where to use System Identification Toolbox in 4 phases designing a Mechantronic system

we use System Identification Toolbox to identify the system (transfer function) from measured data. My problem is to where I can use it in the 4 phases designing a Mechatronic system (Please refer this link: http://en.wikipedia.org/wiki/Model-based_design). As this link says, we use toolbox in the first phase.. that means we need to have a system first, through experiences, we measure input-output data and use toolbox to identify. what kind of system is that?.a system with or without controller. Let me know if i didn't make myself clear. thank you,

Answer 1

[Arkadiy Turevskiy]

I understand your question. The wikipedia article is generalizing things a little bit. In practice, everything depends on a particular problem you are trying to solve. Here are a few scenarios:

1. You are developing a control system for a new device/machine that does not exist yet. To create an initial plant model in this case, you will need to use first-principles modeling (F=ma). Once you get the first prototype, you can start collecting test data and using it to refine your model. You could either create a black-box model (using data only and not taking physics into account at all), or a gray-box model (where you keep physics-based equations and use the data to tune equations parameters so that the model output lines up with the data). System Identification Toolbox can be used for both black-box and grey-box modeling in this scenario. Typically you would collect open loop data in this case.
2. You want to improve the existing controller that is already controlling the machine/device/process, but not as well as you would like to. Another common scenario here is that you changed your machine/device/process somewhat, and the old controller needs to be updated as well. In this case you would typically collect closed-loop data (with controller in the loop), and get a plant model to do control design.
3. Another (less common) application of system identification is model simplification. In this scenario you have a high-fidelity, physics-based model you use for control design and desktop simulation, but want to simplify it (for example, to be able to run it in real-time). You can use System Identification Toolbox for that purpose.

HTH.

Arkadiy