Linearization error "NaN's cannot be converted to logicals"

I get the error. " NaN's cannot be converted to Logicals" when trying to linearize my model. The model runs actually with no problem. the problem arises only when linearizing the model in control design. I don't know what exactly simulink does during linearazation that causes this error.

[Ameer Hamza]

Can you give an example of your model?

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

Answer 1

[Adam Danz]

This will replicate your error:

v = [1 0 1 1 0 NaN 0 1 0 0]; 
logical(v)
Error using logical
NaN's cannot be converted to logicals.

My guess is that you're using logical() to convert a variable with 0s and 1s that also have NaN values.

If that's the case, you need to decide whether NaNs should be a 0 or a 1. In this example, I set NaNs to 0.

v = [1 0 1 1 0 NaN 0 1 0 0];
v(isnan(v)) = 0;  
logical(v)

Or you can remove the NaNs

v = [1 0 1 1 0 NaN 0 1 0 0];
v(isnan(v)) = [];  
logical(v)

[Hossein Rezazadeh]

The model runs actually with no problem. the problem arises only when linearizing the model in control design. So I don't think there is any problem with my code. Besides I am not converting anything to logical in my code. I don't know what exactly simulink does during linearazation that causes this error.

[Adam Danz]

As @Walter alluded to in the comment section under your question, you'll need to provide much more detail for us to blindly troubleshoot your problem.

[Hossein Rezazadeh]

Ok I got it. Apparently during the linearazation some numerical perturbation is fed to the block as input which causes "NAN" at the block output and this was triggering this error. adding v(isnan(v)) = []; to the block output solved the problem.

[Tim]

Hi, could you please explain how exactly you added v(isnan(v)) = []; to the block output?

Answer 2

[Barton Bacon]

This just happened to me and the culprit was a square root block I had recently added to a simulation.... believe it or not. Changing the square root block to a power block and using .5 as the exponent for some reason allowed the linearization to proceed without the "NaN's cannot be converted into logicals" error. MATHWORKS what the heck is going on in the square root and reciprocal square block that would produces this problem in linearization? Those blocks linearize fine by themshelves in a their own block diagram. In a complicated simulation, not so much apparently.

Answer 3

[Welid Benchouche]

hello guys, can you please help me , i get the same error on model predictive control when linearizing Rc=0.15

[Hossein Rezazadeh]

This happens in MPC when your model sends "NaN" to the "mo" inport of the MPC block.

As you have integrators in your model they should be responsible for "NaNs".

Either change the operating point at which you are linearizing the model or change the solver or max step size.

[Apicha Kittirattanachai]

please kindly explain "integrators in your model they should be responsible for "NaNs" "

How are integrators setting for response the NaN value ?

My problem is appear in mpc toolbox when linearize of toolbox.

Thankyou.

Answer 4

[Ahmed Elgohary]

Well, i got the same error with model predictive control during linearization and i could solve the problem by changing the range of simulink operating point as shown at the figure. So it was from [0 1] and i made it from [0.1 1] for example as you want for the upper limit but the main problem was related to using zero for lower limit during the linearization.