Comprehensive mechanical property classification of vapor-grown carbon nanofiber/vinyl ester nanocomposites using support vector machines

• Support vector machines were used to classify VGCNF/VE dataset into 10 classes.
• Resubstitution and 3-folds cross validation techniques were applied.
• The mechanical response value(s) resulting from untested inputs is/are identified.
• The manufacturing lead time of VGCNF/VE will be significantly reduced.
• Usefulness of data mining and knowledge discovery in materials science was proved.

In the context of data mining and knowledge discovery, a large dataset of vapor-grown carbon nanofiber (VGCNF)/vinyl ester (VE) nanocomposites was thoroughly analyzed and classified using support vector machines (SVMs) into ten classes of desired mechanical properties. These classes are high true ultimate strength, high true yield strength, high engineering elastic modulus, high engineering ultimate strength, high flexural modulus, high flexural strength, high impact strength, high storage modulus, high loss modulus, and high tan delta. Resubstitution and 3-folds cross validation techniques were applied and different sets of confusion matrices were used to compare and analyze the classifier’s resulting classification performance. The designed SVMs model is resourceful for materials scientists and engineers, because it can be used to qualitatively assess different nanocomposite mechanical responses associated with different combinations of the formulation, processing, and environmental conditions. In addition, the lead time required to develop VGCNF/VE nanocomposites for particular engineering application will be significantly reduced using the designed SVMs classifier. This work specifically present a framework for a fast and reliable classification of a large material dataset with respect to desired mechanical properties, and can be used for all materials within the context of materials science and engineering.

Figure optionsDownload as PowerPoint slide