3.1.1 NRC Bell 412 Advanced Systems Research Aircraft (ASRA)

The Bell 412 ASRA (Fig. 2) complements the experimental work carried out in the HELIFLIGHT-R simulator through its key capability, the full authority experimental Fly By Wire (FBW) flight control system, allowing the aircraft to be operated as an in-flight simulator [Reference Gubbels and Ellis26]. ASRA’s flight data recording system consists of a range of data monitoring and measurement equipment, facilitating the assessment of the pilot control activity and aircraft task performance analysis. The FBW system enables the implementation and testing of new control laws in flight.

Figure 2. NRC ASRA Bell-412 Aircraft [Reference Perfect, Gubbels, White, Padfield and Gubbels2].

3.1.2 HELIFLIGHT-R simulator

The University of Liverpool’s Flight Science and Technology research group operates a fully reconfigurable research simulator, HELIFLIGHT-R (Fig. 3); acceptance and commissioning of the simulator was completed in 2008 [Reference White, Perfect, Padfield, Gubbels and Berryman27]. The simulator has a three-channel 230° × 70° Field of View computer image generation system, a 6-degree-of-freedom hexapod motion platform, a four-axis force-feedback control loading system and an interchangeable crew station. Flight dynamic models are developed in either FLIGHTLAB [Reference Du Val and He29] or MATLAB®/SIMULINK® and the current aircraft library features a range of fixed-wing, rotary-wing and tilt-rotor aircraft. The outside world environment is generated using the Presagis Creator Pro software, which is integrated with Presagis’ VEGA software into the Liverpool Virtual Environment. The visual and motion cues, together with sound cues, provide a powerful immersive simulation environment for a pilot [Reference White, Perfect, Padfield, Gubbels and Berryman27].

Figure 3. HELIFLIGHT-R Simulator (foreground) [Reference White, Perfect, Padfield, Gubbels and Berryman27].

3.3.1 Control attack

In the LS research, measures of the pilot control activity were used in attempts to quantify the pilot control workload [Reference Perfect, Gubbels, White, Padfield and Gubbels2]. One of the measures used was the control attack parameter, A _η (Equation (1)), introduced earlier. The A _η characterises each discrete control input and is defined as the ratio of the peak rate of control displacement, ${\dot \eta _{pk}}$ , to the magnitude of the change in the control displacement, Δη.

(1) \begin{align}{A_\eta } = \frac{{{{\dot \eta }_{pk}}}}{{\Delta \eta }}\end{align}

In broad terms, higher attack values indicate rapid-small control deflections while a lower attack value indicates slower-large control deflections. As with the ADS-33 attitude quickness parameter [5], the attack approximates the inverse of the time to change for a simple ramp. An example of a high attack and low attack pilot control input are shown in Fig. 4.

Figure 4. Control deflection (lateral cyclic) time history showing high (A_η2) and low (A_η1) values of attack.

The origin of the attack metric in control analysis goes back to Ref. (Reference Padfield, Jones, Charlton, Howell and Bradley17), and relates to the property of a sound envelope in music, where the attack is the time taken for the initial run-up of level from nil to peak, so strictly the inverse of how it is used in control activity analysis. Figure 5 shows an example of two ideal control inputs; decisive and quick, and smooth and slow, having ramps of different slopes.

Figure 5. Ideal control inputs showing two ramps of different slopes; decisive and quick, smooth and slow.

The Attack Number (A _ηN ) and Attack Activity Rate (A _ηR ) metrics are defined as:

1. (a) Attack Number (A _ηN ): the total number of times that the pilot displaces a particular control having a magnitude higher than a specified threshold value. Discussion on the appropriate value of the attack thresholds comes later in this section.

2. (b) Attack Activity Rate (A _ηR ): the ratio of the total number of A _η to the time length of the MTE. It can be defined as the ‘busyness’ metric and an average for an MTE.

3.3.1.1 Control attack threshold

When assessing different attack-based metrics to examine correlations with HQRs awarded by a pilot, it is important to determine a threshold value for what might be described as a productive pilot control input. To explore this approach, the threshold was increased in equal intervals and the impact on the A _ηN and A _ηR registered within different MTE phases was examined.

Figure 6 presents an example of a RS MTE performed in the simulator (see Appendix A: Table A5). Figure 6(a) and (b) show the aircraft’s ground track and the pilot lateral cyclic control activity XA, respectively. The gates (G) for the RS MTE are specified as numbered stars in Fig. 6(a). Figure 6(c) and (d) show the XA control attack for the whole MTE (combined) and spatially segmented-MTE, respectively. Results for two attack thresholds (0.25 and 2.5%) are compared. The control position threshold percentage is calculated for the full control travel range, −50% to +50% (i.e. −6.14:6.33in for XA; −6.1:6.1in for XB; 0:10.7in for XC; −3.92:2.86in for XP).

Figure 6. Control attack threshold selection and segmentation approach, (a) Aircraft track; (b) Lateral pilot stick input; (c) Combined control attack; (d) Segmented control attack.

Using the 0.25% threshold, Fig. 6(c) shows a significant number of the low amplitude-high attack values captured. To spatially identify whether the control activity associated with these attack points was deliberate and required or not, the MTE is divided into segments, enabling examination of pilot compensation in different MTE phases. In this case, five segments were examined based on the positioning of the gates used to traverse the runway (i.e. G4–G6 for the Left Hand Side (LHS) to Right Hand Side (RHS) runway crossing and G9-G13 for the RHS to LHS runway crossing), as shown by the vertical dashed lines in Fig. 6(a) and (b), respectively.

Figure 6(d) shows five subplots associated with these segments. The 1st segment (Start-G4) represents the run-up to the beginning of the MTE (i.e. between −1,000 to 1,500ft runway track) and was used to define the threshold. In this segment, there are no specific task requirements and thus the pilot should not have to apply G&S control inputs, as the simulation model is trimmed, until just before the end of the segment to initiate the turn at G4. Any attack points registered before the turn are considered unproductive from the task perspective, as they do not contribute to G&S elements required along the runway edge. However, this segment does show a relatively large number of low-amplitude high-attack points from Fig. 6(c). The threshold was increased in equal intervals of 0.25% until a single deliberate guidance control attack featured in the first segment, required to initiate the turn. This was achieved at the threshold of 2.5%, as shown in Fig. 6(d)-1st (Start-G4) segment as a star marker. It should be noted that for all the following attack analysis, the threshold of 2.5% is used.

Table 2 shows A _ηN and A _ηR metrics for the segmented attack charts shown in Fig. 6(d), respectively, for 0.25 and 2.5% threshold and their percentage difference. It can be seen that, in the first segment, nearly 90% of the attacks have been filtered out using a 2.5% threshold. To gain more insight and confidence in the suitability of the new attack threshold, a moving window-based time-localised attack metric was formulated for comparison with the TFD metric.

Table 2. Segmented control attack metric calculated using 0.25 and 2.5% threshold for the data presented in Fig. 6(d)

3.3.1.2 Time-varying localised attack activity rate

How the attack parameter is used to correlate with the HQR connects with the idea of a peak or an average rating across the MTE. Using the average A _ηR for the whole MTE can mask local peaks, making it difficult to identify the phase(s) within the MTE where a pilot might be struggling with HQ deficiencies of the vehicle. This section presents a time-varying attack activity rate metric A _ηR ^loc , formulated to capture local control peaks in an MTE that would help in identifying where the pilot is working the hardest, and the dominant area of compensation. The method includes the formulation of a control segmentation approach using a moving 5s window, with 2.5s overlap. The 5s window is selected based on the minimum time needed for the pilot to perceive visual cues and successfully execute a manoeuvre within an MTE [Reference Padfield, Clark and Taghizad31].

Figure 7 shows an example of the XA A _ηR ^loc metric along with the aircraft runway track and pilot control input compared for the two simulated aircraft model configurations flown in the RS MTE [Reference Cameron, Memon, White and Padfield32]. Config A represents a baseline simulation model of the ASRA, whilst Config B represents an improved fidelity simulation model developed to better match flight test data [Reference Cameron, Memon, White and Padfield32]. Figure 7 shows differences in the A _ηR ^loc during different phases through the MTE. As the pilot enters the first LHS-RHS crossing (t = 17.5–27.5s), the A _ηR ^loc increases to >1/s in both cases. The phase (t = 30–37.5s) corresponds to the tracking phase on the RHS edge of the runway. Here, the A _ηR ^loc for Config B increases to 1.4/s, the pilot sustaining this level through to the final tracking on the LHS of the runway, where the A _ηR ^loc reduces to about 0.5/s. For Config A, the A _ηR ^loc reduces to 0.2/s during the RHS runway tracking, but then builds up to a high attack level of 1.8/s during the final LHS tracking; the pilot struggling to maintain track during this, largely, stabilisation phase. Reference (Reference Cameron, Memon, White and Padfield32) discusses the impact of the natural frequency and relative damping of the lateral-directional oscillation on the required compensation, explaining the difference between configurations A and B in terms of these dynamic stability characteristics.

Figure 7. Time-varying localised A_ηR ^loc (star markers define RS MTE phases).

With such large variations in the A _ηR through the MTE, using the average attack can be misleading in terms of the HQR assigned and will not help in the identification of the dominant compensation phase(s).

3.3.1.3 Time-frequency domain and attack-rate metric composite

In the TFD, Fourier transform-based spectrograms [Reference Tritschler, O’Connor, Holder, Klyde and Lampton12] have been used to examine the characteristics of the pilot control inputs, complementary with the A _ηR control metric. A spectrogram is a three-dimensional visual representation of the spectrum of frequencies within the control input as a function of time. It defines the signal in terms of time, frequency and magnitude.

In this paper, composite TFD-A _ηR ^loc charts have been developed to demonstrate the applicability of the A _ηR metric in assessing pilot control compensation. Figure 8 shows another example for the RS MTE. The central spectrogram contour-plot (formulated using a moving 4s window; selected to provide a suitable resolution in both time and frequency, thus highlighting key features of the control signal) shows the time and frequency on the abscissa and ordinate, respectively, with the colour bar indicating the magnitude of the Short-Time Fourier Transform (STFT) of the control signal; the outer plots show the time history (upper), Fast Fourier Transform (FFT, left) and A _ηR ^loc (lower).

Figure 8. Composite TFD-A_ηR metric plots for lateral cyclic control activity [Reference Cameron, Memon, White and Padfield32].

Using this presentation form, details in different phases of the MTE can be observed, along with information for the frequency content as a function of time. The FFT shows the dominant peak around 0.15Hz (0.94rad/s), corresponding to the large control movements during the runway crossing, with a duration of about 10s at 90kn. The spectrogram and A _ηR ^loc variations show where in the MTE the pilot is working the hardest. The two obvious peaks (@ 27.5 and 45s) are at the runway crossings, and particularly re-establishing track over the runway edges. The A _ηR ^loc , in combination with the TFD, provides complementary insight into the compensation required to complete the task.

3.3.1.4 Combined multi-axis attack activity rate metric

So far, the attack metrics have been utilised in single-axis form and do not capture multi-axis control compensation in a complex MTE. A method for calculating a combined attack metric, including all four controls, is required. Moreover, due to the variations in piloting strategy and different levels of control compensation applied by different pilots, it is important that the accumulation methodology is adaptive to these variations.

To address this, a weighted-adaptive formulation has been developed that weights individual control A _ηR contributions with the corresponding fraction of the total control attacks (A _ηNTot ) applied in that axis (Equation (2)).

(2) $$\matrix{{{{{A_{\eta Ni}}} \over {{A_{\eta NTot}}}}} \hfill \cr}$$

where i corresponds to the four (XA, XB, XC and XP) controls.

This is calculated for each control (XA, XB, XC and XP) and then summed to obtain a combined, weighted adaptive metric A _ηRC (Equation (3)).

(3) $$\matrix{{{A_{\eta RC}} = {{\sum\nolimits_{i = XA}^{XP} A }_{nRi}}.{{{A_{\eta Ni}}} \over {{A_{\eta NTot}}}}} \hfill\cr} $$

Alongside the A _ηRC , the combined A _ηR for the primary A _ηR-Pr and secondary A _ηR-Sec controls separately have been computed in the MTE investigations. To calculate these, the same formulation as defined in (Equation (2)) has been used by accumulating the specific controls only. The primary and secondary controls are different for each ADS-33 MTE based on the task performance requirements, see Table 3. The concept of the primary and secondary controls is discussed in the following section.

Table 3. Primary and secondary controls for ADS-33 MTEs

The application of the combined adaptive A _ηRC metric will be demonstrated in Section 4, by examining the correlation with the HQRs awarded by the pilots for a range of MTEs with different task performance requirements and dominant control axes. But first, we return to the separation of control functions as G&S.

3.3.2 Guidance and stabilisation control

Another key interest in the study of pilot control compensation is to identify the regions of G&S components [Reference Jones, Padfield and Charlton23]. To explore this using the attack metric, a new approach based on the concept of the Perfect Pilot (PePi) strategy is introduced in this study.

PePi is defined as the pilot who does not need to apply any compensation to fly an MTE, and so all the control movements (primary and secondary) contribute directly to the task performance. The PePi is based on the guidance control activity only and the guidance frequency for an MTE is approximated as the inverse of the time for the required flight-path change. Another way of looking at this is that the stabilisation activity is only ever required to suppress external disturbances, errors of judgement or natural mode transients, so should be minimal for the types of MTE examined in this paper.

Consider the example of the RS MTE, flown at 90kn, shown in Fig. 9, in which the primary MTE state and flight-path change are the roll attitude and lateral position, and the primary control is XA. Focussing on the first runway crossing, to complete this phase, there are three primary roll attitude changes required; the first to initiate the crossing ( $ \delta\phi \approx +30^{\circ}$ ), second for the roll-reversal ( $ \delta\phi \approx -50^{\circ}$ ) and third for lining up with RHS runway edge ( $ \delta\phi \approx +20^{\circ}$ ). This assumes that the pilot does not level off after the first turn and fly straight for a period during the crossing. Each roll attitude change is achieved with a certain quickness; the ratio of the peak attitude rate to the attitude change [Reference Padfield, Jones, Charlton, Howell and Bradley17], and each roll quickness is achieved by the pilot applying primary XA input. With a Rate Command (RC) response type, two stick movements (i.e. a pulse) are required. Whereas, for an ACAH response type, a single stick movement (i.e. a step or more realistically a ramp) is required. Associated with each stick movement, there is an attack parameter. For the RC aircraft, there are six XA attacks and three for the ACAH, during crossing.

The RS requires the pilot to maintain height and speed within defined margins and pass through the gates within margins on lateral position and heading. The heading is normally controlled by the pilot trying to fly a balanced manoeuvre, so zero lateral acceleration during the turns. Therefore, three secondary controls are required to achieve the full performance standards (longitudinal XB, collective XC and pedal XP). Moreover, to maintain a steady turn, the pilot needs to command both a pitch and a yaw rate; hence, for a right turn, requiring aft stick and, normally, right pedal in the steady turn. The rates are generated by step inputs in XB and XP, hence one attack point each to generate a turn. To complete the runway crossing requires three such inputs, hence three attacks for each control. To maintain height and speed, the pilot needs to increase power, hence collective pitch; again, one attack for each of the three phases of the crossing.

Note that PePi uses all four controls to fly the task, although none of these inputs can be regarded as compensation for cross-couplings, errors of judgement or modal transients (e.g. the lateral-directional oscillation). A real pilot will need to make such corrective, compensatory inputs.

Figure 9. Segmented attack for PePi normalisation, (a) Aircraft track; (b) roll attitude; (c) control inputs; (d) control attack.

For the RS MTE, the runway crossing is assumed to be a single flight-path change. The crossing time is approximately 10s (1,500ft between G4-G6 at 151.8 ft/s), giving a frequency of about 0.6rad/s. A real pilot may elect to divide the crossing into two sub-phases, when we would expect to see higher guidance frequencies and attack values within the control activity. When flying an MTE, PePi will apply a smooth control strategy for the whole manoeuvring phase of an MTE, without periods of no activity. So, the fundamental guidance frequency corresponds to the time for the trim-to-trim, e.g. a runway crossing in the roll-step but the start to stop in cases like the AD or LR MTEs.

With this approach, the real pilot’s control attack can be normalised using the PePi attack to obtain an attack ratio A _ηN-n , informing the identification of G&S contributions to compensation, as A _ηG and A _ηS . A _ηG is defined as the ratio of the PePi to the real pilot’s attack, while the remainder is the A _ηS . Consider the pilot attack’s presented in Fig. 9(d), for all four controls, compared for the two crossings. Table 4 shows the A _ηN-n for each of the controls, along with the corresponding A _ηG and A _ηS . A _ηN for the XA input is 11 in the first crossing and 10 in the second crossing. Normalising these with the PePi attack of six gives A _ηN-n of 1.83 and 1.66, respectively. So, the pilot applies 83% and 66% more XA movements than PePi in the first and second crossings, respectively. With this interpretation, for both crossings, more than 50% of the primary control represents guidance activity. Looking at secondary controls, we see the pilot is busy with XB and XC, controlling speed and height, and XP controlling balance and heading corrections. Again, according to this interpretation, nearly 80% of XB activity relates to stabilisation, including guidance corrections; the corresponding figures for XC and XP are 57% and 40%, respectively.

Table 4. Normalised attack numbers, and guidance and stabilisation attack for the data presented in Fig. 9(d)

Following this strategy, the PePi metric was computed for PH, AD, Pr and LR MTEs; Table 5 shows the total PePi attack numbers for the whole MTEs flown with ACAH and BA/RCAH configurations. The PePi attack breakdown w.r.t specific phases in these MTEs are tabulated in Appendix B.

Table 5. Total PePi attack numbers for ADS-33 MTEs

We return to the investigation of the PePi normalisation by examining its correlation with the HQRs awarded by the pilots in flight and simulation tasks in Section 4. For the PePi metric control accumulation, the A _ηG and A _ηS calculated individually for each control (XA, XB, XC and XP) are averaged to obtain a single combined mean value (Equation (4)), $\overline {{A_{\eta G}}} $ and $\overline {{A_{\eta S}}} $ , for comparisons with HQRs.

(4) \begin{align}\overline {{A_{\eta G,S}}} = \frac{1}{N}\mathop \sum \nolimits_{i = XA}^{XP} {A_{\eta G,Si}}\end{align}

where i corresponds to the four (XA, XB, XC and XP) controls and N is the number of controls.

3.3.3 Cut off Frequency (CoF) metric

The CoF metric as defined in Section 2 is the frequency at which 70% of the control displacement (50% of control power) signal has combined when represented as a gain in the frequency domain [Reference Blanken and Pausder20, Reference Tischler and Remple21]. It is particularly meaningful when the pilot is acting to correct errors from the intended performance in the presence of a random disturbance. Figure 10 illustrates the CoF method in an outer plot (right), in combination with the composite TFD-A _ηR metric plots (Fig. 8), for a PH MTE flown in the ASRA using the BA configuration. The plot shows a ratio between the control input amplitudes summed up to the current frequency (Ση) and the amplitudes summed over the full frequency range (Ση _TOT ). The CoF is the frequency at which this ratio is 0.7.

Figure 10. CoF and composite TFD-A_ηR metric plots for lateral cyclic control activity.

The CoF is calculated using the frequency interval 0.2–2Hz; the lower limit largely removing the lower-frequency guidance element of the control activity from the analysis, and the upper limit removing artefacts (e.g. measurement noise, digitisation, and pilot inceptor dynamics [Reference Tritschler, O’Connor, Holder, Klyde and Lampton12]) introduced in the time-frequency domain transformation [Reference Perfect, Gubbels, White, Padfield and Gubbels2]. Figure 10 shows the CoF metric calculated for the whole PH MTE as 0.6Hz and for the 30s final hover-stabilisation phase as 0.7Hz; so, 16% higher frequency content in the last 30s of the MTE compared to the whole MTE. The A _ηR ^loc increases from the start to the end of the MTE.

In this section, the utility of the CoF metric in understanding the pilot control compensation is demonstrated. However, with its focus on discrete events, the attack metric exposure reflects pilot control compensation better for the complex multi-axis MTEs. The application of the combined adaptive A _ηRC and PePi normalisation metrics are now demonstrated by examining the correlation with the HQRs awarded by the pilots for a range of MTEs with different task performance requirements and dominant control axes.

4.1.1 Case study 1: Flight test

Flight test data gathered for five different MTEs (PH, AD, Pr, LR and RS) flown in the ASRA using the BA and ACAH configurations are assessed (see Table 6). For each case, Table 6 compares the key predicted HQs feature and level with the assigned (pilot) HQs level, rating and experience, and the associated peak phase(s) captured from the A _ηR ^loc metric. Comparisons of the adaptive A _ηRC metric with the HQRs are presented in Fig. 11, showing results for all four controls combined, and separated primary and secondary controls. Each plot shows the average and the peak A _ηR results. Within the figures, the data points are identified with a unique number that corresponds to the specific test case (i.e. MTE and configuration), given in Table 6. The best fit straight-line (dashed) between the HQRs and A _ηRC , are shown, along with the coefficient of determination (r ² ), and 90% prediction-boundary calculated using the T-distribution [Reference Navidi33]. The coloured regions differentiate the HQ levels from the HQR scale, with the associated compensation descriptors on the secondary Y-axis.

Table 6. Details of the cases presented in Fig. 11 (pilot A in ASRA)

Figure 11. HQR vs peak and average A_ηR for (a) Combined, (b) Primary and (c) Secondary controls (Pilot A: Flight Trial).

The first observation is that all results lie in the HQ Level 2 range. The second is that for three of the four MTEs flown with the BA were awarded HQR 6. Third, within this seemingly narrow band, the A _ηR more than doubles, when viewed as the average or peak. Of course, Level 2 is not a narrow band; rather it covers the range from desired performance achieved with moderate compensation to adequate performance achieved with extensive compensation. The change from desired to adequate performance within this band is significant in terms of the expected relationship between the HQR and any compensation metric. It is likely to be nonlinear and the results do suggest this. Figure 11(a) shows a reasonably positive correlation between the HQRs and A _ηRC, with the A _ηRC ^avg showing the highest r ² value of 89%, with the narrowest spread and the prediction boundary amongst all the cases. The correlations for separated primary and secondary controls are lower. This can partially be explained by the different sources of deficiency and consequent compensation in the different MTEs for the same HQR. For example, in Case 8 (LR, BA) the dominant compensation was from the secondary controls (e.g. XB); for Case 9 (PH, BA) the dominant compensation was from the primary controls (XA, XB). The combined peak A _ηR is similar (1.6–1.8) so reflecting the different contributions to overall compensation.

The separation of the controls into primary and secondary is found to be important for understanding the respective contribution of each to the corresponding A _ηRC values since, within each of these MTEs, different controls will be dominant during different phases of the MTEs. To illustrate this, two cases; (1 (AD-ACAH) and 9 (PH-BA)) have been selected from the two extreme HQ rating groups (4 and 6) to represent the moderate and extensive levels of control compensation, where the pilot commented the workload to be tolerable for Case 1 and extremely high for Case 9.

Case 1 is the AD MTE flown with the ACAH configuration and awarded an HQR 4 by the pilot. Figure 11(a) shows that the A _ηRC ^pk is almost double the A _ηRC ^avg . This is dominated by the secondary controls where the difference between the A _ηR-Sec ^pk and A _ηR-Sec ^avg is almost 0.7/s, whereas, in the primary controls, the difference between the A _ηR-Pr ^pk and A _ηR-Pr ^avg is <0.1/s. Between primary and secondary controls, the latter dominates; the A _ηR-Sec ^pk being almost double the A _ηR-Pr ^pk . The pilot commented that he applied “at least moderate compensation particularly in the off-axis channels”. From the TFD-A _ηR ^loc charts shown in Fig. 12, the peaks largely occur during the stabilised hover capture phase with peak A _ηR ^loc > 1.5/s in XA and XP. The pilot commented that due to the poor engine governing dynamics, the control strategy had to be modified slightly from his usual accel/decal strategy. The spectrograms also indicate that there is a hot-spot in the low-frequency region below 0.2Hz in the four controls, which corresponds to the guidance frequency range and reducing significantly at higher frequencies; as expected the guidance-maximum is in the primary XB control.

Figure 12. TFD-A_ηR ^loc composite plots for case (1, AD-ACAH), HQR 4.

Case 9 is the PH MTE flown with the BA configuration and awarded an HQR 6 by the pilot. Compared to the previous case, the overall control compensation is much higher, as shown in Fig. 13. The primary controls, XA and XB, are dominant, with A _ηR-Pr 2.5–3 times the A _ηR-Sec , for both average and peak values. The TFD-A _ηR ^loc plots for the primary controls show A _ηR ^loc increasing from the start to the end of the MTE, with a peak of 3/s in XB at the end of the stabilisation period. The pilot commented that he “cannot relax at any point during the manoeuvre”. Although the pilot commented that “the task required compensation in all axes continuously”, the TFD-A _ηR ^loc plots show the dominant primary controls, with A _ηR for XB remaining above 2/s throughout the MTE. The A _ηR ^loc for XC remains around 1/s and XP remains below 2/s in the initial and final phases with a dip during the hover capture phase for XP. The TFD-A _ηR ^loc also shows high-frequency content in all the controls throughout the MTE. However, at a high-frequency range (>0.5Hz), the amplitudes are lower in secondary controls compared to the primary. For this case the pilot commented that “accuracy was inconsistent, chasing the aircraft rather than controlling it”, “poor stability created issues in pitch, roll and yaw” and “the level of precision could be achieved but the workload was extremely high throughout the MTE”.

Figure 13. TFD-A_ηR composite plots for case (9, PH-BA), HQR 6.

4.1.2 Case study 2: Simulator test

The same aircraft configurations i.e. BA and ACAH, and MTEs, were flown in HELIFLIGHT-R by the same pilot and the results are shown in Table 7 and Fig. 14(a), (b) and (c). It is noted that the A _ηR values are lower, and the gradient of variation with HQR is steeper than in the ASRA flights (Fig. 14(a) compared with 11(a)). Also, the awarded HQRs extend into the Level 3 region. Again, in this section, two different HQR cases 1 and 9 have been selected from two extreme HQ rating groups (i.e. 4 and 7) to represent the moderate and maximum tolerable levels of pilot control compensation.

Table 7. Details of the cases presented in Fig. 14 (pilot A in Simulator)

¹ Note that roll from pitch cross coupling is predicted as level 2 for forward flight.

Figure 14. HQR vs peak and average A_ηR for (a) Combined, (b) Primary and (c) Secondary controls (Pilot A: Simulator Trial).

As in the flight test, the A _ηRC results show a strong level of overall correlation (Fig. 14(a)). However, for the primary and secondary controls (Fig. 14(b) and (c), respectively), the correlation is much lower. As in flight test cases, the BA and ACAH configurations are separated in the HQR levels. The BA configurations are located in the top-right region of the band with the expected higher levels of control compensation (and HQRs ≥ 6). The ACAH configurations are located in the bottom-left region with correspondingly lower levels of pilot compensation (and HQRs ≤ 5). However, when separated into primary and secondary controls, these regions are not as clearly distinguished as with all four controls combined. As discussed for the ASRA data, this is due to the way the compensation is distributed amongst the four controls.

Case 1 is the LR MTE flown with the ACAH configuration and awarded an HQR 4. Compared to the HQR 4 case in the flight test (Fig. 12), the level of compensation as calculated by the TFD-A _ηR ^loc is lower. For the LR MTE, the A _ηRC ^Pk is 1.6/s in flight for the ACAH, compared with 1.0/s in simulation; 2.1/s compared with 1.2/s for the BA configurations. In this case, the secondary controls are dominant with A _ηR-Sec almost three times the A _ηR-Pr , for both average and peak results. The TFD-A _ηR ^loc plots in Fig. 15 provide insight into these differences by showing that, among the three secondary controls, XB and XP appear to have higher peak A _ηR ^loc > 1/s, whilst for XC, the A _ηR ^loc remains <1/s throughout. As in the flight test Case 1, the primary control shows a large low-frequency hot-spot <0.2Hz, which drops at higher frequencies. This low-frequency peak is, again, attributed to the guidance frequency corresponding to the core task frequency. From the control activity and TFD charts, the pilot was active in XB throughout the MTE having peak A _ηR ^loc occurring during the translation phase, while for all the rest of the controls the peaks occurred during the hover capture phase. The pilot commented that “the most difficult element of the task was the longitudinal position keeping”. This was predominantly due to poor outside visual cues in the simulation environment, which in the LR MTE, dominates the pilot’s control strategy. During the assessment of the visual cueing, it was found that due to the hardware limitations, rich textural environment (close to the real-world) could not be fully recreated. This restricted the ability of the pilot to apply accurate translational rate and attitude corrections [Reference Perfect, Gubbels, White, Padfield and Gubbels2].

Figure 15. TFD-A_ηR ^loc composite plots for case (1, LR-ACAH), HQR 4.

Case 9 is the PH MTE flown with the BA configuration and awarded an HQR 7. Figure 14(b) and (c) show that the dominant controls are primary, with the A _ηR-Pr being 2–3 times the A _ηR-Sec , for both average and peak results. Expanding on this using the TFD-A _ηR ^loc plots in Fig. 16, the A _ηR ^loc peaks to 2/s in XA, while for XB it remains ≥1/s throughout. The TFD plots show consistent high-frequency content with multiple peaks around 0.4–0.6Hz for these controls. The pilot commented that “there were continuous control perturbations, the aircraft was easy to excite and never truly stabilised in the hover”. This can be seen from XA and XB control activity having continuous compensation at the hover spot in the last 30s window. The XA plot clearly shows the level of compensation in capturing the hover point (between 18 and 20s) and throughout the 30s hover (20–50s). This is captured as a ramp in the A _ηR ^loc and the first hot-spot in the TFD spectrogram. At the same location in time, hot-spots are shown in XB and XC with A _ηR ^loc s. In this case, the aircraft model was predicted to have Level 2 stability, which required the pilot, during testing, to use maximum tolerable compensation in all axes and is reflected in the spectrograms.

Figure 16. TFD-A_ηR ^loc composite plots for case (9, PH-BA), HQR 7.