Uso de redes neuronales en la optimización del proceso de diseño de mezclas de concretos reforzados con fibras con alto comportamiento de endurecimiento por deformación (HPFRCC)
...
Sánchez Díaz, Jairo Alfredo | 2020
High-Performance Fiber-Reinforced Cementitious Composites (HPFRCC) are used in earthquake resistant constructions where strain hardening and energy absorption capacity are relevant mechanical properties that determine the quality of the mixture. Optimal mixing dosing is usually achieved by experimentation in laboratories, an activity that is a costly as the tests are done with a limited number of dosages.
Currently, there are viable and globally proven dosages with components that are not available in Colombia and have a high import cost. The use of national materials is a viable alternative, however, it is necessary to know the behaviour of the mechanical properties of the HPFRCC that includes materials from the national market. For this reason, the use of Artificial Neural Networks (ANN) is an alternative that provides the possible results of experimentation with high precision at a low cost.
This study proposes different types of ANN architectures, which are trained and evaluated. Two architectures types are selected to estimate the strain hardening and energy absorption capacity of HPFRCC dosages that include fibers available in Colombia and their respective cost. In addition, here are presented dosages with the highest probability of success in strain hardening greater than 0.3% and energy absorption capacity greater than 50 kJ/m3, at the lowest cost. In this way, decision-making is contributed for the development of HPFRCC with materials from the national market.
LEER