose, the strongest interference occurs at the rotor. For the 120ft hose, where the x- and z-position of the helicopter are offset, the strong region of turbulence occurs above the main rotor. These results reflect aspects reported from the initial Airbus A400M flight tests [Reference Osborn8], mentioned earlier in this paper.

Figure 6. Helicopter’s pre-contact position in tanker wake flow for the 80ft (top) and 120ft hose (bottom).

3.2 GHM development

For the real-time helicopter simulation within AVES, DLR’s 2Simulate HeliWorX modeling suite is used, featuring a non-linear implementation of a helicopter model [Reference Gotschlich and Jones21]. The HeliWorX suite originates from a Bo 105 model that applies blade element theory and was further developed to represent the Active Control Technology/Flying Helicopter Simulator (ACT/FHS). The ACT/FHS is a full-authority fly-by wire/light version of the EC135 helicopter and used for research at DLR. In this configuration, HeliWorX was extended with a real-time interference model to simulate local aerodynamic effects, like the air wake of wind turbines, ships or aircraft. Based on this extension of the HeliWorX suite, a Generic Helicopter Model (GHM) was developed and configured to represent a heavy lift cargo helicopter (e.g. CH-53). The GHM will be utilised for HAAR studies in F(AI)²R, and some of its key configuration parameters are listed in Table 1.

After completing the configuration, a three-axes SAS was manually tuned to provide a basic stability augmentation system. For this configuration, quantitative handling qualities levels were examined in terms of the bandwidth and attitude quickness criteria at the relevant flight condition. Gain and phase bandwidth values for the pitch, roll and yaw axes are summarised in Table 2. Results with respect to boundaries for target tracking and acquisition tasks are shown in Fig. 7. It is shown that Level 1 is achieved on all three axes for the bandwidth criterion.

In the Aeronautical Design Standard (ADS) 33E-PRF [22], only the roll attitude quickness is required to be evaluated in forward flight. Lateral step inputs for both left and right inceptor deflection, combined with the responses of the GHM were recorded. The resulting attitude quickness of both directions is shown in the lower right plot of Fig. 7. As it can be seen the minimum attitude changes and the peak roll rate to peak attitude change ratio lead to Level 3 handling qualities in both cases.

Table 1. Key configuration parameters of the GHM

Table 2. Gain and phase bandwidth values of the GHM

Figure 7. GHM handling qualities according to ADS 33E-PRF [22] pitch, roll, yaw bandwidth and roll attitude quickness criterion at 115Kn forward flight.

3.3 Flow field interference

The previously mentioned real-time implementation of the time invariant flow field relies on the approach outlined in Ref. Reference Maibach, Jones and Strbac20. Stored within a three-dimensional lookup table, the flow field interacts with the helicopter at defined Airload Computation Points (ACPs). Using the location of these points relative to the origin of the flow field, the local wind velocities are linearly interpolated. The ACPs are located at the rotor blades comprising ten ACPs per blade, as seen in Fig. 8. For this initial evaluation, ACPs on the fuselage, horizontal and vertical stabiliser were neglected. This was considered acceptable for first tests, but in future evaluations the additional ACPs will also be included. From previous work published in Refs. Reference Maibach, Jones and Strbac20, Reference Strbac, Martini, Greiwe, Hoffmann and Jones23 and Reference Watson, Owen and White24, the amount of ACPs on the main rotor is considered sufficient for piloted simulation. The approach uses one-way coupling only (wind table interference on helicopter), which is also considered sufficient for piloted real-time simulation. The approach has been validated through comparison between ACP output and CFD results in Ref. Reference Strbac, Martini, Greiwe, Hoffmann and Jones23.

Figure 8. Distribution of ACPs over the GHM model.

The original HAAR CFD flow field was cut to a size of +260m in length, ±80m width and ±40m height originating at the FMTA’s nose. Based on the experience of Ref. Reference Fezans and Jann16, the grid had been transformed into a structured equidistant cubic grid with a longitudinal grid discretisation of 1 and 0.5m in lateral as well as vertical direction. Since free airstream, global and local (flow field) winds are separately implemented and superimposed for force calculation later on, airstream was subtracted from the CFD solution.

The flow field is obtained in tanker body fixed coordinates from the CFD solution. To account for the tanker’s angle-of-attack, a real-time rotational transformation was added to the interference model presented in Ref. Reference Maibach, Jones and Strbac20. Instead of rotating the entire flow field once prior to the simulation, only ACP relevant vectors are transformed [Reference Löchert, Schmidt, Jann and Jones19]. This way, the interference model is prepared for rotational movements of the tanker in future work, even for larger flow fields. Yet the tanker is realised as a simple point of mass, in straight and level flight at 115Kn. The flow field origin matches the point of mass and sticks to its translational movements.

The flow field velocities at the rotor hub reference ACP obtained from the trim in the real-time simulation were compared to the velocities at the ideal pre-contact position within the original flow field of the CFD solution. Using the transformation mechanism at a static rotation around the y-axis in the real-time simulation, nearly the same velocity components could be observed in Ref. Reference Löchert, Schmidt, Jann and Jones19. Therefore, the implementation of the transformation is found to be suitable to account for the tankers angle-of-attack.

4.1 Development of HAAR MTEs

Currently, the state-of-the-art design standard, commonly used for rotorcraft development, is ADS33E-PRF [22]. Although forward flight characteristics and ‘mission tasks’ are included in ADS33 it is focused on low-speed test and evaluation. This has been recognised in the community, and ongoing efforts are underway to update criteria for future military capabilities [Reference Berger, Cox, Wood, Ott LTC and De Cecchis25]. In this research effort, the first steps have been taken to define a new repeatable MTE to evaluate handling qualities (HQ) during the HAAR scenario. These will also aid the evaluation of future HAAR assistance and automation systems.

In Ref. Reference Kashawlic, Bender and Schwerke13 three simple flying tasks representing different parts of the contact and refuelling phase were defined. Whilst one of the tasks is directly dedicated to assess the contact and another one to the refuelling period, the third one is used to evaluate the pre-contact position. Hence, the latter task also begins in pre-contact position with the probe tip about 10–15ft aft the drogue and the longitudinal axis of both aligned. After the pilot declares he has stabilised a 30s countdown is started, which is the time the position should be maintained. In Ref. Reference Kashawlic, Bender and Schwerke13 an infrared (IR) illuminator was mounted to the probe which enabled probe drogue alignment evaluation to the pilot and crew. The limits were defined as desired when the IR marker stayed on the drogue’s coupler and adequate in case the IR marker stayed within the parachute material.

4.2 Experimental procedure

In this initial research effort, only the pre-contact phase has been considered, based on the procedure described in Section 3.0. Hence, the mission task is initiated maintaining the observation position outside of the tanker wake. The participating pilots are required to transition in the sequence described in Section 3.0 to begin close formation flight. This is the period of flight prior to the pre-contact position and, therefore, no performance requirements are defined for this segment of the manoeuver. Once stabilised in the pre-contact position 10ft behind the drogue, the formal evaluation is initiated. The pilot is required to hold the helicopter in the desired position for 30 seconds.

Initial position accuracy requirements are proposed for this segment of the task. The drogue used in this investigation has an outer diameter of 1.17m. During the hook-up manoeuver, the pilot must accurately position the probe in order initiate refuelling. Failing to achieve the correct position may result in damage to the drogue and/or to the helicopter. At this phase of the manoeuver, very high accuracy must be achieved. During the pre-contact position, although this level of precision is not required, a high level of accuracy should be demonstrated. Specifically, for these initial evaluations, it is also important to ensure that pilots perform a stable formation in the same position, to compare the subjective assessment from the same position in the static flow field. For these reasons, initial tolerances are defined as shown in Table 3. Figure 9 shows the lateral and vertical limits with respect to the drogue size. This gives an indication how the defined precision relates to the eventual precision required during the hook-up phase of the AAR task.

Table 3. Flying task limits

Figure 9. Proportional visualisation of flying task limits.

During completion of the task, lateral, vertical and heading tolerances can be visually judged by the pilot quite easily. On the other hand, difficulties occur judging the longitudinal distance due to the lack of depth perception within the simulator. Therefore, the longitudinal distance to the ideal pre-contact position is numerically displayed to the pilot. At the same time the flight test engineer has an overview of the deviations in the three axes as well as the heading. These time domain plots additionally display the defined performance boundaries so that the graphs can be used to later determine pilot performance achieved during the task.

The piloted simulations for the evaluation were conducted in AVES, using the motion platform. Due to the large motion excursions expected during completion of the scenario, the motion was tuned with low-motion gains, to ensure that no false cueing was experienced from reaching actuator limits. Each condition was flown multiple times prior to formal assessment, to ensure that the pilot reached a comparable level of performance. The pilot is asked to assess Handling Qualities Ratings (HQRs) and Bedford Workload Ratings (BWRs). For tests with active flow field interference, pilots were asked to additionally award two ratings regarding the turbulence: using the Turbulent Air Scale (TS) and a Ride Qualities (RQ) scale. These scales are discussed in further detail in the following section.

4.3 Turbulence rating scales

The TS, shown in Table 4, is a 10-point scale, whereby the pilot is asked to assess the air’s condition in terms of turbulence. Differences between the turbulence levels are defined both in terms of severity and frequency of turbulence. The method has been developed for use in flight testing [26]. The scale has been used at DLR during flight test programs to identify control equivalent turbulence models (CETI) [Reference Seher-Weiß and Jones27]. During these tests, several limitations and drawbacks of the scale were identified by the pilot. These are discussed in detail in Ref. Reference Seher-Weiß and Jones27.

Due to the limitations found when using the TS scale, a novel RQ scale was developed at DLR, particularly for use in flight tests. This scale is presented in Ref. Reference Seher-Weiß and Jones28. It is intended to be a simple method of determining the level of disturbance felt and the frequency of occurrence. The disturbance is characterised by a number and the frequency a letter. Table 5 shows the definitions for the categorisation in detail.

Table 4. Turbulent Air Scale (TS)

Table 5. Ride Qualities (RQ) scale

4.4 Discussion of results

Due to limited simulator availability, only a single pilot evaluated the scenario in this initial test campaign. The pilot had experience with all assessment methodologies used as well as limited previous experience flying a CH-53 aircraft (about 150 flight hours), typical to the GHM simulated. The pilot had no experience flying AAR missions and, therefore, could not give significant feedback as to whether the scenario was realistic. As the task was new to the pilot, a significant number of training runs were performed prior to any formal evaluations.

An initial issue the pilot encountered whilst completing the task was the lack of available cueing to adequately judge the longitudinal distance relative to the drogue. The introduced numerical indicator was therefore found beneficial in order to meet the performance requirements consistently. Nonetheless, the pilot also reported a proneness for low-frequency pilot-induced oscillations (PIO) when mainly focusing on the displayed distance. This aspect might be attributed to the combination of higher relative velocities, the pilot’s reaction time and the stringent longitudinal performance criteria. Further methods to provide relevant information to the pilot to complete the task will be investigated in future work.

Table 6 shows the ratings obtained during the investigation. Simulations conducted without the disturbance from the flow field (inactive tanker wake) were conducted first to determine the task’s realism and difficulty prior to the activation of the flow field (active tanker wake). As the flow field was not activated, neither large change in the ability to complete the task with changes to the visual scenario respectively position were expected nor that HQRs would vary between the cases.

Table 6. Ratings obtained from the pilot during simulator tests

The pilot achieved desired performance, but stated that it was difficult to obtain and therefore awarded HQR 4 for the case with inactive tanker wake. This corresponds to a qualitative HQ Level 2, whereat the quantitative baseline HQ, determined in Section 4.2, predicted Level 1 for small amplitudes in all axes. For moderate amplitudes in the roll axis, only Level 3 was predicted. However, the pilot was neither used to formation flight nor to HAAR, which means more training could have reduced the pilot’s impression of difficulty. Another aspect mentioned by the pilot was the still missing depth perception. Even with the distance numerically provided to the pilot, in fact helping to achieve desired performance, the benefits of depth perception could not be utilised to free the pilot’s capacities. Both these aspects, more training and depth perception respectively more intuitive cueing, might help to improve performance.

The workload ratings awarded reflect HQRs awarded and confirm that the task requires moderate workload even without the addition of the disturbance from the flow field. The rating indicates that the pilot had insufficient capacity to complete additional tasks whilst performing the manoeuver. The pilot stated that the rating was due to the considerable compensation required in the longitudinal axis. This was minimal in the other axes. Furthermore, the pilot stated that the workload rating depends on the kind of additional task. Additional simple tasks (e.g. providing information about conditions) would be possible throughout, but tasks requiring more attention like navigation cannot be performed.

During tests conducted with active wake interference, HQR 5 was awarded. Since the workload experienced by the pilot did not increase, the degradation of HQR could be interpreted as an inability to achieve desired performance at similar pilot effort. For both cases, with the 80 and 120ft hose length, the pilot awarded similar HQRs. This indicates little differences between both pre-contact positions. The pilot indicated that the task with the 80ft hose length caused slightly less workload than the equivalent 120ft hose scenario. As the pilot first flew the 120ft hose case, this may have been an effect of training and should be investigated in future research efforts.

The pilot’s workload rating was unchanged by the presence of wake disturbance. Pilot comments were similar to the case with and without tanker wake interference. The additional disturbance from the tanker wake did not generally increase workload.

Tendencies on both scales used to rate turbulence were found to be similar. While TS 3 on the 80ft scenario expresses light turbulence requiring corrections in manual control, TS 4 was awarded during the 120ft scenario. Hence, the turbulence within the latter scenario, described by continuous small bumps, is considered moderate. This means the pilot noticed stronger turbulence on the 120ft scenario, which is contrary to results expected prior to the pilot tests. Furthermore, the pilot’s RQ ratings confirm this tendency by increasing the turbulence category from light frequent disturbance (RQ 1-C) in the 80ft scenario to light persistent disturbance (RQ 1-D) in the 120ft scenario. Nevertheless, the wake velocity magnitudes at the rotor hub reference ACP (V_w), shown in Fig. 10, expose higher absolute velocity during the 80ft scenario, reflecting the results from the CFD analysis.

Figure 10. Time history plots of the station keeping phases.

Figure 11. Time varying spectograms of GHM’s rotational rates.

Once fully stabilised, the steady flow field does not provide huge changes in wind velocity. The changes are originated in the helicopter’s movements induced by the pilot, which can be seen in Fig. 10. In the 120ft scenario the pilot notices proximity to the lower boundary in longitudinal direction (Δx) at about 18s, indicating an overshoot relative to the ideal pre-contact position. To remain within the margins the pilot excessively uses the collective (d₀) to reduce speed at 19s, but rapidly comes close to the upper boundary due to overcontrolling. Such a PIO is exemplary of the previously mentioned proneness to PIO and might be avoided by training in this specific case. Starting from the initial intense collective input until ‘recover’ at about 28s the largest changes in tanker wake velocity (V_w) occur. This indicates wake velocity magnitude is provided by the flow field but wake velocity dynamics affecting the helicopter are induced by the pilot.

In Ref. Reference Lampton and Klyde29 tendencies were demonstrated that pilot effort can be described by the activity of power in time frequency spectrograms of the helicopter’s rotational rates. The rate’s power activity, summarised by the average of the time varying power frequency, was found to correlate to HQR. The pilot’s effort distribution over time and frequency can thus contribute to the identification of effort driving factors and possible changes in control strategies. Nonetheless, it must be kept in mind that some aspects (e.g. degraded control power) reduce the correlation and therefore validity.

The spectrograms of the roll and pitch rate during a complete HAAR evaluation run are visualised in Fig. 11. The lowest power activity can be seen in the 120ft scenario with inactive tanker wake (a, b). Medium activity is observed at the 80ft scenario and the highest activity shown by the 120ft scenario, both with active tanker wake disturbance. This also reflects the ratings given by the pilot. While the 120ft scenario is generally characterised by higher activity in the begin and end of a run and lower activity during longitudinal close up, the characteristic is not as pronounced in the 80ft scenario (Fig. 11; c, d). This may be attributed to the pilot’s strategy change to mainly perform the manoeuver by visual references. Generally, the characteristic can be explained by higher efforts to line up probe and hose in the beginning followed by smoothly closing up to the drogue. Then again higher effort is necessary to serve the performance requirements.

However, the largest peaks in power occur on the end of the 30s position keeping phase of the 120ft scenario with active wake disturbance (Fig. 11; e, f). But, the peaks only evolve during this phase. Starting the 30s phase at about 65s, power begins rising 15s later at about 80s. It is also the moment of beginning overcontrolling. In consistence with the observations of the time history diagrams, the main driver for the unexpected turbulence ratings may not be the flow field itself, but turbulence dynamics build up by the pilot.

To reduce over control, options for improved longitudinal cueing and/or an increase of the longitudinal performance requirements should be examined. Moreover, the impact considering additional flow field interaction on the fuselage, tail rotor, vertical and horizontal stabiliser should be analysed. Yet, it is also unclear how the actual drogue movements would impact position keeping and therefore impact wind velocity dynamics. Nevertheless, wake velocity changes would still be pilot induced. This theoretically means combining a trimmed helicopter with perfect position keeping, would not require huge pilot efforts at all. Thus, the tanker wake is recommended to be modeled as an unsteady time-depended flow field, or additional turbulence models should be included in future investigations.