Making of a Tutorial on SRM with losses.

[Ksp-3086]

This discussion is a continuation for the #428 for the SRMs.

[Ksp-3086]

Hello @SebGue ,
You mentioned that you put T=1/p th of a rotation. Is it the rotation in one second or one minute as it depends on the rpm set for the machine. If you are referring to the rotation in one second than I am thinking the formula must be T=60/(p x N). You also mentioned a phase shift, Can I ask how much? How did you reach your conclusion for a phase shift.

So the formula for tau is dependent on T that is fine by me as it was before. When I sent the previous file, which motor did you simulate, the 3 phase one or the 4 phase one?

[SebGue]

Yes, since time is in seconds and electrical period depends on rotational speed it is T = 60 / p / N
The phase shift can be be estimated by the DC current test i ran earlier. I set it to 1/8 of T. So you got to do something like I1 = rectwave(time-1/8)
You also got to consider the order you drive your phases in relation to the rotational direction if you haven't done yet. I.e. my phase order was A D C B.

[Ksp-3086]

So I put the following changes T_P=60/(n*N) and tau=T_P/qs to the code as you said. For the time I put, time1=(linspace(0,T_P,32))-(T_P/8) and also I_1=rect_wav(A,time1-1/8,T_P,tau) at different times. I have also changed the phase sequence but still I am getting unfavorable results. You can take a look at the code if you like.
SRM_innerrotor1.zip

[SebGue]

You don't have to shift change your 'time1'. Only shift the input of rect_wav as above.
But you should take care for your rect_wav function, since it only works for times between 0 and T.

[Ksp-3086]

You mean like I_1=rect_wav(A,time1-1/8,T_P,tau) I got a straight line of torque actually. I even tried I_1=rect_wav(A,time1-T_P/8,T_P,tau) still unfavorable results.

[SebGue]

...sorry, I missed that you substract 1/8 but you got to substract 1/8*T_P.

[Ksp-3086]

Yeah I did that too , right now I am using the code I_1=rect_wav(A,time1-1/8,T_P-1/8,tau) so that it works between 1/8 and T_P supposedly. Still I am getting a flat torque graph due to no current.

[SebGue]

I_1=rect_wav(A,time1-T_P/8,T_P,tau) should work. Whats your current vector/matrix then.

[Ksp-3086]

This is the current matrix and the torque curve is

[SebGue]

With T_P=60/(n*N) n has to be the number of rotor poles, not the stator poles.

[SebGue]

... finally you have should set linspace's endpoint to false

[SebGue]

I set my phase shift by 'hacking' your rect_wav function. but that's quick and dirty and shouldn't be done for reproductive results. So the 1/8th was only a rough estimate.
To get an appropriate phase shift you can do the DC test (no matter what actual machine) for some fraction of your period time. Then you will see, when torque is at its max. and when phase A (B, C, ...) sequence should be started.

For the emf (also known as induced voltage) you may use the calculated value in OutMag.emf. It is basically the time derivative of phase flux linkage (i.e. the magnetic flux linked to the phase coils times the coils number of turns).
For further explainations, I don't have sufficient time to spend :-/

BTW. What's your actual focus as a mechanical engineer?

[Ksp-3086]

So, I made a design for an in wheel motor or a hub motor for automobiles starting from motorbikes first. Although, it is supposed to have better feedback and control over torque and rpm using vehicle control methods, the starting basis is the motor. It was something I though of and my Professor told me to get an efficiency map to get the funding started from investors. I did not have enough data and figures just the design and working principles. You could say the main focus is as an Automotive Engineer as I am doing my masters in that field.

The DC test you are referring to, do you mean I should just test different phase shifts and see proper results which one I get? The graph you showed is completely different than mine which also looks more favorable. So, using the average of out_femm.mag.emf, I can determine if the resulting simulation is favorable or not? I am guessing using V=IR, R is resistance of the wires, the input voltage can be found. If the induced is higher then the simulation must give an error. Did I get that right?

[SebGue]

The DC test is just the thing with the DC current on one phase (preferably phase A) as done before.

As you can model mechanical systems by arrangements of springs, masses, dampers, ..., you can model an electrical system similarly with resistances, inductors, voltage sources, ... Thats an EEC and the number and arrangment of the elements determines the quality of your model.
The simples model you can setup is a resistance (to model winding losses) in series with a voltage source (i.e. the induced voltage in the phase coil, the emf). So your voltage equation will be Va = R * Ia + EMFa with Va ... voltage in phase A, Ia ... current in phase A, EMFa ...

But you may not need that now, if you just want the efficiency map, since you only need mech. power + losses for that.

[Ksp-3086]

I got this from a research paper, although the motor dimensions are different than the one I am using, you can see at high rpm there is no high torque but that is not the case here as I get the same amount of torque at every rpm. That is where this concept comes into play

[SebGue]

I know these figures, since they are normal to electrical machines and I also know you also want to have such figures for your machine.
But you got to understand that for your kind of simulation, no matter what the speed is, the torque will always depend on current. So there is no reduction in torque at higher speeds.
You got to do that on your own at higher speeds. The reason that there is a reduction in torque is due to machines thermal restrictions, limited drive voltage and/or current, ...
So I can't really help you with that.

[Ksp-3086]

Hi @SebGue ,
Is it a bug or something, like one of my files which I have run a many times before, I tried changing the number of FEMM instances to run at the same time to 1 (just to check the outputs). It took a lot of time so I reverted back to 4. Now, whenever I try to run the file it takes a lot of time to run. Previously it took less than 20 minutes but now its taking more than an hour. All the other files are running just fine. I tried closing VS code and opening it too. The problem remains the same. Any fix?

[SebGue]

Don't know whats wrong. You can check the actual number of threads by counting the number of femm instances at a time.

[Ksp-3086]

Wait.. I know what is wrong. The code in which you asked me to do time-T/8 (for phase shift), if I remove that part it runs. The FEMM instances otherwise are not able to reach a solution. It was working fine yesterday morning. Do you think this problem might be temporary? How is it on your end?

[SebGue]

Sorry, I don't know. My simulations run fine. Have you updated pyleecan recently? There were some changes. Maybe there is an issue with the update?

[Ksp-3086]

I have not updated anything but today morning I did update the VS code. It did not seem to work though. It looks more like a problem on my end. One of the FEMM instances does not produce a solution. I noticed when I opened them up in the background.

[Ksp-3086]

I opened up the FEMM directly and ran the simulation by opening the file, using options on it. Then I ran the script again, looks like it is working fine now. The problem was at the FEMM end.

[Ksp-3086]

Hi @SebGue ,
I got the torque as you are seeing right now.

I have sent you the file. Can you take a look at it. Also it is taking abnormally very long to run like 6 hours or something.
SRM_A.zip

[SebGue]

Hello @Ksp-3086

sorry was really busy for some time. Have you made any progress on your issue?
Otherwise I will have a look at your code this week.

Best regards, Sebastian

[Ksp-3086]

My internship is finished and I may not be able to work on the code anymore. I would like to however hand over the code for future references and future help. I also have some suggestions on the GUI but I think it would be better after I am able to do some experimental analysis.

[Ksp-3086]

So here is the code, I hope it could be of some help.
SRM_innerrotor1.zip