Artificial Neural Networks Methods to Analysis of Ultrasonic Testing in Concrete

Nondestructive testing (NDT) techniques are useful tools for analyzing reinforced concrete structures. The use of ultrasonic pulse velocity (UPV) measurements enables the monitoring of changes in some critical characteristics of concrete over the service life of a structure. The interpretation of the data collected allows an assessment of concrete uniformity, and can be used to perform quality control, to monitor deterioration and even, by means of comparison against reference samples, to estimate compressive strength. Nonetheless, the current techniques for UPV data analysis are, on a large degree, based on the sensitivity of the professionals who apply these tests. For accurate diagnosis it is necessary to consider the various factors and conditions that can affect the results. To proper control and inspect RC facilities it is essential to develop appropriate strategies to make the task of data interpretation easier and more accurate. This work is based on the notion that using Artificial Neural Networks (ANN) is a feasible way to generate workable estimation models correlating concrete characteristics, compacity and compressive strength. The goal is to determine if it is possible to establish models based on non-linear relationships that are capable of estimating with good accuracy the concrete strength based on previous knowledge of some basic material characteristics and UPV measurements. The study shows that this goal is achievable and indicates that neural models perform better than traditional statistical models. For the data collected in this work, provided by various researchers, traditional regression models cannot exceed R² = 0.40, while the use of ANNs allows the creation of models that can reach a determination coefficient R² = 0.90. The results make clear that, besides contributing to better the analysis of situations where there is doubts regarding concrete strength or uniformity, neural models are an efficient way to order and transfer unstructured knowledge. It was shown that, given the learning capacity and its ability to generalize acquired information into mathematical patterns, ANNs are a quick and adequate way to model complex phenomena.

References
1. Associacao Brasileira de Normas Technicas, Projeto de Estruturas de Concreto. - NBR 6118, Rio de Janeiro, 2002. 2. Comite Euro-International du Beton, CEB-FIP Model Code 1990. London: Thomas Telford, 1993. 3. Neville, A.M., Propriedades do Concreto. São Paulo: PINI, 1997. 828 p. 4. Bittencourt, G., Inteligência Artificial: Ferramentas e Teorias. Florianópolis: Editora da UFSC, 2001, second edition. 5. Lorenzi, A., Aplicação de redes neurais artificiais para estimativa da resistência à compressão do concreto a partir da velocidade de propagação do pulso ultra-sônico, Porto Alegre, 2009, Tese (doutorado) – Programa de Pós-Graduação em Engenharia Civil, Universidade Federal do Rio Grande do Sul. Escola de Engenharia, 196p. 6. Lorenzi, A., J.L. Campagnolo and L.C.P. Silva Filho, “Application of Artificial Neural Network for Interpreting Ultrasonic Readings of Concrete,” International Journal of Materials and Product Technology, Vol. 26, No. 1-2, pp. 57-70, 2006. 7. MACHADO, M. D., Curvas de Correlação para Caracterizar Concretos usados no Rio de Janeiro por Meio de Ensaios Não Destrutivos. 2005. 265p. Dissertação (Mestrado) – Programa de Pós-Graduação em Engenharia da Universidade Federal do Rio de Janeiro. Universidade Federal do Rio de Janeiro, Rio de Janeiro, 2005. 8. NOGUEIRA, C. L., “Análise Ultra-Sônica da Distribuição dos Agregados no Concreto através de Wavelets,” Proocedings of the XXI Congresso Nacional de Ensaios Não Destrutivos, ABENDE, 2002.
Metrics
Usage Shares
Total Views
12 Page Views
Total Shares
0 Tweets
12
0 PDF Downloads
0
0 Facebook Shares
Total Usage
12