linearising a non-linear plant model

Hello Matlab experts,

I'm trying to linearise a non-linear step-response measured in in a lab experiment. The system is a basically a thermoelectric device (also known as TEC) with heat sinks on both (hot and cold) sides. The temperature is measured on both sides throughout the experiment and the input voltage into the system is logged as well. The nonlinearities of material properties of the TEC result in a rather unusual response to a 5V step input which I tried to linearise with Matlabs "System Identification Tool" (unsuccessfully). Using the system identification tool results and implementing them into Simulink as my plant model gives aboslute wrong responses when again exposed to a step input.

My questions are: Is that the right way of identifying and validating a plant model? Are there any alterntive ways of identifying a plant-model according to its input-output data that I'm unaware of?

Any help is greatly appreciated.

Thanks in advance. Marcus

Answer 1

[Guy Rouleau]

It looks like you are in the right direction:

- The System Identification Toolbox is the tool to identify the relation between input and output data measured experimentally. The important is to pick the right model and methods.... look at the doc and exampls for that.

- If you know what the equations of your system look like, have them implemented in a Simulink model and you need to identify the value of some parameters, then Simulink Design Optimization can help.