Minimization of Internal Shrinkage in Castings Using Synthesis of Neural Networks

Abstract

New methodologies need to be evolved to minimize the shrinkages in castings and to enhance the process of direction solidification so that sound castings can be made. Existing foundry practices employed to achieve these objectives are based on the subjective judgement of the foundry experts. Present state-of-art processes followed to achieve these objectives are based on trial-and-error production, which are intolerably subjective and do not guarantee a satisfactory result. To this end, an intelligent shrinkage minimization module is required which can learn the real behavior of the solidification process so that it can perform the task of casting design feature modification in real time and intensify the process of directional solidification for a given casting. In this research, synthesis of two NN models, such as K-SOM network and BPN, are adopted to tackle the underlying problem. In the test problem considered, the NN-based model was able to minimize the shrinkages by accurately modifying the casting design features and augmenting the process of directional solidification. The proposed methodology simplifies the task of foundry designers to perform casting design modifications in a more flexible and intelligent manner.