Invalid dimensions between blocks in Simulink

Hi!

I'm struggling with the error message of invalid dimensions between two blocks - see figure from simulink.

The gain block uses matrix Bd(set to matrix multiplication(K*u)) which is a 5x2 matrix. The random number has nothing fancy with it, in this case just noise. I've tried using the Signal Specification block as seen above to make the signal appear as a 1x5 vector, but this fails in the same manner.

The errors say "output port widths or dimensions of the Signal specification is a [1x5] matrix"

and "Invalid dimensions specified for input port of gain. Gain has dimensions [5x2] while input is set to dimensions [1x5]. These dimensions is not suitable for matrix product with the input signal"

Any helpers who could help here? Much appreciated!

Answer 1

[A Jenkins]

Do you want a 1x5 array of different random numbers, or the same 5 times?

Either way, if you have u=[1x5] and Bd=[5x2], then you need u*Bd, not Bd*u.

[Andreas Burheim Volden]

Thanks for that! The idea is to use the same block 5 times and create a vector from that particular block that coincides with the gain block as illustrated.

This is supposed to be the state noise component of an LTI system on the form:

xdot = A*x + B*u + Bd*noise

Thats why the figure has matrix multiplication K*u instead of u*K as you suggest. Wouldn't this lead to a different system or am I wrong here? Cheers!

[A Jenkins]

You should be able to build any size u matrix you want using "Vector Concatenate".

However, the inner dimensions of your matrix product must agree. So if your system has p inputs, q outputs, and n state variables, then x=nx1, A=nxn, B=nxp, and u=px1.

What is the number of states, n, in your system? If you want Bd*noise and have r noise inputs, then Bd=nxr and noise=rx1.

[Andreas Burheim Volden]

Thanks again!

For this system, I have 5 states and two inputs and two outputs, thus a 2x2 system.

Bd is a 5x2 matrix, and noise should, as you point out, be a 2x1 matrix. This would make all the summing terms, A*x + B*u + Bd*noise, 5x1 vectors which coincide with the state vector derivative xdot. And thus by integrating xdot I get back to the state vector x which also is [5x1]. Cheers!