Investigation into hybrid data mining and soft computing techniques to aid to design of composite joints

The evaluation and prediction of the failure probability and safety levels of composite components and structures is of extreme importance in structural design and manufacturing. A new application of data mining techniques for predicting the behavior of pin-loaded composite joints is presented. The proposed system consists of combining different data mining and soft computing techniques such as classification and clustering with fuzzy logic. By using these techniques, the relationship between different parameters, such as edge distance and tensile strength of composite joints, is modeled. A classification approach based on fuzzy clustering yielded the best predictive results.