Thank you for having me today. My name is Tarek Mohammed, and I'm from Kite Mill. I'm also, this work is part of my PhD study in Pune, where it's supervised by Professor Lorenzo Pagano. So, the title is fault-oriented control of airborne wind energy system with multi-copter fixed-wing UAV configuration.

So, the model considered in this study is a model from Kite Mill, five kilowatts. So, this model is ground-based generation, and it is a YOYO principle, and it has rigid kite and VTOL system, vertical takeoff and landing system.

So, the VTOL system is used here just for two transition phases, takeoff and landing. And this method has been used by many companies. So, in one of the EU published report for airborne wind energy technology, in this, they describe the barrier

toward the commercialization of this technology. So, one of the barrier is system safety and reliability. This study tried to make contribution toward solving the problem of system safety and reliability.

So, we look at the system as a multi or over-actuated system, where we have three effects that we want to make, Yaw, pitch and row, but we have many actuators that can do this system. So, the first thing that we try to apply controller location scheme.

So, the controller location scheme we use is a daily chain method, constraint controller location. For example, if you want to go to waypoint, to the north and then to the east, we can do this either by the discrete surface

or the VTOL system. So, we want to make some desired row moment. So, yes, we can define some barometer, B delta aileron, this coefficient, yes, for the desired row moment,

we can know the desired aileron deflection and in case of saturation, then we can, yes, we can allocate the unattainable moment to another effector, like the rotor in this case.

And yes, so, and if we have another effector that can initiate a row moment, we can also, yes, use the daily chain method to make a row moment. So, yes, how to compute this barometer?

We use the model-based method. So, we have the aerodynamic data of the kite, we have the barometer of the thrust or the VTOL system and we can online compute the dynamic control

effective risk matrix, which includes that barometer. This could be computed online. So, why we use the easy chain approach?

Because during flight, we can have a sudden failure in the actuator and if we are able to detect this failure, we can instantly saturate the first control actuator and allocate all the moment to the next actuator

and this is how it simply works. So, for the sake of this controller, we make some residual generator to detect the failure in some actuator. So, this is the general view of the control architecture.

So, we have a feedback controller, normal feedback, linear feedback controller for the flight and for the hoovering phase and we have a switch that can switch between the output from these two controller

based on the pace of the flight and yes, what we should mention here is this feedback controller, the command signal is the moment. What is the desired moment? And then we take this moment and we allocate it first to the surfaces because it is much more cheaper to use

than the VTOL system. And yes, we calculate the dynamic control effectiveness matrix online. Yes, we have the alpha, beta and airspeed. So, and then in case of surface saturation or failure,

we can allocate the desired moment to the VTOL system. So, we did two simulation, one for tethered flight, one for untethered flight. For the tethered flight, we have like to follow some way point,

go to the north and then here to the east. So, this is the result of the simulation. So, the first scenario, we have no any failure in any system. The first column, you have the body velocity in X, Y, Z axis. Then we have here the Euler angles

and here we have the rate, the rolling rate and here we have the position, the both position, the north position, east position and altitude. And we can see that it follow the way point, go to the north and then steer to the east

and maintain the altitude. So, this is how the controller perform in case of no failure. So, this is how the actuator, we have the aileron, elevator, rudder and we put some traction force, just for the sequence of study and this is how the VTOL RPM behave.

This is how the digit chain work. So, then we inject, in the second scenario, we inject an aileron failure. The aileron failure, yes, you can see here,

we just put the aileron to zero degree and as you see the results here, it can go, yes, it still follow the way point, go to the north and then steer to the east with some degradation in performance but it should be acceptable to land the kite

or to glide it to the safe point. Then we did some, so because it's simulation, so we, yes, we did some more failure. So, here we just put all the discrete surfaces to zero and this is how the VTOL system behaves

and this was the performance of the digit chain and the controller scheme we have proposed in this study. It still follow the way point but yes, there are some degradation in performance but it should be acceptable. And this is how the residual generator behave.

So, as you see in this in the case of aileron failure, we have, yes, the residual generator in the x axis, it has some sudden jump but for the other axis,

it's kind of no any sudden jump as here and this in case of three discrete surface failure, how the residual generator behave. So, we applied the second simulation with the tither slide, with the tither extreme segment,

spring dumber element connected in series and the simulation is about just circling in the point above the pulley. So, this how the tither model behave.

So, in this simulation, we have east and north position trajectory. The yellow line is we just give some turning rate

to the controller, we want just to circle. So, this how in the yellow line is how the controller behave in case of no failure. If we inject an aileron, an elevator failure, it will be in the here in this point, this is in the dash straight line.

So, as you see here, it's still the VTOL system can handle the failure and can provide the good turn rate control. But in case of the rudder failure, the kites start to diverge out of the circling turn rate

and that might be because of the VTOL system has not enough yawing authority to keep it in the track. So, this work will be published in the CCTA conference

in the next August. Yes, and the next, yes, the idea basically is to build sophisticated system model and then apply fault-tolerant control scheme. And yes, then go to design a safety design approach for the airborne wind energy system.

If you have any questions. Thank you. We have time for two questions, so, yes, please. Hi, have you assessed performance for a non-zero constant elevator or aileron deflection

for your system response? No, actually, just put it at zero. The failure, the failure that we assessed. But, yeah, yes, we didn't make a simulation for this. Okay, do you expect in a real system most of the failures will result

or have some hardware backup to set it back to zero? But it depends on the system itself. So, if in case of failure, we can add some mechanism to release the rod, maybe the mechanical rod to make it really moving. But other than that, it could be that you've been bitching up or bitching down.

This will be kind of excessive moment to handle. So, but in this case, the perfect thing is just to switch directly to the VTOL system. If you detect such kind of failure and then the disc receptors will have minor effect because it's slow speed. Yeah, another question here? Okay, good.

Maybe last question before closing this session. Yeah, there is a question there. Hi, thank you. I was wondering how you did the online system identification of the control effectiveness? So, it was not a system identification.

It was data from CFD analysis and some tool I use to map the thrust for a given RPM. So, then using this data for a given airspeed, for a given angle of attack and side slip angle, we can estimate the forces, the aerodynamic forces

and the thrust forces from the VTOL system. Okay, thank you. Thank you. Yeah, thank you all for your participation in this session and the speakers also for these nice presentations. Hope to see you in other sessions also. Yeah, it's time to have a break.