In general, when using trainedModel.predictFcn, the observation data that is passed to trainedModel.predictFcn should contain the same features (same columns) that were in the original dataset that was loaded into the Regression Learner app. The predictFcn will handle the feature selection and PCA, if any. Given the error message seen above, it looks like perhaps 'Tyre' was not one of the featues loaded into the Regression Learner app when building that model.
To look into this further, please share samples from the model training data, the trained model, and samples of the test data.
For examples of using models exported from Learner Apps in App Designer, see these two answers:
If this answer helps you, please remember to accept this answer.
