System identification relies on which type of data to build a plant model?

• A. Internal state logs from the controller.
• B. Input-output data.
• C. Color and branding information of components.
• D. Weather data during testing.

Answer: (B)

The central idea is that you model a plant by observing how it responds to known inputs over time. To build a model, you need pairs of input and output signals that reflect the system’s true dynamics. By applying a known input (like a step or ramp) and watching the resulting output, you can infer how the plant behaves, including delays, time constants, and how it reacts to disturbances.

Internal state logs from the controller don’t reliably reveal the plant’s dynamics because they show only what the controller did, not necessarily how the plant itself responds, and some states may be hidden or altered by the control law. Color and branding information of components don’t affect the physical behavior of the system. Weather data during testing can introduce external disturbances but isn’t the data used to identify the plant’s internal relationship between input and output.

So, the data used to build the plant model are the input-output measurements: what you apply to the system and how the system responds, captured over time.