Classification Microstructure of Al-Si Casting Alloy by Deep Convolutional Neural Networks Technique

The Al-Si alloy is widely used in many industries because of its good fluidity. The addition of alloying elements helps to modify the eutectic silicon morphology from plate-like into a fine fibrous structure. This method, called "modification," offers better mechanical properties than the normal hypoeutectic Al-Si alloy. The evaluations of modification levels are based on an American Foundry Society standard, which describes each modification level using the wall chart microstructure. However, this method is quite dependent on human experience and could introduce human-caused errors into comparative metallographic analyses. To avoid the above problem, the development of a deep convolutional neural network model for the modification level prediction from the Al-Si microstructure image has been introduced. The development performs well on the classification task since it has an accuracy of 93.54%. The precision score, recall score, and F1-score of the testing set were 0.9404, 0.9345, and 0.9360, respectively.