Model Order Reduction- The next BIG thing for Advanced Electric Drive

We electrical engineers are like the small brother of our big brother- the mechanical engineer. We borrow so many nice ideas from them including Finite Elements, and the latest in the sleeve is this Model Order Reduction. Let me assure you in my humble opinion that it will pave the way for the 'Next Generation' of electric drive like : DIGITAL TWIN where we are more focussed on the condition monitoring, system performance and efficiency than individual performance.

What is Model Order Reduction?

You can think the Model order reduction something like a FFT (No exactly, just an analogy) in the Galerkin state space. Think of it like the frequency spectum for your normal machine where you take the most important (most of the time the fundamental) frequency to to calculate the motors behaviour which gives you a reasonable accuracy. Model order reduction also do the same, but unlike fourier transformation, you need to do proper orthogonal decompostition on your finite element system. Though there are other methods but lets ignore that part now. What you do is that you approximate the higher order system with a reduced lower order system, and smartly projected with some operator the reduced solution in the original state space. Sound complicated huh!!

Some simple Example

Think of a higher order polynomial equation, and you choose only few order (terms) that gives you some close approximation of the original equation. Another analogy of order reduction

Lets see the picture below. Consider , 400 rank are needed to perfectly describe the picture G, then you will see how with lower approximation of model order reduction we can almost get quite good resemblance (rank 30) of the orginal picture (rank 400).

How it works for Electrical Machine?

The reason we cannot use the super accurate finite element models in our real time controller is the incumbency of the computational time. It slow down everything as the real time needs computation in the micro seconds order. So here comes the Order Reduction.

It simply reduces the finite element order of the original system of equations and take into consideration the only important sectors of the machine. For example, the airgap, the teeth, the shape of the permanent magnet etc, and gives you a close approximation of the original finite element model.

So What ?

Well, it is a method which is clearly much more complex and accurate than the simplistic lumped parameter model and takes considerably less time than the original model. So if you have a non-linear system, or highly changing system, then even with a simple control strategy you can use order reduction model, and it will give you super awesome result.

I have personally investigated a little bit this subject and finds it fascinating. You can read our paper here, and let me know your opinion. This is still a very new thing, but certainly and eventually it will start making drives awesome.

https://ieeexplore.ieee.org/document/7842577/

Thank you for reading this. If you want more information, feel free to contact us. We would be glad to co-operate cuz' Sharing is caring!!!