Reverse 2D-lookup table in Simulink

I tried to implement a reverse 2D lookup table in Simulink according to the following description by Fangjun Jiang:

https://de.mathworks.com/matlabcentral/answers/12809-reverse-2d-lookup-table?s_tid=sug_su

First, I wrote the following Matlab code for testing and it does exactly what I expect:

    % Absolute eccentricity range    
    e = 0:1e-06:1.5e-05;
    % Piston phase angle range
    rad = 0:2*pi/400:2*pi;
    % Load capacity matrix
    load('Tragkraft_alpha_AK_kpl_V03.mat');
    load_cap = Tragkraft_alpha_AK_kpl_V03;
    % Instant piston phase angle
    rad_dyn = 3.12588469032184;
    % Reverse 2D lookup table for load capacity
    lookup = interp2(e,rad,load_cap,e,rad_dyn);
    % Instant load capacity
    load_dyn = 659.331131439451;
    % Instant absolute eccentricity
    e_dyn = interp1(lookup,e,load_dyn);

After that, I tried to implement it in Simulink using a MATLAB function block with the following code:

function e_dyn = fcn(rad_dyn, load_dyn, load_cap)

    % absolute eccentricity range  
    e = 0:1e-06:1.5e-05;
    % piston phase angle range
    rad = 0:2*pi/400:2*pi;
    % 2D lookup table for load capacity
    lookup = interp2(e,rad,load_cap,e,rad_dyn);

% instant absolute eccentricity
e_dyn=interp1(lookup,e,load_dyn);

However, this leads to the following error message that I don't really understand. In my understanding the Simulink model should do exactly the same as the Matlab code. Can anybody tell me how to fix that in Simulink?

[VBBV]

rad_dyn = rad_dyn*ones(1,length(e))

[VBBV]

That difference is probably due to Simulink block functionality which puts constraints on input data as opposed to Matlab coder. Moreover you can also use mixed array sizes for interp2 function as you want (scalar) for the purpose of reverse look up table . You can browse through the doc page of interp2 function output argument section for interpolated values (Vq).

[Bruce]

Thanks, I just tried that, but it results in the same error message. I still don't understand why the code works in Matlab but not in Simulink.

... attached are the original files for testing (if required)

Answer 1

[Bruce]

I noticed that there was already an explanation of this issue in the link I mentioned above. The MATLAB function block in the Simulink model requires the following changes in order to work:

• use of "griddata" instead of "interp2"
• add "coder.extrinsic('griddata');" at the top of the Matlab function
• e_dyn = 1; to Initialize the output at the top of the Matlab function
• change e from a row vector to a column vector

Finally, the code inside the MATLAB function block looks as follows:

function e_dyn = fcn(rad_dyn, load_cap, load_dyn)
    
    coder.extrinsic('griddata');
    e_dyn = 1;
    % absolute eccentricity range
    e = (0:1e-06:1.5e-05)';
    % piston phase angle range
    rad = 0:2*pi/400:2*pi;
    % 2D lookup table for load capacity
    lookup = griddata(e,rad,load_cap,e,rad_dyn);

% instant absolute eccentricity
e_dyn=interp1(lookup,e,load_dyn);

Although I have absolutely no idea why this works, the result is the same as the output of the Matlab function, so it seems to make sense. Also, for whatever reason, the output calculation becomes extremely slow.

[VBBV]

Ok. The key thing (line) is invoking the matlab built in function inside the Simulink function block although it was not recommended by @Fangjun Jiang

coder.extrinsic('griddata')

The rest of changes are not significant. You can try the same with interp2

Answer 2

[VBBV]

@Bruce Xq, Yq must be vectors , in your one is scalar and other is vector.

e = 0:1e-6:1.5e-05

e = 1×16

1.0e-04 * 0 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800 0.0900 0.1000 0.1100 0.1200 0.1300 0.1400 0.1500

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

rad_dyn = 3.12588469032184

rad_dyn = 3.1259

    % piston phase angle range
rad = 0:2*pi/400:2*pi

rad = 1×401

0 0.0157 0.0314 0.0471 0.0628 0.0785 0.0942 0.1100 0.1257 0.1414 0.1571 0.1728 0.1885 0.2042 0.2199 0.2356 0.2513 0.2670 0.2827 0.2985 0.3142 0.3299 0.3456 0.3613 0.3770 0.3927 0.4084 0.4241 0.4398 0.4555

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

% piston phase angle range  rad = 0:2*pi/400:2*pi;
load_cap = rand(401,16)

load_cap = 401×16

0.9390 0.7404 0.0375 0.4748 0.5578 0.6625 0.5913 0.8934 0.5613 0.6913 0.5698 0.7728 0.0915 0.6569 0.2616 0.7420 0.5898 0.3545 0.1735 0.8986 0.2358 0.3774 0.2710 0.3005 0.3594 0.7609 0.9310 0.2273 0.0780 0.8546 0.3428 0.1685 0.7407 0.6071 0.3289 0.3032 0.8605 0.3039 0.3185 0.0074 0.0197 0.2289 0.6588 0.1055 0.7994 0.7097 0.3390 0.9664 0.0773 0.7881 0.2613 0.0009 0.1449 0.3495 0.3655 0.7452 0.7094 0.7467 0.5933 0.4940 0.2385 0.9605 0.5026 0.7729 0.0445 0.4355 0.5632 0.1824 0.6726 0.5115 0.8234 0.6539 0.3724 0.1295 0.5974 0.7898 0.0420 0.1607 0.4936 0.3356 0.8658 0.5376 0.6603 0.7400 0.4651 0.4632 0.2891 0.1489 0.7257 0.9054 0.9649 0.8331 0.3219 0.9556 0.5288 0.6005 0.4665 0.5594 0.4608 0.6730 0.9842 0.8845 0.0180 0.4513 0.4014 0.4267 0.4705 0.8162 0.5023 0.2621 0.2733 0.6841 0.1051 0.8237 0.5663 0.0619 0.1383 0.9129 0.2330 0.1048 0.6653 0.6792 0.1665 0.9511 0.4487 0.1645 0.9818 0.9896 0.1134 0.5379 0.3903 0.0331 0.6306 0.6620 0.4031 0.7992 0.4596 0.3366 0.7462 0.0770 0.5829 0.2536 0.5996 0.1671 0.0019 0.2616 0.1290 0.7993 0.0818 0.5686 0.7396 0.6163 0.9185 0.4611 0.7098 0.9392 0.2135 0.0523 0.2622 0.1000

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

lookup = interp2(e,rad,load_cap,e,rad_dyn*ones(1,length(e)))

lookup = 1×16

0.5883 0.9483 0.0752 0.8895 0.3999 0.4260 0.4265 0.6782 0.3920 0.4980 0.5230 0.6050 0.0885 0.7243 0.5768 0.0679

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Bruce]

Thanks for the answer, but then I do not understand why it does work in Matlab with essentially the same code. Futher, providing a scalar as one input is actually the intention of the reverse lookup table. It shall provide one breakpoint as an output based on the table data and a dynamic changing (scalar) breakpoint value as an input.