Ensaios de ondas ultrassônicas e redes neurais artificiais na avaliação da resistência à compressão do concreto

Abstract

Stripping of formworks in reinforced concrete structures can only be done when concrete is hardened enough to carry the loads without generating unacceptable deformations. These are two important demands from Brazilian Code NBR 14931-2004. To get those conditions the structural design engineer should inform the building engineer the minimum values of compressive strength and elasticity modulus that must be obtained, simultaneously, to strip of the formworks on the specified age. To determine the in situ compressive strength of concrete in an specific structural elements is usual to perform destructive tests using test standard samples made with the same concrete that will be used in the element. This situation, however, does not configure a strict attendance to the demands of the Brazilian Code since the conditions of the tests in standard samples are quite different from the concrete in the structural element. The research discusses the possibilities to adopt and follow those requirements using non destructive ultrasonic waves tests associated with artificial intelligence tools. Although the ultrasound test is relatively simple to perform, the interpretation of its results brings important difficulties, since it is influenced by several factors. The research, then, explored the possibility to use potentialities of Artificial Neural Nets simulations associated with ultrasonic wave tests to estimate compressive strength of the concretes. Two distinct ways to investigate the subject were used: experimental and numerical computational simulation. In the experimental program, nine different concrete mixtures, 162 standard test samples 10x20 cm and 27 prisms with dimension of 25x25x50 cm were made. The test samples were tested in direct compression on three different ages - 7, 28 and 60 days - and ultrasonic wave tests were made in the prisms on the same ages. With the results from experimental tests, computational simulations using Artificial Neural Networks to obtain a mapping among the problem variables length of the prisms, metacaulim content, aggregate diameter, age of the test samples and ultrasonic speed - and the output properties which was the compressive strength of the concrete. Obtained results showed that the simulations with Artificial Neural Networks together with ultrasonic wave tests are import tools that can help engineers to evaluate the compressive strength of in situ concrete