Problems in SRM modelling

[Ksp-3086]

I have modelled an SRM in pyleecan GUI. However, I need to run the FEMM simulation to get iron and copper losses and torque results. I have the following problems as of now ( as I am new to python and pyleecan both):

1. How do I define the phase current for switching circuits?
2. How do I calculate losses (both iron and copper losses, if I can get overall losses that would be great too")
3. How do I output the data to a dat file so that I can use the data for future references and in other software like MATLAB?

[SebGue]

Hello Ksp,

from your question I guess there is no such tutorial yet? Would you mind to create a tutorial on that if I guide you through?
Best regards, Sebastian

[Ksp-3086]

Yeah, sure that would be helpful. I do have a slight idea on how to do the step functions (as SRM works on that) but I have not worked on python programming, so yeah it would be a huge help.

[SebGue]

Okay, then I think it would be best to start with the tutorial on FEMM simlations. There you already got individual phase currents defined as function of time. You only need to replace them.
Since I don't know your actual current waveform there may be several ways to define them, depending on your pereference (e.g. with fourier series or step functions or ...). In any case I would recommand to define a function that returns the current waveform, so you can reuse it and your code is more readable. E.g.

def rect_wave(time, freq, width, phase):
    current = ...
    return current

Then with the currents you can run your simulation and check if you get the expected torque values.
We will address the other points after that.

OT: Do you have further experiance in SRM expecially in 6 phase machines for sinewave drives?

[Ksp-3086]

Ok, I have gone through the tutorial and yeah I think I need a rectangular wave form as a function of time. I just have some problem defining the rectangular waveform. I do not have much experience in SRM and 6 phase machines for sinewave drives.

Also, pyleecan does not let me define more or less than 3 phases for SRM, it always gives out an error. I am also a bit confused what does pole pairs mean here? As far as I remember pole pairs meant N-S pairs, but here that does not seem to be the case

[SebGue]

Oh, it's even easier than I thought. There is a dedicated numpy function for rectangular signals.
Aside from my question, do you need other than 3 phases now? Maybe you can post your current code to make your questions more clear.
Edit: ... pole pairs should of course be N-S pairs.

[Ksp-3086]

I think I do not need more than 3 phases for now but it does give an error of I need to increase it in the future. The error is

Also about the pole pairs take a look at the images where I used 2,6 and 12 pole pairs. Aside from the change in position of the windings there was not much change which is why I am confused.



The dedicated numpy function is for square waveforms as it is shown on the site as I can see and not for the rectangular one. For this tutorial, I think we would need the current in each phase to be a positive constant value for 1 cycle and 0 for the next two. So that is somewhat different

[Ksp-3086]

Hi @SebGue ,
I got the function for the rectangular wave and I put it in. As I am new to python, how can I return the value needed as you can see in the code below.

[SebGue]

For the multiphase error, I have to check if the winding generator is able to generate such winding. Anyway you can allways setup an user winding with phase count other than 3.

The windings for the different pole pair numbers seem to be valid. What did you expect there?

Side note: The machine with 12 pole pairs would not be used with 'normal' machines due to its low winding factor. Is this a valid pole/slot combination in SRM?

If you offset the numpy rect function and use a duty cycle of 33% it should also work.
But you fourier series is also okay. You only got to init the x variable e.g. with numpys zeros or zeros_like functions.
Further for indexing in python you need to use square brackets [ ]

Hope that help you for now.

[Ksp-3086]

How do I setup phase count other than 3? Is it by changing the code in json file?

Visually, the pole pairs kind of look the same. I think I am just confused in that part like, the dotted represents current going inwards and the other one outwards. If I set pole pairs to 2, then there should be only 2 N-S pairs but there seem to be 3 for each phase (like 3 dotted and 3 checkered)

12 pole pairs is not valid, I was just playing with the GUI to see different outcomes or if it gives an error.

Also, I was not able to find a proper documentation on rect function

[Ksp-3086]

So, what does this error mean exactly

[SebGue]

Hi, I think there is some mismatch in terminology here. For non SRM the stator allways has a pole count other than the slot count. For SRM I guess the slot respectively tooth count is equal to the pole count?!

So what is the topology you actually want to design for now, i.e. how many rotor, stator slots and how many coils/phases do you want to have?

(For phase count other than 3, I would first design some similar winding with 3 phases with the generator, export the csv, edit it to my needs and import it as a user winding then.)

With 'rect' I meant the rectangular signal mentioned above.

[SebGue]

... sorry, just saw that the rectangular signal function is not of numpy but scipy.

[Ksp-3086]

As the SRM is outer rotor and for an automobile, stator slots-18, rotor slots-12, and phases -6 ( which I was not able to do due to the error shown earlier). I could share the json file but I do not know how to do it

[SebGue]

If you are on the website for writing comments, you just need to drag your file into the comment box.

[Ksp-3086]

SRM_1_outer_rotor.zip
Here only zip file upload was possible

[SebGue]

Hello, you can import the attached csv on an user winding. I don't know if this is the winding arrangement you want to achive. But if you open the csv file you will see how it works and how you can setup your own winding.
SRM_1_outer_rotor_Winding.csv

[SebGue]

I will also update the import, so you can also import fractions of winding definition later. So repeating arrangements will be easier to setup.

[Ksp-3086]

So, I was going through the csv file you sent me. I understood that in (1,2,18,6) 2 means pole pair, 18 means no of slots, 6 means- no of phases but what does 1 mean? The rad layer and tan layer mean radial and tangential layer of winding, right?

Hi, I think there is some mismatch in terminology here. For non SRM the stator allways has a pole count other than the slot count. For SRM I guess the slot respectively tooth count is equal to the pole count?!

So what is the topology you actually want to design for now, i.e. how many rotor, stator slots and how many coils/phases do you want to have?

(For phase count other than 3, I would first design some similar winding with 3 phases with the generator, export the csv, edit it to my needs and import it as a user winding then.)

With 'rect' I meant the rectangular signal mentioned above.

Also you mentioned a mismatch of terminology, for SRM, the number of poles on stator must be more than no of slots on rotor to prevent locking as far as I know. I read it in a research paper while I was gathering some information on them. So, how do I approach this problem to rectify the error

[Ksp-3086]

SRM_1.zip
here is the code I have written so far

[SebGue]

The tuple (1, 2, 18, 6) is defining the winding matrix size. The meaning should be (no. rad. layers, no. tan. layers, no. slots, no. phases).

I haven't run your code but it seems reasonable. I only would propose the following changes to rect_wav:

def rect_wav(A, time, T, tau, phi=0):
    d = tau / T
    x = zeros_like(time)
    for n in range(len(time)):
        x[n] = (A * d) + (2 * A / d) * (
            sin(pi * n * d) * cos(2 * pi * n * (time[n] + phi) / T) / n
        )
    return x

Atleast you have to inline 'return' with 'for'. With the 'phi' parameter you only got to define one time vector.
I also changed your indexing to 0 based as this is python standard. I hope the 1 based index was not by intention.

[Ksp-3086]

Yes, the 1 based was not as I am used to MATLAB coding more so it is kind of new to me. Can you tell me what to do about the error of lamination I am getting, as I am not able to run the simulation due to it.

[SebGue]

Missed that error....
You have to set the stator polepairs = rotor polepairs, even if that maybe not valid in terms of SRM terminology.

[Ksp-3086]

ok, and where do I set that I mean how? like in the library for SRM or somewhere?

and one more thing, I plotted the wave, it doesn't look rectangular to me. I set A=10 in this case. Do you think it is still fine?

[BonneelP]

Hello,

I have opened a PR related to the pole pair number for SRM: #430
I have started by removing the check on the get_pole_pair_number for SRM (and for now I return stator.winding.p)

For the definition of pole pair for SRM, would it be helpful to return p=rotor.Zs/stator.winding.qs ? Or should we return stator.winding.p ? (I'm not an expert of this kind of machine)

We also need to define the "set_pole_pair_number". For now it would set stator.winding.p=p and rotor.Zs = 2*p. As p is required to define the winding we could add a "set_pole_pair_number" method to MachineSRM that would only set stator.winding.p (and won't change rotor.Zs).
We can also hide the p widget on the first step for SRM machines and add a p widget on the winding definition step.

What do you think would be the best solution ? I plan to work on the release tomorrow so we should be able to include these modifications.

Best regards,
Pierre

[Ksp-3086]

I am also not that big of an expert but I do know these points but I think you should return the "statorwinding.p" for pole pair definition.

The p widget gets me confusing sometimes (like you can see in the above images #428 (comment)) , I do think it is better to add it on winding definition step. I am not sure how the code fully works but what does "set_pole_pair_number" refer to? Is it stator poles or the rotor poles. Both give different meanings so its a little vague for me.

I don't think that no. of poles on stator affect Zs or slots on rotor. I have read about 12/6, 12/8 and 12/10 SRM, so I think that option is not needed.

Also, a side suggestion. If possible add an option about singly salient and doubly salient construction. For the singly salient construction no. of pole pairs*2=no of slots but in doubly salient construction no. of pole pairs= no of slots. It would be easier if this is included although you may already know that.

[BonneelP]

The p widget gets me confusing sometimes (like you can see in the above images #428 (comment)) , I do think it is better to add it on winding definition step. I am not sure how the code fully works but what does "set_pole_pair_number" refer to? Is it stator poles or the rotor poles. Both give different meanings so its a little vague for me.

For most of the machine, we have to enforce the rotor and stator pole pair number matches. This is why the option to select p is set on the "Machine Type" step. The widget on this step calls "set_pole_pair_number" which differ from one machine to another. So the simplest solution is to define that "set_pole_pair_number" for MachineSRM to only set stator.winding.p.

We still have to set stator.winding.p since it is required by the star of slot algorithm to generate the winding pattern.

Also, a side suggestion. If possible add an option about singly salient and doubly salient construction. For the singly salient construction no. of pole pairs*2=no of slots but in doubly salient construction no. of pole pairs= no of slots. It would be easier if this is included although you may already know that.

Can you provide us with schematics/screen shots/publications... about the machines you are trying to define/simulate (if there are not confidential) ?

Best regards,
Pierre

[Ksp-3086]

Singly salient type SRM

and more widely used doubly salient

[Ksp-3086]

Hi all,
I thought about decreasing the pulse width (the on-off of a circuit). The graph can be seen like this

The ripple has been reduced but the torque is all negative

Looking back at the code of the tutorial, I saw rot_dir and comp_rot_dir, I think these mean the rotation direction but of what current or the rotor? I am not sure how to use them.

[EmileDvs]

Hello,

At the end we went with:

* MachineSRM.get_pole_pair_number = self.rotor.get_Zs()

* stator.winding.p = stator.slot.Zs // stator.winding.qs

If I understood/remember correctly what Emile suggested, for comp_felec we should use machine.get_pole_pair_number() (which anyway makes more sense for MachineUD). I will correct it in #437.

Best regards,
Pierre

Yes exactly the frequency of pulse in each phase should be Zr*N0/60 as returned by comp_felec() since get_pole_pair_number() now returns Zr (number of rotor slots) for SRM.

@Ksp-3086 The torque curve still seems very distorted in my opinion, as you said in your update the rotor initial position seems not to be the best. Assuming you have phase A on first stator tooth, and you start feeding phase A with DC current, the middle of the first rotor slot should be aligned with the middle of the first tooth. That way, the first stator tooth fed by phase A current should attract the rotor tooth coming after the first rotor slot.

Concerning torque sign, it is indeed related to the rotor direction. For AC machines, rot_dir is directly deduced in motor convention by comp_rot_dir method, which calculate unit mmf of stator winding and deduce the rotation direction of the fundamental field. For your case I think it should only follow the winding phase pattern. If you feed phase A-B-C etc., the rotor should move in the direction to be successively in front of stator teeth fed by A-B-C etc.

Best regards,
Emile

[Ksp-3086]

Hi @EmileDvs ,
So, According to the current graph in which every subplot shows current in different phases, the rotor should follow the direction in clockwise if I am correct. What I am not able to understand is why the ripple and negative values if it is rotating in the clockwise direction (which I am assuming the code does by default, if it doesn't then please tell me).

First I thought that a high number of rotor poles was the problem and I decreased them, only the amount of ripple is lessened and it has become negative.

[Ksp-3086]

Hi all,
I was trying different configurations and these are some main problems that I am facing.
I had simulated this design in ANSYS first. (like a month before)

At 1500 rpm I get the torque graph as

The losses are

The current phases graph can be seen as

although the same code gives these subplots in MATLAB

The main problems in every design are

1. The SRMs are not able to run at high rpm no matter what the design, I have tried reducing poles and everything. I think it is mainly a problem of FEMM (I am not so sure though)
2. The winding losses are huge for SRMs. In every design I made I got 6 or 5 figures digit for losses even though the output power was 3-4 digits. I do not know how exactly are they being calculated here but it seems a problem.

It can be entirely possible that these two problems are arising due to the input current I am giving them but input current only has values of Amplitude A or 0 (to make it a rectangular wave). What do you guys think?

[SebGue]

Hello Ksp,

for your first point "The SRMs are not able to run ..." I don't know what you mean by that.
The winding losses are easy to validate by I²R. So you got to check your winding (resistance).

To check the simulation I would start with a constant (DC) current in one of the phases. The result should be an alternating torque curve. There you can check if the number of periods is correct. Also if your currents phase is right. Further you can easily validate the current losses.

If you still get invalid results you may disable periodicity first. Maybe there is still some issue with that for SRM?

Regarding the current waveforms, I'm wondering why there is a time scale of 1 second? If you got 1500 rpm the time of one period should be way shorter. (The triangular shape is because of the sampling of the signal where you only got 32 sampling points I guess. But maybe there is some staircase type of plot for better visualization.)

Seeing forward to see some results :-)

Best regards, Sebastian

[Ksp-3086]

Hi Sebastian,
My apologies for being somewhat ambiguous on "The SRMs are not able to run.." . I mean that at high rpm there is no torque, it is a straight line. In my case at somewhat 1800 rpm.

As suggested by you I ran a constant DC Current in just one of the phases, on disabling and enabling the periodicity of the motor. I got the same torque results. It alternates for some amount of time then it becomes 0.

The total time of simulation is 1 second and you are right the time period must be shorter and it is. As you can see I am getting a emulation of a rectangular waveform. The higher the rpm the more the peaks or in this case the positive values of current appear in one second.

What are your suggestions on the time-torque curve?

[SebGue]

I can't give you a suggestion since your torque curve suffers from aliasing. You got to set some sufficient time period and/or number of time steps first. You should have at least 4 steps per pole pair and per rotor rotation.

[Ksp-3086]

Right now, this is the type of motor that I am trying to run. It has 2 pole pairs per phase. As you can see there are 4 phases, that means 4x4=16 time steps minimum.

The time period is calculated from the code.

As you can see that the time period depends upon the number of phases and the rpm. The number of time steps set by me is 32 right now over a iteration of 1 second. How much would you recommend me to increase these values (time steps and amount of time simulated) so that results are somewhat favorable?

[SebGue]

Right, with your T (=qs*tau) of 0.03s you will simulate 1 electrical period, i.e. 1/2 rotor rotation which is fine. (Don't know why you want to simulate 1 s of time. That would require approx. the 30 fold number of time steps and give no additional information.)

[Ksp-3086]

I did not actually choose 1 sec for any profound reason. I was just using the value for all the other designs I made and just making changes to number of time steps needed.

So, you are saying that number of time steps is fine. I thought it was a bit extra. So, do you have any suggestions?

[SebGue]

The number of time steps depend on the details you want to resolve. But 32 for a half rotation should be okay for now to check if your simulation settings are valid.

[Ksp-3086]

So the torque graph at high speeds is not favourable I am guessing something is wrong but I am unable to actually figure out what. Why does it drop to 0 at high speeds?

[SebGue]

Therefore you should do the 'DC current test' with e.g. 0.03 s period and 32 steps @ 1000 rpm first. After you get valid results, you can also increase speed (reduce period) and see if results stay valid.

[Ksp-3086]

Yes, I have tried that. I have sent a graph of exactly the same input values you mentioned. These do not look valid to me. First some amount of torque and then 0.

[SebGue]

Okay, then you should attach your current files (simulation script and the machine). Have you checked your current twice?

[Ksp-3086]

code.zip
These are the files I am using. There are two json files. SRM_innerrotor has 3 phases and SRM_innerrotor1 has 4 phases. I have checked my current too right now I was using 200 Amps.

[SebGue]

I haven't started your simulation, but what I see if I break at simu_SRM.mag = MagFEMM(... is, that your current only has 200 amps for the first 2 time steps. (Besides you are using 100 time steps here.)

[SebGue]

... I think the only thing to do is to fix the time1
time1 = linspace(0, T, 32)

[Ksp-3086]

So at high speeds my code is not able to compute the current. Thanks for the insight I could not have figured that out. I can try another way to get a rectangular wave. I am still using your suggestion time1=linspace(0, T, 32). It is not a full rectangular wave but a full time period, I guess it should be enough. I will get back to you with the results.

[SebGue]

I don't know what you mean by "It is not a full rectangular wave but a full time period, ..".

For the high speeds you actually don't need them to be calculated, since you only got a static FEA calculation.
You can do a single simulation at arbitrary speed and calculate any other desired speeds from that results. (Thats also the reason the iron loss model got a speed override parameter.)

[SebGue]

So I ran you simulation and the results look valid to me, i.e. there are 3 torque periods (per half turn) due to the 6 rotor pole pairs.

Now with this curve you can also choose some optimal phase shift of your current wave forms.

[SebGue]

... for the winding losses, they seem also valid (if you have a look at the single phase losses loss_st_wdg = loss_stator_wind.get_along("time", "phase")["Loss"])
It shows around 200.000 W for phase A. Since your winding resistance is roundabout 5 Ohm and your current is 200 Amp, P = 200²*5 = 200.000 holds.
As I said before, you got to check your winding for that issue.

BTW the following code will improve the loss calculation for asymmetric currents.

loss_st_wdg = loss_stator_wind.get_along("time", "phase=sum")["Loss"].mean()
print(f"stator winding loss = {loss_st_wdg} W")

[Ksp-3086]

Yeah, I did the calculations using the resistivity in the file and the length and area. I agree. I will need to increase the area and decrease the number of turns. Some questions that I have

like what is Kfill here. I made out that it was the percentage of the area covered by the winding in the slot against the total area. Sometimes it actually bugs out and does not change though.

@BonneelP Some suggestions for the next update for the GUI mainly

1. If it is possible that details to every parameter can be just written instead of being shown when hovering a cursor over them. I mean right now to see what a parameter means, we hover mouse cursor on the parameter. Some values which are calculated like Kfill are not mentioned fully what they mean. So like it could be better for starters to like know about that.
2. Loading json files of motors in the GUI to have a better look at the motor designed. (It could be helpful to those who are just starting the software)

[Ksp-3086]

Hi all,
Reducing the number of poles on the rotor, I get some favorable results while running a 2/6 pole SRM.

The ripple is still there. If I increase the number of poles to 4 on this rotor, I get negative torque

Maybe the pulse length I am taking is too long or the starting angle is not as I think it is. Any suggestions?

One more thing, there is something else that I noticed, that when running the motor at high rpm, it is supposed to produce less torque due to back emf but this was not the case, I got same graphs at low and high rpm.

[SebGue]

Hello Ksp,

actually your results doesn't look too favorable to me.
I did some slight modification to your recently attached simulation and get the following results.

This is no offend but I think you still have to do some research on how electrical machines (and SRM in particular) work. Then you will also find the issue in your simulation easily. It's in your current waveform.

BTW I was wrong with "... with your T (=qs*tau) of 0.03s you will simulate 1 electrical period, i.e. 1/2 rotor rotation which is fine.", since it isn't 1 electrical period. It is only 1/2 rotation.

Could you tell why the machine should produce less torque with high rpm due to emf. I have a guess what you mean but that doesn't apply to this simulation.

In general all the recent discussion isn't related to a pyleecan issue anymore so my proposal would be to close the issue and e.g. open a dedicated 'Discussion' which could also be linked to this issue.

Best regards, Sebastian

[Ksp-3086]

Hello Sebastian,
Yeah you are right that there is still a lot of research that I need to do to see how electrical machines work as my bachelor was in mechanical so I got only some amount of knowledge on E Machines. Can you tell me what modification you did?

[SebGue]

I set T to the 1/p-th fraction of a rotation and tau to T/qs. This is because an electrical period is the time between one pole is aligned with given phase and the next pole is aligned with this phase.
Further you got check the linspace 'endpoint' parameter in regards of periodicity.

[SebGue]

,,, and added slight phase shift.

[Ksp-3086]

Hello Sebastian,

As you said I am closing this issue and opening a discussion #449 . I am going to ask you a question on your method as I am still not fully clear what I did wrong and what exactly you did?