Simulink Matlab function block output is inferred as a variable-size matrix, but its size is specified as inherited or fixed

Hello,

I have implemented the following simulink to generate a prediction using a trained neural network.

When I try to run the model, I receive the follwong error message:

'input4NN' is inferred as a variable-size matrix, but its size is specified as inherited or fixed. Verify 'input4NN' is defined in terms of non-tunable parameters, or select the 'Variable Size' check box and specify the upper bounds in the Size box.

The preprocessing of the data in the corresponding matlab function block looks as follows:

function [enablePrediction, input4NN] = fcn(IVT_Current, IVT_Voltage, maxTemp, minVoltage, SOClast)
    data = [IVT_Current IVT_Voltage maxTemp minVoltage SOClast];
    persistent inputStream;
    IPlength = 300;
    if isempty(inputStream)
        inputStream = data;
    else
        inputStream = [inputStream; data];
    end
    if size(inputStream, 1) >= IPlength
        enablePrediction = true;
        newestDataStream = inputStream((end-IPlength):end, :);
        input4NN = [newestDataStream(:, 1); newestDataStream(:, 2); newestDataStream(:, 3); newestDataStream(:, 4); newestDataStream(:, 5)];
    else
        enablePrediction = false;
        input4NN = zeros(1500, 1);
    end
end

I have read some forum entries regarding the variable-size error, but I was able to resolve it that way.

I need an IP vector of (1500, 1) of non-variabel-size for the prediction block. So OP of the preprocessing function should also be (1500, 1). I already created a variabel "input4NN" that es stored inside the model workspace and specified the dimensions of (1500, 1) as fixed.

I'm thankful for any further ideas/help.

Answer 1

[Fangjun Jiang]

There are issues with the code. The size of "inputStream" is going to grow indifinitely.

I suggest using a Buffer block to replace this MATLAB Function block.

If you don't have the Buffer block in your Simulink and its toolbox, I suggest assigning input4NN = zeros(1500, 1) at the beginning of the code. Use a counter to determine enablePredition. Assign input4NN like below

TempData = [newestDataStream(:, 1); newestDataStream(:, 2); newestDataStream(:, 3); newestDataStream(:, 4); newestDataStream(:, 5)];

input4NN(1:1500)=TempData(:);

[Fangjun Jiang]

Or, making this one line change should resolve the "input4NN is inferred ..." error.

input4NN(1:1500) = [newestDataStream(:, 1); newestDataStream(:, 2); newestDataStream(:, 3); newestDataStream(:, 4); newestDataStream(:, 5)];

[Elias Niepötter]

Thank you for the help, works now!