Eemods 20.05 11h

0:03

S… Speaker 1 (Eemods 20.05 11h)

Welcome everyone,

0:04

S… Speaker 1 (Eemods 20.05 11h)

I am your founder from Johannes Kessler University Lins in Austria.

0:16

S… Speaker 1 (Eemods 20.05 11h)

when we end this session,

0:19

S… Speaker 1 (Eemods 20.05 11h)

in case our speakers and people that ask

0:23

S… Speaker 1 (Eemods 20.05 11h)

questions,

0:24

S… Speaker 1 (Eemods 20.05 11h)

we just finish

0:28

S… Speaker 1 (Eemods 20.05 11h)

on time every thought,

0:29

S… Speaker 1 (Eemods 20.05 11h)

so I ask you to do this.

0:31

S… Speaker 1 (Eemods 20.05 11h)

And for the questions,

0:33

S… Speaker 1 (Eemods 20.05 11h)

so after the talks,

0:35

S… Speaker 1 (Eemods 20.05 11h)

if you want to ask something,

0:37

S… Speaker 1 (Eemods 20.05 11h)

please raise your hand and when it's your turn,

0:40

S… Speaker 1 (Eemods 20.05 11h)

please introduce yourself with the affiliation and then write

0:45

S… Speaker 1 (Eemods 20.05 11h)

your questions.

0:46

S… Speaker 1 (Eemods 20.05 11h)

So,

0:47

S… Speaker 1 (Eemods 20.05 11h)

the first talk is from Peter Savino

0:51

S… Speaker 1 (Eemods 20.05 11h)

-Tatterson.

0:52

S… Speaker 1 (Eemods 20.05 11h)

from campus drive.

0:54

S… Speaker 1 (Eemods 20.05 11h)

The title here is a little different than here,

0:57

S… Speaker 1 (Eemods 20.05 11h)

but it's called motor -suitual modeling.

1:00

S… Speaker 1 (Eemods 20.05 11h)

So I guess we take the most recent one.

1:03

S… Speaker 1 (Eemods 20.05 11h)

And we have a bio that is a long time

1:07

S… Speaker 1 (Eemods 20.05 11h)

by Inuit.

1:08

S… Speaker 1 (Eemods 20.05 11h)

So our speaker has an electrical engineering

1:12

S… Speaker 1 (Eemods 20.05 11h)

degree from the University of Southland,

1:14

S… Speaker 1 (Eemods 20.05 11h)

Denmark in 2000,

1:16

S… Speaker 1 (Eemods 20.05 11h)

a PhD degree from Glasgow University in 2008.

1:21

S… Speaker 1 (Eemods 20.05 11h)

And he has worked for 25 years at Downport

1:25

S… Speaker 1 (Eemods 20.05 11h)

and dealing with motor design and

1:29

S… Speaker 1 (Eemods 20.05 11h)

optimization as well as motor and transportation.

1:32

S… Speaker 1 (Eemods 20.05 11h)

Thank

1:37

S… Speaker 1 (Eemods 20.05 11h)

you very much.

1:38

S… Speaker 1 (Eemods 20.05 11h)

Without this one,

1:42

S… Speaker 1 (Eemods 20.05 11h)

I guess I have.

1:43

S… Speaker 1 (Eemods 20.05 11h)

I also hope that everybody got the caffeine and

1:47

S… Speaker 1 (Eemods 20.05 11h)

sugar.

1:47

S… Speaker 1 (Eemods 20.05 11h)

But sugar was a little great.

1:48

S… Speaker 1 (Eemods 20.05 11h)

So,

1:49

S… Speaker 1 (Eemods 20.05 11h)

yeah,

1:50

S… Speaker 1 (Eemods 20.05 11h)

as the title suggests,

1:51

S… Speaker 1 (Eemods 20.05 11h)

motor switch laws modeling.

1:53

S… Speaker 1 (Eemods 20.05 11h)

Maybe you could even have said simple screwing motor

1:57

S… Speaker 1 (Eemods 20.05 11h)

switch laws modeling,

1:58

S… Speaker 1 (Eemods 20.05 11h)

but we will get to that.

2:02

S… Speaker 1 (Eemods 20.05 11h)

The phenomenon of switching frequency

2:06

S… Speaker 1 (Eemods 20.05 11h)

related losses in motors has been around for as long as we have had.

2:10

S… Speaker 1 (Eemods 20.05 11h)

And in general,

2:13

S… Speaker 1 (Eemods 20.05 11h)

the tendencies and the trends of these losses are well understood.

2:17

S… Speaker 1 (Eemods 20.05 11h)

They can be based on empirical data,

2:20

S… Speaker 1 (Eemods 20.05 11h)

they are described extensively

2:24

S… Speaker 1 (Eemods 20.05 11h)

in literature,

2:25

S… Speaker 1 (Eemods 20.05 11h)

also by standards,

2:28

S… Speaker 1 (Eemods 20.05 11h)

how to deal with them or

2:32

S… Speaker 1 (Eemods 20.05 11h)

how to...

2:32

S… Speaker 1 (Eemods 20.05 11h)

observe them.

2:33

S… Speaker 1 (Eemods 20.05 11h)

And if you are searching

2:37

S… Speaker 1 (Eemods 20.05 11h)

for really precise magnitudes for a given motor

2:42

S… Speaker 1 (Eemods 20.05 11h)

design, they can also be calculated quite precisely with fine

2:46

S… Speaker 1 (Eemods 20.05 11h)

element tubes.

2:47

S… Speaker 1 (Eemods 20.05 11h)

But as we heard from yesterday,

2:50

S… Speaker 1 (Eemods 20.05 11h)

there is a trend

2:55

S… Speaker 1 (Eemods 20.05 11h)

System designers are considering larger and larger system

2:59

S… Speaker 1 (Eemods 20.05 11h)

models where you have sub -components of systems.

3:02

S… Speaker 1 (Eemods 20.05 11h)

So basically multi -domain modeling could be a

3:06

S… Speaker 1 (Eemods 20.05 11h)

water application or a drilling application or something like that

3:10

S… Speaker 1 (Eemods 20.05 11h)

where you have several domains,

3:12

S… Speaker 1 (Eemods 20.05 11h)

mechanical domains or hydraulic domains,

3:14

S… Speaker 1 (Eemods 20.05 11h)

electrical domains in one complete model in order to assess

3:19

S… Speaker 1 (Eemods 20.05 11h)

the whole system.

3:21

S… Speaker 1 (Eemods 20.05 11h)

And in these,

3:23

S… Speaker 1 (Eemods 20.05 11h)

we would call these either digital twins or

3:28

S… Speaker 1 (Eemods 20.05 11h)

systems approach to modeling.

3:30

S… Speaker 1 (Eemods 20.05 11h)

And in these cases,

3:32

S… Speaker 1 (Eemods 20.05 11h)

the motor model itself is typically a,

3:35

S… Speaker 1 (Eemods 20.05 11h)

it's definitely not a fine element model,

3:39

S… Speaker 1 (Eemods 20.05 11h)

because nobody has the patience to wait for that in these

3:43

S… Speaker 1 (Eemods 20.05 11h)

cases.

3:43

S… Speaker 1 (Eemods 20.05 11h)

So it's typically something that's based on,

3:47

S… Speaker 1 (Eemods 20.05 11h)

say, well -known equipment.

3:49

S… Speaker 1 (Eemods 20.05 11h)

diagrams.

3:50

S… Speaker 1 (Eemods 20.05 11h)

And the question then is,

3:52

S… Speaker 1 (Eemods 20.05 11h)

if we take that as a basis,

3:56

S… Speaker 1 (Eemods 20.05 11h)

then how well are these switching losses actually modeled in these cases?

4:00

S… Speaker 1 (Eemods 20.05 11h)

And before we get to that,

4:03

S… Speaker 1 (Eemods 20.05 11h)

let's just get some fundamental things straight here.

4:06

S… Speaker 1 (Eemods 20.05 11h)

If we take a,

4:09

S… Speaker 1 (Eemods 20.05 11h)

let's say that we make a 3 watt measurement

4:13

S… Speaker 1 (Eemods 20.05 11h)

on a given machine,

4:15

S… Speaker 1 (Eemods 20.05 11h)

in this case a 55 kilowatt induction machine,

4:19

S… Speaker 1 (Eemods 20.05 11h)

running at some speeds,

4:20

S… Speaker 1 (Eemods 20.05 11h)

half the nominal speed in this case,

4:22

S… Speaker 1 (Eemods 20.05 11h)

hard low,

4:23

S… Speaker 1 (Eemods 20.05 11h)

switching frequency of 2 kHz,

4:26

S… Speaker 1 (Eemods 20.05 11h)

then we can calculate the instantaneous power from phase

4:30

S… Speaker 1 (Eemods 20.05 11h)

voltage times phase current and sum it

4:34

S… Speaker 1 (Eemods 20.05 11h)

all over the three phases and then we can plot it in the time domain here like in the upper

4:39

S… Speaker 1 (Eemods 20.05 11h)

part and it will look like this.

4:41

S… Speaker 1 (Eemods 20.05 11h)

So, an oscillating signal but with a

4:45

S… Speaker 1 (Eemods 20.05 11h)

mean value somewhere,

4:47

S… Speaker 1 (Eemods 20.05 11h)

in this case here by the red line,

4:50

S… Speaker 1 (Eemods 20.05 11h)

which is the average power and thereby the real power being

4:54

S… Speaker 1 (Eemods 20.05 11h)

delivered to the motor.

4:56

S… Speaker 1 (Eemods 20.05 11h)

If we take this signal in the frequency...

5:00

S… Speaker 1 (Eemods 20.05 11h)

domain it will look like this cloud of

5:04

S… Speaker 1 (Eemods 20.05 11h)

peaks along the frequency range but the only

5:08

S… Speaker 1 (Eemods 20.05 11h)

power being delivered is actually the DC

5:12

S… Speaker 1 (Eemods 20.05 11h)

component of this everything else is just oscillations and

5:18

S… Speaker 1 (Eemods 20.05 11h)

that's because you have multiple frequencies interacting with each other but as we know

5:22

S… Speaker 1 (Eemods 20.05 11h)

power is only generated by voltage and frequency

5:26

S… Speaker 1 (Eemods 20.05 11h)

of equal

5:27

S… Speaker 1 (Eemods 20.05 11h)

sorry,

5:28

S… Speaker 1 (Eemods 20.05 11h)

voltage and current of equal frequency being

5:32

S… Speaker 1 (Eemods 20.05 11h)

determined by their respective sizes and

5:38

S… Speaker 1 (Eemods 20.05 11h)

then the angular distribution between them.

5:41

S… Speaker 1 (Eemods 20.05 11h)

If we

5:45

S… Speaker 1 (Eemods 20.05 11h)

kind of expand...

5:49

S… Speaker 1 (Eemods 20.05 11h)

these spectrums here into real power delivered at

5:53

S… Speaker 1 (Eemods 20.05 11h)

respective frequencies then it will look like this we will have the

5:57

S… Speaker 1 (Eemods 20.05 11h)

power at fundamental frequency

6:02

S… Speaker 1 (Eemods 20.05 11h)

here

6:03

S… Speaker 1 (Eemods 20.05 11h)

And then we will have some components at higher frequencies

6:07

S… Speaker 1 (Eemods 20.05 11h)

also.

6:07

S… Speaker 1 (Eemods 20.05 11h)

We will also have some here in the

6:12

S… Speaker 1 (Eemods 20.05 11h)

lower harmonic screen.

6:15

S… Speaker 1 (Eemods 20.05 11h)

So not associated with switching but with machine imperfections.

6:21

S… Speaker 1 (Eemods 20.05 11h)

Either on balance or slot harmonics

6:25

S… Speaker 1 (Eemods 20.05 11h)

or something like that,

6:27

S… Speaker 1 (Eemods 20.05 11h)

winding harmonics.

6:28

S… Speaker 1 (Eemods 20.05 11h)

Something like that.

6:30

S… Speaker 1 (Eemods 20.05 11h)

This perhaps is not so illustrative as to where is the power,

6:35

S… Speaker 1 (Eemods 20.05 11h)

where are the main components of this spectral power

6:39

S… Speaker 1 (Eemods 20.05 11h)

being delivered.

6:39

S… Speaker 1 (Eemods 20.05 11h)

So another measure would

6:44

S… Speaker 1 (Eemods 20.05 11h)

be this one here where we basically sum the real power over

6:48

S… Speaker 1 (Eemods 20.05 11h)

the frequency and that will then give us this plot here

6:52

S… Speaker 1 (Eemods 20.05 11h)

and I think that gives a clearer picture of where in the frequency range

6:56

S… Speaker 1 (Eemods 20.05 11h)

is the power actually being delivered.

7:00

S… Speaker 1 (Eemods 20.05 11h)

Please note that the scale here does not start at zero because

7:06

S… Speaker 1 (Eemods 20.05 11h)

the majority of power is still being delivered at fundamental frequency

7:10

S… Speaker 1 (Eemods 20.05 11h)

which in this case is high hertz because the motor is running in some hard

7:14

S… Speaker 1 (Eemods 20.05 11h)

load condition and then we can simply by summing this spectrum

7:19

S… Speaker 1 (Eemods 20.05 11h)

here we will see how the power increases

7:23

S… Speaker 1 (Eemods 20.05 11h)

until it converges to the red line here which is also this line

7:28

S… Speaker 1 (Eemods 20.05 11h)

here

7:30

S… Speaker 1 (Eemods 20.05 11h)

Yeah, that kind of shows us where is it that the

7:35

S… Speaker 1 (Eemods 20.05 11h)

losses occur and how big are they.

7:37

S… Speaker 1 (Eemods 20.05 11h)

And here we can see in this case the majority of the switching related losses is

7:41

S… Speaker 1 (Eemods 20.05 11h)

taking place at two times the switching frequency of the dimmerer.

7:45

S… Speaker 1 (Eemods 20.05 11h)

So,

7:48

S… Speaker 1 (Eemods 20.05 11h)

if we should try to model this in a,

7:52

S… Speaker 1 (Eemods 20.05 11h)

you could say,

7:52

S… Speaker 1 (Eemods 20.05 11h)

commonly available dynamic motor simulation to be

7:57

S… Speaker 1 (Eemods 20.05 11h)

put in flex,

7:58

S… Speaker 1 (Eemods 20.05 11h)

which is,

7:59

S… Speaker 1 (Eemods 20.05 11h)

those of you that are familiar with the matter,

8:02

S… Speaker 1 (Eemods 20.05 11h)

You will know that this is a plug -in to MATLAB objects.

8:05

S… Speaker 1 (Eemods 20.05 11h)

It can also run in its own environment.

8:08

S… Speaker 1 (Eemods 20.05 11h)

There are other plug -ins to MATLAB,

8:12

S… Speaker 1 (Eemods 20.05 11h)

for instance Simscape,

8:13

S… Speaker 1 (Eemods 20.05 11h)

that will also present you with a ready -to -go motor model

8:17

S… Speaker 1 (Eemods 20.05 11h)

that you just need to configure,

8:19

S… Speaker 1 (Eemods 20.05 11h)

and then you will basically see how it performs.

8:21

S… Speaker 1 (Eemods 20.05 11h)

It will basically give you this

8:25

S… Speaker 1 (Eemods 20.05 11h)

level of detail here.

8:26

S… Speaker 1 (Eemods 20.05 11h)

You can see that it's based on a conventional interface.

8:31

S… Speaker 1 (Eemods 20.05 11h)

equivalent circuit with these components and

8:36

S… Speaker 1 (Eemods 20.05 11h)

even if it's not so obvious then with the

8:40

S… Speaker 1 (Eemods 20.05 11h)

simscape it's

8:45

S… Speaker 1 (Eemods 20.05 11h)

the same level of detail for both simscape and flex

8:50

S… Speaker 1 (Eemods 20.05 11h)

so basically I would expect them to be completely identical

8:55

S… Speaker 1 (Eemods 20.05 11h)

If we then mimic the operation point that I had

8:59

S… Speaker 1 (Eemods 20.05 11h)

from the measurement,

9:00

S… Speaker 1 (Eemods 20.05 11h)

so basically configuring the flex model with the same

9:04

S… Speaker 1 (Eemods 20.05 11h)

inductances,

9:05

S… Speaker 1 (Eemods 20.05 11h)

reactances,

9:06

S… Speaker 1 (Eemods 20.05 11h)

resistances,

9:07

S… Speaker 1 (Eemods 20.05 11h)

all that, and then run it with the same switching frequency and

9:12

S… Speaker 1 (Eemods 20.05 11h)

take note conditions,

9:13

S… Speaker 1 (Eemods 20.05 11h)

then we can repeat the calculations

9:18

S… Speaker 1 (Eemods 20.05 11h)

and then see how the power is delivered.

9:21

S… Speaker 1 (Eemods 20.05 11h)

And in this case you can see we have roughly the same fundamental power here,

9:26

S… Speaker 1 (Eemods 20.05 11h)

around 7 point something kilowatts.

9:28

S… Speaker 1 (Eemods 20.05 11h)

And we also have some representation of the power

9:33

S… Speaker 1 (Eemods 20.05 11h)

being delivered at a higher frequency,

9:35

S… Speaker 1 (Eemods 20.05 11h)

but they are very very small.

9:38

S… Speaker 1 (Eemods 20.05 11h)

Basically they are almost visible.

9:40

S… Speaker 1 (Eemods 20.05 11h)

So using a model like this will not correctly

9:45

S… Speaker 1 (Eemods 20.05 11h)

present switching induced losses

9:49

S… Speaker 1 (Eemods 20.05 11h)

in the motor.

9:51

S… Speaker 1 (Eemods 20.05 11h)

And if we think a little bit about it,

9:53

S… Speaker 1 (Eemods 20.05 11h)

then it's not so strange actually,

9:55

S… Speaker 1 (Eemods 20.05 11h)

because, again,

9:57

S… Speaker 1 (Eemods 20.05 11h)

this is just another representation of the same...

10:00

S… Speaker 1 (Eemods 20.05 11h)

circuitry that was used in flex.

10:01

S… Speaker 1 (Eemods 20.05 11h)

If we analyze this circuit and calculate

10:06

S… Speaker 1 (Eemods 20.05 11h)

the input impedance,

10:07

S… Speaker 1 (Eemods 20.05 11h)

it will look something like this.

10:10

S… Speaker 1 (Eemods 20.05 11h)

And if we make it good about this,

10:12

S… Speaker 1 (Eemods 20.05 11h)

it will look something like this,

10:15

S… Speaker 1 (Eemods 20.05 11h)

depending on the gene parameters and everything,

10:17

S… Speaker 1 (Eemods 20.05 11h)

of course.

10:18

S… Speaker 1 (Eemods 20.05 11h)

But what is obvious,

10:20

S… Speaker 1 (Eemods 20.05 11h)

or what is perhaps interesting or relevant for this discussion here,

10:23

S… Speaker 1 (Eemods 20.05 11h)

is how does the impedance behave in

10:27

S… Speaker 1 (Eemods 20.05 11h)

the, you could say,

10:28

S… Speaker 1 (Eemods 20.05 11h)

kilohertz range,

10:29

S… Speaker 1 (Eemods 20.05 11h)

which we have up here.

10:30

S… Speaker 1 (Eemods 20.05 11h)

And you can see,

10:32

S… Speaker 1 (Eemods 20.05 11h)

first of all,

10:33

S… Speaker 1 (Eemods 20.05 11h)

it has some magnitude,

10:34

S… Speaker 1 (Eemods 20.05 11h)

but the face angle is almost zero.

10:37

S… Speaker 1 (Eemods 20.05 11h)

So it's almost completely an electric circle.

10:42

S… Speaker 1 (Eemods 20.05 11h)

And then it's clear that we can't produce any power of that.

10:45

S… Speaker 1 (Eemods 20.05 11h)

But of course those of you that are familiar with Motus

10:49

S… Speaker 1 (Eemods 20.05 11h)

would say that there is obviously something lacking in this module,

10:52

S… Speaker 1 (Eemods 20.05 11h)

right? At least there is something lacking here.

10:54

S… Speaker 1 (Eemods 20.05 11h)

So why didn't they present us

10:58

S… Speaker 1 (Eemods 20.05 11h)

with a parameter

11:03

S… Speaker 1 (Eemods 20.05 11h)

like this?

11:04

S… Speaker 1 (Eemods 20.05 11h)

I don't know.

11:05

S… Speaker 1 (Eemods 20.05 11h)

They didn't,

11:06

S… Speaker 1 (Eemods 20.05 11h)

but we can of course include it in our analysis here and see

11:11

S… Speaker 1 (Eemods 20.05 11h)

what happens.

11:12

S… Speaker 1 (Eemods 20.05 11h)

How did we do that?

11:13

S… Speaker 1 (Eemods 20.05 11h)

Then what happens?

11:14

S… Speaker 1 (Eemods 20.05 11h)

And if we introduce this well

11:18

S… Speaker 1 (Eemods 20.05 11h)

-known call loss resistance,

11:20

S… Speaker 1 (Eemods 20.05 11h)

we will add another 0 -1 pole to the transfer function and

11:25

S… Speaker 1 (Eemods 20.05 11h)

that will then give us the red characteristic here.

11:27

S… Speaker 1 (Eemods 20.05 11h)

The blue is the simpler one from the previous slide.

11:30

S… Speaker 1 (Eemods 20.05 11h)

So it will affect the...

11:36

S… Speaker 1 (Eemods 20.05 11h)

impedance in the gigahertz range but will

11:40

S… Speaker 1 (Eemods 20.05 11h)

it then be a good representation of core losses you can see so

11:45

S… Speaker 1 (Eemods 20.05 11h)

if we include this because

11:49

S… Speaker 1 (Eemods 20.05 11h)

we can do that in reflex we can simply add it and then modify

11:54

S… Speaker 1 (Eemods 20.05 11h)

some of the equations here then

11:58

S… Speaker 1 (Eemods 20.05 11h)

we will get a clearer picture of

12:02

S… Speaker 1 (Eemods 20.05 11h)

or at least we can see something

12:05

S… Speaker 1 (Eemods 20.05 11h)

in the Kilohertz range here,

12:07

S… Speaker 1 (Eemods 20.05 11h)

but it is nowhere near enough when we compare it against the measurements.

12:11

S… Speaker 1 (Eemods 20.05 11h)

It's only,

12:12

S… Speaker 1 (Eemods 20.05 11h)

in this case,

12:13

S… Speaker 1 (Eemods 20.05 11h)

50 watts,

12:14

S… Speaker 1 (Eemods 20.05 11h)

which we have 500 watts measured.

12:17

S… Speaker 1 (Eemods 20.05 11h)

So clearly,

12:18

S… Speaker 1 (Eemods 20.05 11h)

this is not enough.

12:19

S… Speaker 1 (Eemods 20.05 11h)

And then we could,

12:20

S… Speaker 1 (Eemods 20.05 11h)

of course, start to modify this one,

12:23

S… Speaker 1 (Eemods 20.05 11h)

but we also don't want to do this,

12:25

S… Speaker 1 (Eemods 20.05 11h)

because this one is the one that represents the fundamental core losses,

12:28

S… Speaker 1 (Eemods 20.05 11h)

and we should not achieve it in that way.

12:31

S… Speaker 1 (Eemods 20.05 11h)

So something is missing.

12:36

S… Speaker 1 (Eemods 20.05 11h)

What we could do is to perhaps extract the

12:40

S… Speaker 1 (Eemods 20.05 11h)

spectrum impedances from the measurements we just did,

12:43

S… Speaker 1 (Eemods 20.05 11h)

because we have made three phase voltages,

12:46

S… Speaker 1 (Eemods 20.05 11h)

three phase current measurements,

12:49

S… Speaker 1 (Eemods 20.05 11h)

so we can analyze

12:53

S… Speaker 1 (Eemods 20.05 11h)

them in the frequency domain,

12:55

S… Speaker 1 (Eemods 20.05 11h)

get magnitudes and phases,

12:56

S… Speaker 1 (Eemods 20.05 11h)

and then actually extract the impedances for each frequency

13:00

S… Speaker 1 (Eemods 20.05 11h)

when we have this.

13:02

S… Speaker 1 (Eemods 20.05 11h)

and that will give us something that you could say comparable but not

13:06

S… Speaker 1 (Eemods 20.05 11h)

identical to the Bode plot before and what I have done here is

13:10

S… Speaker 1 (Eemods 20.05 11h)

I wouldn't say I have cheated but I have figured out all

13:15

S… Speaker 1 (Eemods 20.05 11h)

the impedances that are calculated you could say

13:19

S… Speaker 1 (Eemods 20.05 11h)

for currents that are less than 1 % of the fundamental because

13:23

S… Speaker 1 (Eemods 20.05 11h)

otherwise this calculation would present

13:28

S… Speaker 1 (Eemods 20.05 11h)

me with an impedance for every

13:30

S… Speaker 1 (Eemods 20.05 11h)

frequency, including that which is also just noise,

13:33

S… Speaker 1 (Eemods 20.05 11h)

so I would just have a picture full of crosses all over.

13:36

S… Speaker 1 (Eemods 20.05 11h)

So let's just observe the impedances where we have a

13:40

S… Speaker 1 (Eemods 20.05 11h)

current of a certain magnitude.

13:42

S… Speaker 1 (Eemods 20.05 11h)

And that gives us something like this.

13:45

S… Speaker 1 (Eemods 20.05 11h)

And that tells us

13:49

S… Speaker 1 (Eemods 20.05 11h)

something.

13:51

S… Speaker 1 (Eemods 20.05 11h)

It tells us that it's kind of...

13:55

S… Speaker 1 (Eemods 20.05 11h)

And you can discuss how obvious it is,

13:57

S… Speaker 1 (Eemods 20.05 11h)

but maybe you could argue that this curve seems to flatten

14:01

S… Speaker 1 (Eemods 20.05 11h)

out here and have higher frequencies.

14:03

S… Speaker 1 (Eemods 20.05 11h)

Also you could argue that there is a downward tendency

14:08

S… Speaker 1 (Eemods 20.05 11h)

in the phase angle once you increase the frequency.

14:12

S… Speaker 1 (Eemods 20.05 11h)

So how

14:16

S… Speaker 1 (Eemods 20.05 11h)

could we limit that?

14:17

S… Speaker 1 (Eemods 20.05 11h)

And a simple way to do that could simply be to add a

14:22

S… Speaker 1 (Eemods 20.05 11h)

kind of a parasitic resistor here.

14:25

S… Speaker 1 (Eemods 20.05 11h)

And if we then repeat the analysis,

14:27

S… Speaker 1 (Eemods 20.05 11h)

that means that we will add another pole in the third

14:32

S… Speaker 1 (Eemods 20.05 11h)

function.

14:32

S… Speaker 1 (Eemods 20.05 11h)

And that means that we will eventually,

14:34

S… Speaker 1 (Eemods 20.05 11h)

say,

14:34

S… Speaker 1 (Eemods 20.05 11h)

create a break -off frequency

14:40

S… Speaker 1 (Eemods 20.05 11h)

here somewhere.

14:41

S… Speaker 1 (Eemods 20.05 11h)

That means that eventually,

14:43

S… Speaker 1 (Eemods 20.05 11h)

at a high frequency,

14:45

S… Speaker 1 (Eemods 20.05 11h)

the motor will become purely resistant.

14:47

S… Speaker 1 (Eemods 20.05 11h)

And that means we will definitely have some losses here.

14:52

S… Speaker 1 (Eemods 20.05 11h)

So of course,

14:52

S… Speaker 1 (Eemods 20.05 11h)

depending on the value of this,

14:54

S… Speaker 1 (Eemods 20.05 11h)

we can shape the impedance transfer function so

14:59

S… Speaker 1 (Eemods 20.05 11h)

that it could...

15:00

S… Speaker 1 (Eemods 20.05 11h)

fit somehow the measurements.

15:02

S… Speaker 1 (Eemods 20.05 11h)

If we do that,

15:05

S… Speaker 1 (Eemods 20.05 11h)

and if we also include this parasitic resistor

15:09

S… Speaker 1 (Eemods 20.05 11h)

in the plank structure,

15:10

S… Speaker 1 (Eemods 20.05 11h)

and spend a little time to find

15:14

S… Speaker 1 (Eemods 20.05 11h)

the right value for it,

15:15

S… Speaker 1 (Eemods 20.05 11h)

then we can actually get

15:20

S… Speaker 1 (Eemods 20.05 11h)

to a magnitude that seems to fit,

15:22

S… Speaker 1 (Eemods 20.05 11h)

and which also seems to still use,

15:25

S… Speaker 1 (Eemods 20.05 11h)

say,

15:27

S… Speaker 1 (Eemods 20.05 11h)

maintain the fundamental properties of the machine.

15:30

S… Speaker 1 (Eemods 20.05 11h)

So basically we are not really affecting the performance

15:37

S… Speaker 1 (Eemods 20.05 11h)

of the circuit at nominal frequencies.

15:40

S… Speaker 1 (Eemods 20.05 11h)

So just get the value right

15:44

S… Speaker 1 (Eemods 20.05 11h)

and it works.

15:46

S… Speaker 1 (Eemods 20.05 11h)

It's just the thing about getting the value right.

15:51

S… Speaker 1 (Eemods 20.05 11h)

Okay,

15:51

S… Speaker 1 (Eemods 20.05 11h)

just to illustrate here,

15:54

S… Speaker 1 (Eemods 20.05 11h)

just to maybe sum up a little bit,

15:58

S… Speaker 1 (Eemods 20.05 11h)

if we compare,

16:00

S… Speaker 1 (Eemods 20.05 11h)

I would call it model sophistication,

16:03

S… Speaker 1 (Eemods 20.05 11h)

that's probably a bit rich,

16:05

S… Speaker 1 (Eemods 20.05 11h)

since I'm basically just adding some resistance,

16:08

S… Speaker 1 (Eemods 20.05 11h)

but in the default case

16:12

S… Speaker 1 (Eemods 20.05 11h)

that Plex will present you with,

16:14

S… Speaker 1 (Eemods 20.05 11h)

that would be the blue cross here.

16:19

S… Speaker 1 (Eemods 20.05 11h)

There are no iron loss considerations whatsoever.

16:22

S… Speaker 1 (Eemods 20.05 11h)

That would be kind of the default.

16:24

S… Speaker 1 (Eemods 20.05 11h)

If we add the fundamental frequency for all of these systems,

16:28

S… Speaker 1 (Eemods 20.05 11h)

we will not really affect the magnitude,

16:31

S… Speaker 1 (Eemods 20.05 11h)

but a little bit the phase of the impedance.

16:34

S… Speaker 1 (Eemods 20.05 11h)

But if we add this additional parasitic

16:38

S… Speaker 1 (Eemods 20.05 11h)

resistor...

16:39

S… Speaker 1 (Eemods 20.05 11h)

then we will lower the magnitude of the

16:44

S… Speaker 1 (Eemods 20.05 11h)

impedance but we will also lower the phase angle and then overlay them with

16:48

S… Speaker 1 (Eemods 20.05 11h)

the measurements we can actually get relatively close

16:52

S… Speaker 1 (Eemods 20.05 11h)

so you could say we can use measurements

16:56

S… Speaker 1 (Eemods 20.05 11h)

like this to find an indication of what should this parasitic resistor be

17:00

S… Speaker 1 (Eemods 20.05 11h)

now of course

17:06

S… Speaker 1 (Eemods 20.05 11h)

I think I just wanted to show that from

17:10

S… Speaker 1 (Eemods 20.05 11h)

the measurements,

17:11

S… Speaker 1 (Eemods 20.05 11h)

depending on what the switching frequency is,

17:13

S… Speaker 1 (Eemods 20.05 11h)

we still get like a trend line in

17:18

S… Speaker 1 (Eemods 20.05 11h)

the extracted

17:22

S… Speaker 1 (Eemods 20.05 11h)

impedances here.

17:23

S… Speaker 1 (Eemods 20.05 11h)

I tried the three different switching frequencies,

17:26

S… Speaker 1 (Eemods 20.05 11h)

but they seem to respect the same lines

17:32

S… Speaker 1 (Eemods 20.05 11h)

or patterns.

17:35

S… Speaker 1 (Eemods 20.05 11h)

And again,

17:36

S… Speaker 1 (Eemods 20.05 11h)

only based on measurements,

17:37

S… Speaker 1 (Eemods 20.05 11h)

and this is actually well known,

17:39

S… Speaker 1 (Eemods 20.05 11h)

if we have 2 kHz as a default,

17:43

S… Speaker 1 (Eemods 20.05 11h)

if we increase the switching frequency to 5 kHz,

17:47

S… Speaker 1 (Eemods 20.05 11h)

the losses will reduce.

17:48

S… Speaker 1 (Eemods 20.05 11h)

Then I try to increase

17:53

S… Speaker 1 (Eemods 20.05 11h)

it to 10 kHz,

17:53

S… Speaker 1 (Eemods 20.05 11h)

but what happens at the same time with the drive configuration I used is that...

17:58

S… Speaker 1 (Eemods 20.05 11h)

I will increase the switching frequency but the pattern will also change to something that

18:02

S… Speaker 1 (Eemods 20.05 11h)

contains more harmonics so in this case the

18:07

S… Speaker 1 (Eemods 20.05 11h)

losses didn't really decrease further but

18:11

S… Speaker 1 (Eemods 20.05 11h)

that's because it's another so it's basically not a fair comparison maybe

18:16

S… Speaker 1 (Eemods 20.05 11h)

I should actually have left this out but it's just to say that sometimes

18:21

S… Speaker 1 (Eemods 20.05 11h)

the pattern and the switching frequency they can pull in different

18:25

S… Speaker 1 (Eemods 20.05 11h)

directions

18:27

S… Speaker 1 (Eemods 20.05 11h)

Now, if we try to simulate the

18:32

S… Speaker 1 (Eemods 20.05 11h)

same thing basically,

18:33

S… Speaker 1 (Eemods 20.05 11h)

how well does the model then respond to

18:37

S… Speaker 1 (Eemods 20.05 11h)

these observations that we know?

18:40

S… Speaker 1 (Eemods 20.05 11h)

So

18:45

S… Speaker 1 (Eemods 20.05 11h)

again, I simulated around 490

18:50

S… Speaker 1 (Eemods 20.05 11h)

watts with 2 kilohertz.

18:53

S… Speaker 1 (Eemods 20.05 11h)

If I increase the switching frequency to 5 kHz,

18:56

S… Speaker 1 (Eemods 20.05 11h)

the losses reduce,

18:58

S… Speaker 1 (Eemods 20.05 11h)

and if I increase it further without changing

19:02

S… Speaker 1 (Eemods 20.05 11h)

the pattern, the losses will reduce a bit further,

19:06

S… Speaker 1 (Eemods 20.05 11h)

but not by the same...

19:10

S… Speaker 1 (Eemods 20.05 11h)

proportion.

19:10

S… Speaker 1 (Eemods 20.05 11h)

And that is actually also in line with the common knowledge

19:14

S… Speaker 1 (Eemods 20.05 11h)

that you can increase the switching frequency but when you increase

19:19

S… Speaker 1 (Eemods 20.05 11h)

it beyond a certain point you will not really get any rewards for it.

19:22

S… Speaker 1 (Eemods 20.05 11h)

So that's actually answered very well.

19:26

S… Speaker 1 (Eemods 20.05 11h)

And it's

19:30

S… Speaker 1 (Eemods 20.05 11h)

not a perfect correspondence with

19:34

S… Speaker 1 (Eemods 20.05 11h)

measurements but at least the trend is right.

19:39

S… Speaker 1 (Eemods 20.05 11h)

I also tried to,

19:41

S… Speaker 1 (Eemods 20.05 11h)

using the 2 kHz switching frequency as a reference,

19:44

S… Speaker 1 (Eemods 20.05 11h)

try to see what would happen if I changed the magnetic operating

19:48

S… Speaker 1 (Eemods 20.05 11h)

point of the machine by simply reducing or increasing the voltage,

19:52

S… Speaker 1 (Eemods 20.05 11h)

the fundamental voltage,

19:54

S… Speaker 1 (Eemods 20.05 11h)

so that would be the 25 Hz component.

20:00

S… Speaker 1 (Eemods 20.05 11h)

The benchmark here is at 490 versus 500 watts

20:05

S… Speaker 1 (Eemods 20.05 11h)

at 170 volts in this particular operating

20:09

S… Speaker 1 (Eemods 20.05 11h)

point.

20:09

S… Speaker 1 (Eemods 20.05 11h)

If I reduce it,

20:11

S… Speaker 1 (Eemods 20.05 11h)

I reduce the losses.

20:13

S… Speaker 1 (Eemods 20.05 11h)

That's also the case over here.

20:15

S… Speaker 1 (Eemods 20.05 11h)

If I increase it,

20:16

S… Speaker 1 (Eemods 20.05 11h)

I increase the losses.

20:18

S… Speaker 1 (Eemods 20.05 11h)

So the friction stimulation losses also do it over here.

20:22

S… Speaker 1 (Eemods 20.05 11h)

It's not a perfect match,

20:23

S… Speaker 1 (Eemods 20.05 11h)

but again, the trend is correct.

20:26

S… Speaker 1 (Eemods 20.05 11h)

And to be honest,

20:27

S… Speaker 1 (Eemods 20.05 11h)

I was actually a little bit surprised because the modification is really simple.

20:31

S… Speaker 1 (Eemods 20.05 11h)

It's just one resistor.

20:32

S… Speaker 1 (Eemods 20.05 11h)

I also tried to do the same

20:36

S… Speaker 1 (Eemods 20.05 11h)

for a permanent magnet

20:40

S… Speaker 1 (Eemods 20.05 11h)

machine that Plex will also allow us to...

20:50

S… Speaker 1 (Eemods 20.05 11h)

They present different models,

20:52

S… Speaker 1 (Eemods 20.05 11h)

configurations,

20:54

S… Speaker 1 (Eemods 20.05 11h)

some of them you can configure with flux maps,

20:57

S… Speaker 1 (Eemods 20.05 11h)

some of them you can also just configure with circuit parameters.

21:02

S… Speaker 1 (Eemods 20.05 11h)

But again,

21:02

S… Speaker 1 (Eemods 20.05 11h)

we can make the same modifications.

21:06

S… Speaker 1 (Eemods 20.05 11h)

So this resistor is not here by default.

21:09

S… Speaker 1 (Eemods 20.05 11h)

So I simply just add it in the flex circuit.

21:12

S… Speaker 1 (Eemods 20.05 11h)

and the difference is actually the same when

21:17

S… Speaker 1 (Eemods 20.05 11h)

I compare against measurements which is the blue line here and

21:23

S… Speaker 1 (Eemods 20.05 11h)

again I find the right value for this resistor then I can actually

21:27

S… Speaker 1 (Eemods 20.05 11h)

get it to behave quite similar

21:32

S… Speaker 1 (Eemods 20.05 11h)

in terms of where the

21:36

S… Speaker 1 (Eemods 20.05 11h)

power is delivered at which frequencies whereas if

21:40

S… Speaker 1 (Eemods 20.05 11h)

I don't do anything I will not

21:42

S… Speaker 1 (Eemods 20.05 11h)

I will not calculate any mouse losses at all.

21:44

S… Speaker 1 (Eemods 20.05 11h)

So the yellow one here,

21:47

S… Speaker 1 (Eemods 20.05 11h)

that's the default PLEX model,

21:49

S… Speaker 1 (Eemods 20.05 11h)

and that will not generate any switching related losses.

21:52

S… Speaker 1 (Eemods 20.05 11h)

So if that's important or

21:56

S… Speaker 1 (Eemods 20.05 11h)

interesting, then we need to do something about it.

22:00

S… Speaker 1 (Eemods 20.05 11h)

At least if we are relying on,

22:02

S… Speaker 1 (Eemods 20.05 11h)

say,

22:03

S… Speaker 1 (Eemods 20.05 11h)

model complexity like PLEX or SimScale,

22:05

S… Speaker 1 (Eemods 20.05 11h)

because they are not there as default.

22:10

S… Speaker 1 (Eemods 20.05 11h)

Yeah, so just to maybe summarize a little bit,

22:13

S… Speaker 1 (Eemods 20.05 11h)

it's a really simple modification,

22:16

S… Speaker 1 (Eemods 20.05 11h)

but it seems to capture at least most of the trends correctly,

22:21

S… Speaker 1 (Eemods 20.05 11h)

provided that we can find a suitable value for this resistor.

22:25

S… Speaker 1 (Eemods 20.05 11h)

It may not only

22:29

S… Speaker 1 (Eemods 20.05 11h)

be in terms of losses that it could be interesting,

22:31

S… Speaker 1 (Eemods 20.05 11h)

it could also be if there are THDR considerations,

22:34

S… Speaker 1 (Eemods 20.05 11h)

maybe for cable loss calculations or something like that.

22:39

S… Speaker 1 (Eemods 20.05 11h)

between drive and motor and yeah

22:43

S… Speaker 1 (Eemods 20.05 11h)

exactly what I say here if we have system models where

22:47

S… Speaker 1 (Eemods 20.05 11h)

we also consider losses in cables for

22:51

S… Speaker 1 (Eemods 20.05 11h)

instance yeah and that's also very nice but

22:56

S… Speaker 1 (Eemods 20.05 11h)

how should we actually attempt this

23:02

S… Speaker 1 (Eemods 20.05 11h)

dimensioning of the resistor values.

23:03

S… Speaker 1 (Eemods 20.05 11h)

It is based on many factors,

23:06

S… Speaker 1 (Eemods 20.05 11h)

machine non -linearity,

23:08

S… Speaker 1 (Eemods 20.05 11h)

machine design,

23:10

S… Speaker 1 (Eemods 20.05 11h)

all kinds,

23:11

S… Speaker 1 (Eemods 20.05 11h)

machine size of course,

23:12

S… Speaker 1 (Eemods 20.05 11h)

type.

23:12

S… Speaker 1 (Eemods 20.05 11h)

We can find them from measurements if we do

23:16

S… Speaker 1 (Eemods 20.05 11h)

these spectral analysis,

23:18

S… Speaker 1 (Eemods 20.05 11h)

but in many cases that's not really an option.

23:21

S… Speaker 1 (Eemods 20.05 11h)

Maybe you don't even have the machine present,

23:25

S… Speaker 1 (Eemods 20.05 11h)

maybe you only have a data sheet values of it and you try to see how it fits into

23:29

S… Speaker 1 (Eemods 20.05 11h)

your system.

23:32

S… Speaker 1 (Eemods 20.05 11h)

but these unknowns can actually be said about

23:36

S… Speaker 1 (Eemods 20.05 11h)

many things of the machine also like non -linearities

23:40

S… Speaker 1 (Eemods 20.05 11h)

or even the core loss resistor itself it's not always

23:44

S… Speaker 1 (Eemods 20.05 11h)

well defined what the actual value is so

23:51

S… Speaker 1 (Eemods 20.05 11h)

one approach to find out

23:55

S… Speaker 1 (Eemods 20.05 11h)

What this value should be would be to simply measure a lot of motors

23:59

S… Speaker 1 (Eemods 20.05 11h)

and then see is there a trend line in them that we could

24:03

S… Speaker 1 (Eemods 20.05 11h)

rely on.

24:04

S… Speaker 1 (Eemods 20.05 11h)

We could

24:08

S… Speaker 1 (Eemods 20.05 11h)

also do something else and basically this is what

24:12

S… Speaker 1 (Eemods 20.05 11h)

I tried to at least start here with these

24:16

S… Speaker 1 (Eemods 20.05 11h)

measurements on four different types,

24:18

S… Speaker 1 (Eemods 20.05 11h)

two induction motors.

24:20

S… Speaker 1 (Eemods 20.05 11h)

permanent magnets and then a synchronous reluctance type of different

24:24

S… Speaker 1 (Eemods 20.05 11h)

powers and these values are simply found by trial and error or

24:28

S… Speaker 1 (Eemods 20.05 11h)

by extracting from the from the impedance so

24:33

S… Speaker 1 (Eemods 20.05 11h)

there's not really any

24:38

S… Speaker 1 (Eemods 20.05 11h)

systematic to this but there's one thing that could perhaps help

24:42

S… Speaker 1 (Eemods 20.05 11h)

us a little bit

24:43

S… Speaker 1 (Eemods 20.05 11h)

Because we could say that we don't want the resistor to

24:47

S… Speaker 1 (Eemods 20.05 11h)

interfere with the fundamental frequency

24:51

S… Speaker 1 (Eemods 20.05 11h)

of the motor.

24:53

S… Speaker 1 (Eemods 20.05 11h)

So we could say that the parallel connection of

24:57

S… Speaker 1 (Eemods 20.05 11h)

the parasitic resistor and the leakage

25:02

S… Speaker 1 (Eemods 20.05 11h)

inductance of the motor

25:04

S… Speaker 1 (Eemods 20.05 11h)

that will basically create a pole I call it the leakage pole and

25:08

S… Speaker 1 (Eemods 20.05 11h)

we can then design this leakage pole we can choose the resistive

25:13

S… Speaker 1 (Eemods 20.05 11h)

component because we know what the inductor component is then we can choose

25:17

S… Speaker 1 (Eemods 20.05 11h)

the resistive component to at least not interfere

25:21

S… Speaker 1 (Eemods 20.05 11h)

with the fundamental problems of the machine then we have not ruined

25:26

S… Speaker 1 (Eemods 20.05 11h)

anything so we

25:30

S… Speaker 1 (Eemods 20.05 11h)

will not

25:32

S… Speaker 1 (Eemods 20.05 11h)

we will not ruin the fundamental properties of the machine but we will to some

25:36

S… Speaker 1 (Eemods 20.05 11h)

extent include at least some of the trends that we

25:41

S… Speaker 1 (Eemods 20.05 11h)

know will happen in the switching frequency range

25:45

S… Speaker 1 (Eemods 20.05 11h)

that are otherwise completely ignored so it's a little bit pragmatic here

25:49

S… Speaker 1 (Eemods 20.05 11h)

but this could be an approach to at least set some limits to the range

25:53

S… Speaker 1 (Eemods 20.05 11h)

of what this resistor would be

26:00

S… Speaker 1 (Eemods 20.05 11h)

Because at this stage of model sophistication that we have,

26:04

S… Speaker 1 (Eemods 20.05 11h)

then I would say that any improvement is better than the alternative.

26:08

S… Speaker 1 (Eemods 20.05 11h)

There could still be a lot of work in this,

26:12

S… Speaker 1 (Eemods 20.05 11h)

but I will end it here and say thank you very much.

Eemods 20.05 11h

خلاصه

از هوش مصنوعی در مورد این رونوشت بپرسید