Alternative Methods for Porosity Prediction in Aluminum Alloys

The occurrence of porosity during metal solidification is one of the major issues that impact the quality of castings. Quantitative information on the development of porosity is particularly important for safety critical components, such as automotive chassis parts and airframe primary structures. In this paper, we present two approaches to predict the location and volume fraction of porosity for aluminum alloy A356. In the first approach, the application of Neural Networks to predict porosity is examined. Results are compared with the established criteria functions and reported experimental findings. Neural Networks are shown to predict the occurrence of porosity with higher confidence than the existing thermal parameter based criteria functions. In the second approach, microporosity evolution is modeled mathematically. This model incorporates various solidification phenomena such as dendrite formation and growth, hydrogen evolution at the solid/liquid interface, solidification shrinkage, interdendritic fluid flow, and formation and growth of pores. Preliminary results indicate that the porosity distribution predicted by the proposed model is in good agreement with experimental measurements.