Development of a Series Hybrid Multirotor

Hybrid-electric power is an appealing technology for multirotor platforms due to its ability to enhance the range of the vehicle while providing low emissions and the precise thrust control required for vehicle stability. The development of a multirotor utilizing hybrid propulsion is an essential step for CfAR as it will be the basis of research on this technology in flight. The designed multirotor, MIMIQ (Modular Inertia Matching Quadcopter), is a 32kg quadcopter with a motor-to-motor diameter of 2.7m which will require a total of 3,700W of power to hover. This demand is primarily met by the series hybrid generator onboard which requires the battery to only supplement a portion of that power. At peak thrust however, the motors will demand 12,000W from the system. During these rapid bursts of energy the battery plays a critical roll in providing an immediate response to the power demand which the generator would not be able to provide otherwise. This multirotor is designed to accommodate a wide range of propulsion configurations to optimize performance. In addition, the mechanical characteristics are also intended to be easily adjustable. Parameters such as center of mass, inertia, and motor distance can be adjusted to mimic characteristics of future flight vehicles that implement hybrid systems. Before integration, ground tests were performed on various off-the-shelf hybrid generators in order to evaluate their performance and reliability to ensure that they are well suited for the MIMIQ. To characterize the different modes of operation an extensive performance map was pursued. A number of parameters were measured over time with the most important being load, power, and fuel consumption. It has been proven that at low loads the generator can provide power to the propulsion system while charging the onboard batteries. However, at higher loads both the generator and the battery pack are required to provide power. The tests demonstrated that the integration of a hybrid system into MIMIQ is possible. It was also verified that when the engine is integrated into MIMIQ, both the generator and batteries have to provide power simultaneously in all flight phases.


Introduction and Background
A hybrid electric propulsion system (HEPS) is characterized by the use of electric power in combination with another type of power, typically an internal combustion engine (ICE).HEPS provides the designer the option to exploit the advantages of each type of power while minimizing their weaknesses to increase the efficiency of the vehicle [1].There are two main configurations of HEPS, parallel and series architecture.The scope of this paper will discuss the design and modelling of a custom series HEPS and the integration of a commercial unit into a custom multirotor.
In a series system, the ICE supplies mechanical power to a generator motor which transforms it into electrical power.The power can then be used directly by the electric propulsion motors or stored in batteries.This architecture has been widely studied in recent times and has several advantages over parallel HEPS.The key advantages stem from the simplicity of the system and it's ease of integration due to the engine being directly coupled to the generator motor.It is shown in literature that while the series hybrid propulsion system is 20-40% heavier than a standard system, it will still yield a fuel consumption reduction of 13-24% [2].This same paper also demonstrated that it is key to optimize ICE sizing to the power consumption and operational altitude as the degree of hybridization is sensitive to how it is sized to the platform.It is shown that parallel HEPS does provide better performance than series [2][3] with a fuel consumption decrease of 30-50% in comparison to a standard propulsion system.The increased performance of parallel is attractive [2], however the considerable reduction in complexity that series offers makes it a much more desirable candidate for a HEPS propulsion system on a multirotor.
The majority of work in literature is focused around the development of theoretical HEPS models to compare various configurations and degrees of hybridization.As the technology has progressed through the years, several off-the-shelf units have come into production from various manufacturers.Due to this, experimental flight data to be acquired relatively easily from which better series HEPS models can be built.In order to integrate these systems, CfAR has designed the MIMIQ (Modular Inertia Matching Quadcopter) flight test vehicle which primarily serves as a test platform for these HEPS technologies.In addition, the motor-propeller placement, MTOW, and propulsion specifications is intended to replicate future large format VTOL projects.The design and manufacturing of a large format multirotor is predicted to be an invaluable stepping stone to developing future VTOL and multirotor vehicles that could benefit from the integration HEPS technologies.
During the design of the MIMIQ, a similar project for an unmanned vertical take-off aircraft with a significant load handling capacity was launched at the Institute of Aviation in Warsaw (ILOT).The Unmanned Aerial Vehicle (UAV) was intended to be powered by a series hybrid engine.During the concept phase, the payload underwent several changes.These amendments necessitated the creation of a numerical instrument for designing a suitable propulsion system for the aerial mechanism.The numerical model was intended to facilitate the creation of future hybrid designs.The amount of data available, especially for low volume combustion engines was not sufficient enough to predict it's performance, generated power, torque, and fuel usage.The selected engine to build this model from (Desert Aircraft DA-70) had provided a relatively small amount of data.It contains maximum torque and power curves at open throttle in addition to a set of fuel usage data, all of them as function of rotational velocity.Thus the simulation model had to be supplemented with theoretical dependencies for other work scenarios.
The aforementioned observations prompted the creation of a test bench to explore the hybrid system's attributes.At the outset, the strategy called for designing a low-power motor-based system, of which the characteristics are typically the least documented.By pursuing this tactic, it became feasible to amass the maximum amount of knowledge about low-power hybrid propulsion.

Design Overview and Initial Constraints for designed Multirotor
The design of the MIMIQ project had many driving factors.The primary goal is to use the platform to study and analyze hybrid series generators to determine how the performance of the units could be optimized to maximize flight time and efficiency.An additional goal however was to match several key characteristics of a future VTOL project.Characteristics such as propulsion system design, MTOW, and motor placement in order to replicate how the vehicle would behave using hybrid technologies.The first goal required modularity so that the MIMIQ could accommodate several different generator systems while the latter goal drove design decisions in air frame sizing, motor selection, and avionics.
Initially the MTOW of the MIMIQ was desired to be 25kg, however during the conceptual design phase it was realized that the MTOW would need to be closer to 30kg.Motor sizing for MIMIQ was done by bench testing several combinations of motors and propellers.The T-Motor MN805-S 120KV with 30x10.5 carbon fiber propellers [4] was determined to be the best combination for the needs of MIMIQ.At 52.7V the motors produced a maximum thrust of 166.5N each, this results in a peak thrust to weight ratio of 2.27.At hover the total power consumption is estimated to be 3,700W.For most generators this was determined to be a charge depleting configuration where the generator cannot solely sustain hover and requires the battery to slowly deplete.Generators were selected primarily based on capability and power output, but metrics such as weight and fuel efficiency were also considered to ensure that the HEPS would fit within the air frame.The primary generator candidates were the Pegasus Aeronautics GE70 [5], Currawong Engineering Cortex-50 [6], and Löweheiser HYBGEN 32 [7].To begin initial testing, the Pegasus Aeronautics and Currawong units were purchased.The GE70 power output far surpassed any other generator on the market and produced 3,500W.The Cortex-50 was quoted to produce only 2,200W, however had extensive onboard telemetry that would greatly benefit research efforts.In comparison, while the GE70 offers better performance, the onboard telemetry is far less accessible than the Cortex-50 which is why although less powerful the Cortex-50 was considered to be much better suited for a research platform.For this reason it was decided that the Cortex-50 generator would be integrated into MIMIQ for the first round of flight testing.In this configuration the estimated endurance of the MIMIQ is expected to be 50 minutes which as mentioned is limited by battery capacity due to the generator not being able to solely sustain hover.Due to this charge depleting configuration, a 14s 23,000mAh lithium polymer (LiPo) battery pack is required to sustain the estimated 50 minutes of flight.
The airframe of MIMIQ is built using primarily carbon fiber tubes and plates to ensure rigidity of the structure.Key vehicle parameters of MIMIQ are shown below in table 1 with a mass breakdown shown in figure 1  It is clear based on preliminary calculations that the balance between the goals of the project created a MIMIQ that was oversized for the Cortex-50 which resulted in a limited flight time.
An airframe that was optimized for the size of the generator could operate charge sustaining rather than charge depleting which would greatly increase flight times.However, the goal of the project was not to perfectly pair the airframe and generator but to study hybrid systems in flight to predict their behaviour on multirotor and VTOL vehicles.From this data, future optimization strategies could then be derived.For a hybrid multirotor, 50 minutes of flight is not long, however it does provide more than enough time to research generator performance in flight and was deemed suitable for CfAR's application.

Experimental Setup
Prior to integration, the Cortex-50 underwent a series of bench tests by CfAR to assess the performance and reliability of the generator.The goal of bench testing was to validate manufacturer metrics of generator performance and to understand how the engine operates during starting and power generation.
Contrary to Currawong Engineering's recommendation of using a 14s battery [6], two 6s Tattu LiPo batteries were used in series to effectively create a 16,000mAh 12s battery.14s was recommended due to the generator motor KV and RPM being optimized for a 14s voltage rather than 12s.This would result in additional power generated that could not be extracted using 12s.Initial tests were all performed at 12s, however future tests are planned that will utilize 14s packs to determine what additional benefits can be found.The Array 3756A 5kW DC programmable electronic load was used to demand power from the generator during testing.This made it possible to apply an electrical load to the generator and provided a way to easily adjust the load requirements between tests.CEquip was the primary control interface to the engine and is developed by Currawong specially for this application.It is used to control the Cortex-50 and obtain test data as it allows for the simultaneous monitoring of several different systems and provides real-time graphics in addition to logging all telemetry data.The brushless motor is driven by the MGM HBCi-20063 electronic speed controller.The electronic speed controller facilitates the operation of the brushless motor as a starter for the internal combustion engine and as a generator.The circuit has been engineered to facilitate dual power inputs which accommodates either battery power or a bi-directional power supply as needed.The data acquisition system utilised National Instruments cRIO-9045 real-time controller as its foundation.The majority of measurements are conducted using 4.20mA current signals which reduced the impact of electromagnetic interference on the measurements.

Bench Testing
The first round of CfAR testing consisted of load tests that were aimed at understanding how the Cortex-50 operated during different power regions.A load ramp test was performed where the load applied was within the region where the engine can supply the entirety of the load power while still charging the battery.This provided an understanding of how the generator power modulation unit (PMU) handles battery charging.A high load test was also done to maximize the generator output to therefore demand power from the batteries to assist in supplying power to the load.This test verified the performance of the electrical harness under high load and provided additional detail into how the PMU operates in this power band.Figure 6 and figure 7 detail the power flow within the systems during testing.The grey line represents the load that was on the system which was applied using the programmable load.The blue and orange lines represent battery and generator power respectively.Negative values represent power entering the PMU while positive represents power leaving.Vibration testing was also performed with the engine integrated into MIMIQ while the vehicle was suspended from each motor using bungee cords.This aimed to determine what frequencies the structure resonates at from engine vibrations and whether the rigidity of the airframe would be enough to withstand resonance without affecting autopilot controllability.The primary concern is that the autopilot will over correct to structural resonant modes and will aim to compensate with the motors.If this happens it could further drive the vibrations and result in a IOP Publishing doi:10.1088/1742-6596/2716/1/0120216 loss of control.By testing the generator at different power output levels the resonant peaks were mapped to determine if filters could then be applied at select frequencies to ensure the autopilot does not react to engine induced structural vibrations.An indication that the airframe is not stiff enough would be if resonance was observed at angular frequencies below 20Hz.If this was the case it would be challenging to discern resonance from vehicle maneuvers and could result in controllability issues if filters are implemented.To test this, the IMU onboard the autopilot was used and data was recorded internally.To load the generator, a similar method to the load ramp is used.The programmable load was set to increase the load by 100W every 10 seconds from idle to 2,400W so the generator power output could be correlated to structural vibration.
To test the ILOT HEPS test rig, it was connected to a bi-directional power source equipped with an emulator for batteries.Tests were performed at increasing throttle opening levels denoted in the percentages 30%, 40%, 50%, 60%, and 70%.At each throttle opening level measurements were taken at no load and at three load levels which corresponded to successive values of charging power.Results collected from testing were used for numerical model validation.

Discussion
Through load testing the Cortex-50 using a 12s battery pack the following performance specifications summarized in table 4 were found.
In figure 6 the battery charge power can be seen to taper off as the load on the engine begins to saturate the throttle range at 2,300W of load.Before that however, the engine was able to manage battery charging while still supporting the load demand on the system.In the load ramp test, the battery is not at full capacity and therefore the generator supplies power.In the high load test however, the battery is nearly at full capacity.The PMU recognizes this and rather than charging the battery it instead pulls power from it to supplement the load while operating the engine at a lower throttle to meet the load demand even though it is under 2,300W as shown in figure 7. Once the load demand does exceed 2,300W, the load saturates the throttle range and the battery is immediately requested to supply the additional power.Figure 7 shows a small section of the generator power plot that is positive at the start if the test.This represents the PMU providing power to the generator motor to start the engine.Engine vibrations proved to not be a concern as demonstrated during vibration testing.
Figure 8 shows the acceleration measurements measured and processed by the autopilot during testing.The angular acceleration fast Fourier transform plot (FFT) indicated that the largest structural resonance occurred at 110Hz with the earliest sign of resonance at 60Hz.The results from vibration testing demonstrated that the airframe can withstand engine vibrations without causing controllability issues.The acceleration power spectral density plot does indicate low frequency linear modes, however this is not as much of a concern as it is easier for the autopilot to handle.
Through testing, it is clear that the Cortex-50 is a good engine for research.It is highly configurable, provides a reliable and comprehensive stream of telemetry, and can support 50 minutes of MIMIQ flight time.That said, it is also clear that for this vehicle it is under powered.While it can support a reasonable multirotor flight time, the airframe mass is not optimal for the size of the engine.This can be seen when comparing the endurance to an electric only configuration and then to the higher power, better suited GE70 configuration.Using the same battery and propulsion system, the estimated endurance for a 23.3kg electric only vehicle is 29 minutes whereas preliminary estimations show that the GE70 can far surpass this with a estimated flight time of 2 hours.The Cortex-50 is expected to be better suited for a vehicle with an MTOW of 20kg where it achieves similar performance to the GE70 configuration with 1 hour 45 minutes of flight time.Figure 9 shows MIMIQ fully assembled with the Cortex-50 and truss assembly visible.To validate the ILOT simulation tool, similar test runs were made on both the test bench and in the simulation tool.This allowed the plots of both approaches to be overlaid for visual correlation.It was shown that the higher the throttle opening level, the better the correlation.In addition, lowering the generator load showed better results.This is expected to be related to the lack of manufacturer data available for scenarios other than propeller load where the load can self-adapt to available power.This is in contrast to a HEPS application where the generator applied load is defined from the control unit.In the numerical model, the charging current was used as reference for control.

Conclusion and Future Work
It is clear from market research, bench testing, and the custom development of flight platforms and generator units that series HEPS show promise for the application of extending the range of multirotor vehicles.In addition, since these systems are self contained, integration onto flight platforms is and has been demonstrated to be relatively easy when compared to the benefits they provide.The simplicity of these systems will enable the MIMIQ to fly a variety of off the shelf HEPS technologies therefore maximizing the experimental data it is able to collect.As expected, it is clear after testing the Cortex-50 that it is under powered for the size of the MIMIQ platform.Based on test data for the propulsion system and generator, an optimal multirotor MTOW would be 20kg to result in a charge sustaining configuration where the endurance is limited by the fuel capacity, this allows the battery pack to be relatively small.This configuration would result in a flight time of 1 hour and 45 minutes in comparison to the current 32kg configuration where the flight time is 50 minutes.Even in an oversized airframe however, it is clear upon testing and preliminary estimations that the integration of a HEPS system clearly increases flight time substantially when compared to a fully electric configuration.
Future rounds of CfAR ground testing will be done using a 14s pack before flight testing to determine how much additional power can be achieved.During flight test campaigns, initial rounds will be done electric only to characterize the vehicle and tune its control parameters.Once completed, flight testing will commence using the hybrid generator where flight data can be recorded and analyzed to determine how the system performs.
Current state of work at numerical simulation tool resulted in better understanding of interaction between components of HEPS.Not much data is needed for preliminary high power output performance prediction, but to calculate some data, such as fuel usage and low power operation, additional data (more than manufacturer provides) for combustion engine is needed.For example fuel usage is not a simple function of rotational velocity and throttle opening, when in generating mode engine load can be selected arbitrarily.
To further progress the ILOT simulation model development, a wider variety of test scenarios are to be tested (more steady state testing as well as dynamic testing).To facilitate the development of the numerical model, additional work is scheduled to outfit the stand with power receivers.Additionally, four test benches will be incorporated, complete with drive motors and propellers which will enable the hybrid engine to replicate the interactions of the propulsion motors.This setup will simulate a quadcopter that utilizes a series HEPS in a controlled laboratory test environment.Further development of the custom hybrid stand involves additional system testing and adjustments for flying platform applications.Future testing will utilise battery power which will facilitate the completion of the numerical model.

Figure 1 .
Mass distribution of MIMIQ in kilograms.

Figure 2 .
Configuration of CfAR test bench.The test rig prepared by the ILOT team to validate the numerical tool and collect data for low-power HEPS consisted of a 2-cylinder, 2-stroke engine (Desert Aircraft DA-70)[8] and a T-Motor V10 160KV brushless DC motor[9].The motors have been mounted on platforms with rail guides which allows for the test stand to be adaptable.This allows easy disassembly of the individual components of the hybrid system.

Figure 6 .
Figure 6.Power flow of load ramp test.Figure 7. Power flow of high load test.

Figure 7 .
Figure 6.Power flow of load ramp test.Figure 7. Power flow of high load test.

Table 3 .
ILOT test bench summary Figure 3. Configuration of ILOT test bench.