Principal Component Analysis combined with a Self Organization Feature Map to determine the pull-off adhesion between concrete layers

• New proposition of determination of the pull-off adhesion between concrete layers was presented.
• Principal Component Analysis and Self Organization Feature Map have been used in analysis.
• Genetic Algorithm was used to optimize the weights.
• Higher correlation than using conventional approach was observed.
• Presented analysis is applicable to the same grade of concrete with similar characteristics.

This study attempted to use Principal Component Analysis (PCA) combined with a Self Organization Feature Map (SOFM) to determine the pull-off adhesion between concrete layers. Also Genetic Algorithm (GA) was used to optimize the weights. Finally a constant model was selected among all of the PCA_SOFM combinatory models. To evaluate the precision of this model, it was compared to Multilayer Perceptron (MLP) model as well as Feed Forward (FF) model. The results indicated that the PCA_SOFM model had more ability, precision and flexibility in forecasting the pull-off adhesion between concrete layers parameter than the two mentioned models.