Diagnostic analyses of concrete dams on site, by flat jack tests, computer simulations and parameter identifications

Title: Diagnostic analyses of concrete dams on site, by flat jack tests, computer simulations and parameter identifications
Author(s): G. Cocchetti, R. Fedele, T. Garbowski, G. Maier, A. Tavakoli
Paper category : conference
Book title: RILEM Symposium on On Site Assessment of Concrete, Masonry and Timber Structures - SACoMaTiS 2008
Editor(s): L. Binda, M. di Prisco, R. Felicetti
Print-ISBN: 978-2-35158-061-5
e-ISBN: 978-2-35158-075-2
Publisher: RILEM Publications SARL
Publication year: 2008
Pages: 571 - 579
Total Pages: 9
Nb references: 9
Language: English

Abstract: In this communication an experimental-numerical method is outlined for the in situ assessment of possibly deteriorated elastic properties, tensile and compressive strength of dam concrete, and for the estimation of the local stress states in concrete dams. Such novel method is based on computer simulations and inverse analysis combined with the traditional flat jack technique. The following operative stages have to be carried out in situ: generation of suitably located slots on the dam surface; application of pressure by flat jacks; measurement of consequent displacements; identification of the sought parameters by inverse analyses based on the experimental data. The identification of elastic moduli and stresses is achieved by two slots; the assessment of tensile and compressive strengths require five and four slots, respectively. In each test, the relative displacements of monitored points are measured during the application of the pressure by the flat jacks in the selected slots. Each experiment is simulated by a finite element model. In a deterministic framework, the unknown parameters are obtained as a solution of a (nonlinear) mathematical programming problem in which a suitable objective function is minimized, specifically a function which quantifies the discrepancy between the experimentally measured quantities and those computed by means of the mathematical model as a function of the unknown parameters. Such minimization of the discrepancy function is performed in situ, by means of a portable computer endowed with an artificial neural network trained once for all through the above finite element model. The parameter identification in a stochastic context can be achieved by means of a modified Bayes technique, leading, in a batch way, to parameter estimates endowed with a covariance matrix, which quantifies their degrees of confidence. The main features and some numerical validation tests of the proposed method are presented and its novelties and potentialities are briefly discussed.

Online publication: 2009-05-26
Publication type : full_text
Public price (Euros): 0.00