"Unable to resolve the name"

I am trying to code an optimization script to maximize the lift-drag ratio of a turbine blade, as functions of geometrical features on a Bezier curve. I have trained models using data generated from CFD on Regression Learner App, using Gaussian Process Regression, and I tried following this tutorial: https://www.mathworks.com/videos/part-4-problem-based-nonlinear-programming-1549458887351.html

I get the following error related to the models being called:

Unrecognized function or variable 'model_name'

For some reason, I can't get the script to run models (in the form of 1x1 structures) to recognize the models already loaded on the workspace. And this error... I actually don't understand:

Function evaluation failed while attempting to determine output size. The function might contain an error, or might not be well-defined at the automatically-chosen point. To specify output size without function evaluation, use 'OutputSize'.

Here's how my code looks:

Optimization problem:

LDRmax = optimproblem;

Variables

V_inf = optimvar('V_inf','LowerBound',10,'UpperBound',22);
Base_R = optimvar('Base_R ','LowerBound',3,'UpperBound',8);
LA_Ang = optimvar('LA_Ang ','LowerBound',1,'UpperBound',5);
UA_Ang = optimvar('UA_Ang ','LowerBound',1,'UpperBound',5);
UA_Height = optimvar('UA_Height ','LowerBound',0.001,'UpperBound',0.005);
LA_Height = optimvar('LA_Height ','LowerBound',0.001,'UpperBound',0.005);
X1 = optimvar('X1 ','LowerBound',0.09,'UpperBound',0.41);
X2 = optimvar('X2 ','LowerBound',0.09,'UpperBound',0.41);
UA_BA = optimvar('UA_BA ','LowerBound',45,'UpperBound',90);
LA_BA = optimvar('LA_BA ','LowerBound',45,'UpperBound',90);
AoA = optimvar('AoA ','LowerBound',-20,'UpperBound',20);
UA_BL = optimvar('UA_BL ','LowerBound',0.02,'UpperBound',0.1);
LA_BL = optimvar('LA_BL ','LowerBound',0.02,'UpperBound',0.1);

Multi-variate Optimization Expression:

edit ldrobj
% I transfer the objective function to an external function that should
% call and process the models laoded in the workspace. Matlab can't seem to do that
obj = fcn2optimexpr(@ldrobj,Base_R,LA_Ang,UA_Ang,UA_Height,LA_Height,X1,X2,UA_BA,LA_BA,LA_BL,UA_BL)
prob.Objective = obj
LDR = 1/DLR
CM

Supporting functions:

function f = ldrobj(Base_R,LA_Ang,UA_Ang,UA_Height,LA_Height,X1,X2,UA_BA,LA_BA,LA_BL,UA_BL)
x = [Base_R,LA_Ang,UA_Ang,UA_Height,LA_Height,X1,X2,UA_BA,LA_BA,LA_BL,UA_BL];
ALL = ALLGPO.predictFcn(x); % name.predictFcn are the trained models I used for this optimization problem,
AUL = AULGPO.predictFcn(x); % I am calling several of these. The script can't seem to see them
clear x;
x = [V_inf,Base_R,LA_Ang,UA_Ang,LA_Height,X1,X2,UA_BA,LA_BA,AoA,UA_BL,ALL,AUL];
Dyn_Vel = DynVelGPO(x);
clear x;
x = [V_inf,X2,LA_BA,AUL,Dyn_Vel];
y = [V_inf,LA_Ang,UA_Ang,UA_Height,LA_Height,X2,UA_BA,LA_BA,AoA,LA_BL,AUL,ALL,Dyn_Vel];
f = CdGPO(x)/ClGPO(y);
end

Please thank you!

[Aquatris]

ALLGPO and AULGPO object(?) are not defined in the ldrobj function space. Where do you define them? Did you try giving them as input to the ldrobj function?

[Roberto Louis]

Thanks for your response @Aquatris,

'...giving them as input to the ldrobj function", did you mean like this?

f = ldrobj(a lot of inputs, ALLGPO,AULGPO, etc...)

Where ALLGPO and AULGPO are two of the objects I'm trying to integrate. I just tried that now but the script still can't recognize them.

May I ask how else I could define objects in the function space?

[Githin George]

Assuming that ALLGPO and AULGPO are objects available in the MATLAB Base Workspace, you can use the "evalin" function to access these objects in the Function Workspace.

The usage is as follows:

v = evalin('base','ALLGPO')

This way you can access the objects in the base workspace without passing new arguments to your function.

For more information, refer https://www.mathworks.com/help/matlab/ref/evalin.html

[Roberto Louis]

I will try this @Githin George

Thanks!

[Stephen23]

The recommended approach is to pass them as input arguments, e.g.:

https://www.mathworks.com/help/matlab/math/parameterizing-functions.html