Hi @Gabriele
If my understanding is correct, the reason it linearizes the Lookup Tables to zero is that such blocks are classified as discrete blocks and, therefore, do not have defined Jacobians. Although these blocks are programmed to produce an output value by interpolating a table of input values according to the numerical method specified in the block parameters, the Jacobian exists only when a mathematical function is available to describe the input–output relationship, 
, and when the function is differentiable at the point of interest.
, and when the function is differentiable at the point of interest.Since the data is available to construct the lookup tables, you can use MATLAB's curve-fitting tools to find a mathematical function that creates a visually smooth curve between the data points, clearly illustrating the underlying trend for each table. You can then replace the lookup table block with the user-defined function (Fcn block) and specify the mathematical expression in the block parameters. The mathematical functions for 1-D Lookup Tables are the simplest to find as long as the data points do not exhibit the piecewise function-like trends.
If you are having trouble finding suitable mathematical functions for some lookup tables, please create a new post for each table (as they are considered individual math problems) and share the corresponding data.
Note: To improve the user experience, I believe that documentation for some Simulink blocks should indicate whether these blocks are linearizable or if linearization is not applicable.
