Intelligent prediction modeling of the post-heating mechanical performance of the brick powder modified cement paste based on the cracking patterns properties

This paper presents an intelligent modeling approach for the prediction of compressive and tensile strength of thermally degraded cement matrix modified with the brick powder. For this purpose, artificial neural networks were used, the support vector machine approach and classical functional models were used for comparison. What indicates the novelty of the developed model is the fact that as inputs were used the quantitative characteristics of the cracking patterns (CPs), which were formed due to the elevated temperature interaction. To date, such an approach has not been reported in the literature so far. The conducted research indicated that the models based on the CPs parameters have very high accuracy and even higher accuracy than those based on a very popular measure, i.e., the ultrasonic pulse velocity. This makes that such an approach can be successfully applied in engineering practice for prediction of mechanical characteristics of thermally degraded cementitious composites.