In order for Simulink to perform its tuning capabilities, it must be able to linearize the plant model. Simulink performs linearization using a block-by-block approach, then combines the individual block linearizations to product the overall system linearization. The linearization of each block can use one of two different methods:
1) Analytical linearization: obtain the Jacobian from the block
2) Numerical perturbation: perturb the block by a small amount and compute a linear model based on the response to these perturbations
Strong discontinuities in a block (such as those generated by PWM signals) can cause poor linearization results, e.g. linearizing to zero or an infinite gain. Therefore, such blocks require special handling.
For more on this, see the following page, which describes common problematic blocks and potential solutions for linearizing them:
Refer to the following page when troubleshooting linearization results in general, which provides a number of useful resources and approaches for obtaining better linearization results:
To handle poor linearization results, you can specify a custom linearization for a block if it is known. This can be down using either of these two approaches:
1) Specify using a MATLAB expression: https://www.mathworks.com/help/slcontrol/ug/specify-linear-system-for-block-linearization-using-matlab-expression.html
2) Specify using a function: https://www.mathworks.com/help/slcontrol/ug/specify-d-matrix-system-for-block-linearization-using-function.html
An alternative approach is to use the System Identification Toolbox to identify a linear system for the block. An example of how to do that can be found here:
