Casting defects such as gas and shrinkage porosity are common in aluminum alloys, which seriously affect their performance. In this study, a multi-scale prediction model for defects in cast aluminum alloys based on cellular automaton models was developed independently. To verify the accuracy of the model, X-ray computed tomography (XCT) was utilized to characterize the size and morphology of the microporosities in different parts of the components in three dimensions. A comparative analysis of the predicted and experimental results revealed that the model was capable of efficiently predicting the microporosities sizes in various parts of the components, with the most accurate predictions for the equivalent diameter of the microporosities, exhibiting an average relative error of approximately 25%. This multiscale prediction model can effectively accelerate the optimization of the wheel casting process, improve wheel production efficiency and reduce production costs.