Estimación de la incertidumbre de un perceptrón multicapa para la modelización del ensayo de resistencia a compresión del concreto de alta resistencia mediante la aplicación del método de Monte Carlo

• Autores: Isabel Moromi Nakata, Francisco García Fernández, Ana Victoria Torre Carrillo, Pedro Espinosa, Luis Acuña Pinaud
• Localización: Revista de la Construcción, ISSN-e 0718-915X, ISSN 0717-7925, Vol. 17, Nº. 2, 2018, págs. 319-329
• Idioma: español
• Enlaces
  • Texto completo (pdf)
• Resumen

  • The use of artificial neural networks as a modelingtoolfor the physic-mechanical properties of diversematerials has experienced greatadvances in the last ten years,mainly dueto the increased in computing capacities of computers.This technique has been used in many different fieldsof science and its effectiveness is sufficiently proven. Its applicationin the particle board industry complies with the requirements of the test regulations for the use in production control, as an alternative methodto normalized one. However, in spite of providing a result with a great approximation, they do not indicate anything about the uncertainty of the result.This last point is crucial when the resultshaveto be comparedwith a product standard.There are internationally accepted deterministic techniques for obtaining the uncertainty of a test result, always starting from the knowledge of the function that relates the measure with the measurement parameters. However, these techniques are not entirely adequate for the case of excessively complex functions such as an artificial neural network. In these cases, the use of stochastic simulation methods such as the Monte Carlo method is more appropriate.In this article, an artificial neural network will be developed to obtain the compressive strength of high-strength concrete to later obtain the uncertainty by a Monte Carlo simulation