Uso de redes neurais artificiais na determinação de dosagens para produção de blocos de concreto com resíduos da construção civil.

Abstract

This work aims at investigating the performance of concrete blocks produced with civil construction wastes and the use of Artificial Neural Networks (ANNs) to define a model capable of predicting the efficiency of unknown dosages. The procedure consisted in molding structural and non-structural blocks (with materiais duly characterized) and obtaining its physical-mechanic characteristics, emphasizing the compressive strength and the water absorption. The tests were carried out at 7 and 28 days of cure. The substitutions of natural. aggregates by civil construction wastes were in the proportions of 0%, 30%, 60% and 100%. Most of the results obtained for the alternative blocks showed values in accordance with the limits established by the Brazilian Association of Technical Norms - ABNT norms, what contributes to the increase of the credibility in components manufactured with civil construction wastes. After this phase, characteristics of the compositions of molded blocks were related to the results of compressive strength and water absorption through ANNs. It was modeled an ANN for the structural blocks and another for the non-structural ones. Both ANNs presented results equally reliable, what revealed them sufficiently useful in the definition of theoretical supports for the study of waste incorporation in civil engineering components. These models may be thus useful for reducing time and expenses with experiments durihg the determination of the ideal composítion in alternative concrete blocks.