This research aims to advance the state-of-the-art requirements for helicopter pilot training in flight simulators contributing to the rotorcraft safety enhancement framework. It has been conducted as part of the Marie Sk\l odowska-Curie European Joint Doctorate NITROS (Network for Innovative Training on Rotorcraft Safety).
Training has always been the traditional answer to help pilots deal with flying vehicles and scenarios that, without adequate preparation, would otherwise be unforgiving. Due to the risks and costs involved in training for such critical circumstances, the exclusive use of in-flight training is untenable, especially for helicopters. A combination of simulator and in-flight training is the solution adopted to reduce accident rates and human fatalities in a safe and efficient manner and to fulfill the ever-harsher mandate for flawless performance required by the military domain. Inevitably, the use of simulation to support pilot training brings forward the issue of skills and performance transfer from the simulator to the actual aircraft, which is addressed in this thesis in relation to helicopters.
The primary focus of flight simulation transfer-of-training research is to assess how learning a task in a flight simulator affects the trainee's performance capabilities in the same task in the actual aircraft. To explicitly measure the transfer of behavior learned in a certain setting (e.g., a simulator) to the evaluation setting of interest (e.g., a real aircraft), transfer-of-training experiments are one of the few available methods for direct evaluation of the training effectiveness. To measure pilot transfer of skills at least two groups of participants are required. The speed of learning in the actual aircraft by one (or more) "experimental" group(s), previously trained in the simulator, need to be compared with the learning performance of a "control" group having received no special previous training. While this design enables to directly assess the effectiveness of a simulator, it requires strictly balanced groups according to participants' relevant prior training and experience to deliver meaningful results.
Several variations of this basic transfer model, named a true-transfer design, have also been proposed. The most popular is the simulator-to-simulator transfer model, also known as quasi-transfer design. In quasi-transfer experiments, participants are not transferred to the real-world setting, but to a different, often more realistic or enhanced, simulation environment. The quasi-transfer paradigm relies on the assumption that the more realistic simulator acts as a valid replacement for the actual aircraft. Although this is a strong assumption, its effectiveness for evaluating skill transfer is corroborated by experimental evidence. Furthermore, a quasi-transfer design avoids the costs, hazards, and scheduling hindrances (e.g., interruptions due to bad weather) of a true-transfer experiment and offers the possibility of safely investigating dangerous situations such as engine failures. Another issue arising from true-transfer studies (and from flight tests in general) is the reliability of the performance measurements in the real-world setting. Moreover, there are inevitable psychological differences in a pilot's mindset between training in a simulator or in the actual aircraft. This is not necessarily a disadvantage from a training perspective, because relieving the trainee of the stress and the workload deriving from auxiliary duties (e.g., safety and flight regulation aspects, communication, periodic systems monitoring, etc.) enables devoting more mental resources to learning.
For this thesis, three quasi-transfer-of-training experiments were conducted to test the effectiveness of flight simulator training for two different helicopter tasks: hover and autorotation. The ability to hover, i.e., to remain in a nearly stationary flight condition, is the main capability that differentiates helicopters from fixed-wing aircraft. The ability to autorotate, i.e., to keep the rotor spinning by means of the airflow passing through it, is an essential emergency maneuver that enables helicopter pilots to often safely reach the closest suitable landing site in the event of an engine failure.
These two maneuvers were not randomly chosen. The choice was based on the fact that hover and autorotation pertain to different phases of the helicopter pilot training syllabus. While both maneuvers need to be mastered by helicopter pilots, hover is generally the very first maneuver that student pilots learn to perform, whereas autorotation is practiced only when the trainee demonstrates a sufficient level of proficiency in maintaining/controlling the airspeed and the rotorspeed. Therefore, hover and autorotation can be characterized as a "basic" and an "advanced" maneuver, respectively. Furthermore, hover is performed in normal operating conditions, whereas an autorotation represents an abnormal mode of operation for helicopters and is thus performed only in emergency circumstances. On the other hand, hover is performed by helicopter pilots on a daily basis (or at least every time they fly). Fortunately, nowadays engine reliability is high and they seldomly fail, meaning that real power-out autorotations are not performed often. However, to be prepared for a potential occurrence, simulated engine-failures (generally with a power-recovery, i.e., terminating in a hover) are practiced during recurrent training and proficiency checks. It is therefore evident that issues in simulator training of the hover maneuver need to be assessed especially in relation to novices (ab-initio training), while those in simulator training of the autorotation maneuver require a focus on experienced pilots (recurrent training).
The type of maneuver (e.g., basic or advanced), the operating condition (e.g., normal or abnormal mode of operation), and the trainees' characteristics (e.g., novice or experienced pilots) are all factors that play a role in the level of simulator fidelity needed for effective training. In contrast to the unquestioning and unceasing pursuit of high fidelity, which is typical of the simulation industry and is also supported by current regulations for flight simulator training devices, there is increasing evidence that adding more fidelity beyond a certain point results in a diminished degree of transfer of skills, especially for nonexpert pilots. Indeed, high fidelity also means high complexity, which generally requires more cognitive effort, thus increasing the trainee’s workload, which may, in turn, impede simulator learning.
With the goal to seek more clarity with respect to the relation between fidelity and training effectiveness, a first quasi-transfer-of-training experiment was conducted, in which the simulator's objective fidelity (i.e., the quality of the cueing systems) was the independent variable. Two groups of task-na\"ive learners (a total of twenty-four participants) underwent a hover part-task training program, formulated according to Cognitive Load Theory, an instructional design theory that reflects the way humans process information. The experimental group first trained in a low-fidelity simulator (a Computer Based Trainer at the Max Planck Institute for Byological Cybernetics) and then transferred to a high-fidelity setting (the CyberMotion Simulator at the Max Planck Institute for Byological Cybernetics), while the control group received all its training in the high-fidelity simulator. The two groups were balanced according to participants' manual control skills, which were evaluated through a pre-experimental aptitude test (a two-axes compensatory tracking task). During the evaluation phase, which both groups performed in the high-fidelity simulator, no statistically significant differences were found between the two groups in all the dependent measures. Of course, this does not directly imply that the two simulators are equally effective, as the hover part-task training program likely had a mitigating effect, supporting the idea that the lack of simulator objective fidelity can be compensated by the use of instructional design (i.e., a proper training program tailored to the trainees' needs). This can be verified in future experiments using a third group of task-na\"ive learners, trained with a different hover training program in the low-fidelity simulator who are then transferred to the high-fidelity setting to prove this hypothesis.
This thesis also describes two quasi-transfer-of-training experiments that focused on autorotation and had the same setup (the SIMONA Research Simulator at Delft University of Technology), but used two different helicopter flight mechanics models, characterized by a different level of fidelity. The lower-fidelity model was chosen to gain a simple understanding of the flight dynamics in autorotation, that could then be more easily extended to a higher-fidelity model. These experiments, in which the helicopter dynamics were chosen as the independent variable, were motivated by an example of in-flight-to-in-flight negative transfer of training reported in several helicopter accidents. Indeed, many engine failure accidents result from an apparent loss in rotor performance (different helicopter dynamics), which is unexpected for pilots who only practiced autorotations with a power recovery (i.e., terminating in a hover). For helicopters with free-turbine engines, even in a ground-idle setting, the engine still transmits some power to the rotor. This is a clear example of in-flight-to-in-flight negative transfer of training: practicing power-recovery autorotations (task A) interferes with learning or performing real power-out autorotations (task B) for helicopters with free-turbine engines, due to the fact that there is a crucial mismatch between the helicopter dynamics characteristics in the two task situations. Here, a pilot's mental model of the helicopter is not representative of the actual helicopter, which requires a different control strategy than learned during training.
Experienced helicopter pilots participated in the two quasi-transfer-of-training experiments on autorotation. They were divided in two groups, which were balanced according to participants' background (license type) and experience (flight hours), and performed a straight-in autorotation maneuver with two different helicopter dynamics presented in a different sequence. The two dynamics used in the experiments were selected to require a different level of pilot control compensation. To this end, a sensitivity analysis on the helicopter eigenmodes was performed to understand which design parameters control the autorotative flare index, a metric to evaluate autorotative performance in terms of available energy over required energy and thus influence helicopter dynamics in autorotation. This was achieved through the structural evaluation and comparison of the helicopter natural modes of motion in steady-descent autorotation. Thirty-two configurations were compared by individually varying the main rotor blade chord, the main rotor radius, the main rotorspeed and the helicopter weight from the baseline value (Bo-105 helicopter) to get eight different values of the autorotation index, spanning from 5 to 40 ft3/lb. This range was chosen after comparing the autorotative flare indices for various existing helicopters. Among these configurations, the two requiring the most and the least pilot control compensation were selected.
In the first experiment on autorotation, fourteen pilots performed the straight-in autorotation maneuver controlling a 3-degrees-of-freedom (DOF) longitudinal dynamics + rotor speed DOF model. Ten pilots performed the same task with a 6-DOF rigid-body dynamics + rotor speed DOF model in the second experiment. In both experiments clear positive transfer was found from the most to the least demanding helicopter dynamics, but not the opposite. This is observed especially for the rate of descent at touch-down, which is considered the key indicator of a smooth landing. The outcome of these two experiments suggests the need to update the current simulator training syllabus for autorotation to include a wide range of helicopter configurations with different handling characteristics. Such configurations can be obtained for example considering different models of the same helicopter family, to give to the trainee the opportunity to familiarize with helicopters with different sizes, dynamics and "feel". This can help inexperienced pilots to better understand that an autorotation is not a "by-the-numbers" procedure and that adaptability and judgement of the pilot should always cover a prominent role in the accomplishment of the task.
To strenghten the experiments on autorotation, a thourough analysis was conducted to investigate the effects of the rotorspeed degree-of-freedom in autorotation on the classical rigid-body modes. Although the developed 3-DOF and 6-DOF models are characterized by a different level of fidelity, good agreement in terms of stability characteristics of the longitudinal modes of motion was found between the two models. Especially the phugoid and the heave-subsidence modes are strongly affected by the additional rotorspeed degree of freedom, meaning that autorotation requires a different stabilization strategy by the pilot with respect to straight level flight. On the contrary, the pitch subsidence in both models and the lateral-directional modes in the 6-DOF rigid-body helicopter model do not change significantly in steady-descent in autorotation with respect to straight level flight.
In conclusion, this thesis provides enhanced insight into helicopter pilot training in flight simulators by addressing two critical training tasks, hover (Part I of this thesis) and autorotation (Part II of this thesis), that represent two of a helicopter's most unique capabilities. With these new insights, this thesis lays the foundations for an enhanced understanding of the future requirements for helicopter pilots training in flight simulators, which will become even more important considering the current trends towards Urban Air Mobility. Indeed, the transition from helicopters as a niche sector in the aerospace industry to the widespread future use of personal aerial vehicles (PAVs) based on rotorcraft concepts needs to be accompanied by a disruptive change in aviation regulations, encompassing every aspect of safety, including training. Even though these future PAVs will be characterized by a high level of automation, the human operators will keep playing an important role in the safe operation of the flight, hence raising the need to develop training requirements for PAV pilots.