Statistical Parameters of Steel Rebars of Reinforced Concrete Existing Structures

Abstract

Historical and cognitive investigations supported by in-situ and/or laboratory tests are needed for a robust reliability assessment of existing structures. Indeed, an adequate knowledge of material properties and their statistical description is the basis for carrying out accurate reliability analyses and verifications on the investigated structures. In this paper, a procedure for the definition of pdfs of mechanical parameters of steel rebars is proposed based on secondary experimental test data. This information is very helpful for the reliability assessment of existing r.c. buildings, where estimation of statistical parameters of mechanical properties of steel reinforcement is very difficult. In fact. It must be highlighted on the one hand that direct information about the examined structure are commonly not sufficient, on the other hand that the number of rebar samples extracted from the structure, if available, is so limited that it does not allow a complete statistical analysis. The first step has been the collection of experimental acceptance tests carried out by Department of Civil and Industrial Engineering of University of Pisa on steel rebars of reinforced concrete (r.c.) structures during the 1960s. The yield strength and the tensile strength are extrapolated for each sample defining a significant database of experimental test results for existing r.c. structures. Then, probability distribution models for the mechanical properties of steel reinforcement have been defined as already done by the authors for concrete strength. A cluster analysis has been carried out based on the Gaussian Mixture Model applying the Expectation-Maximization algorithm to identify homogeneous material classes and their associated pdfs of material mechanical parameters. The main advantage of proposed procedure consists in its "blindness", In fact, not requiring subjective information like pre-classification of data, the methodology is not sensitive to alterations caused by engineering judgement or by inexact identification of declared strength class of the tested samples, due for example to downgraded materials.

Explainable Artificial Intelligence for Understanding the Ageing Classes of Reinforced Concrete Bridge Components

“…Asset managers identify the factors that determine the evolution of deterioration for each segment and develop a deterioration model for each HG. Recent studies have shown that HGs can be identified for building materials with respect to their mechanical properties as well as bridge components with respect to their condition evolution, by analysing digital MMSs with cluster algorithms (Croce et al, 2018(Croce et al, , 2020Marsili et al, 2023): such approaches, while very promising, have not yet been duly explored. In addition, it is important to emphasized that although it is appropriate to identify groups of objects or components that degrade at similar rates, the ability to recognize such groups based on the characteristics of the bridge and its environment must be analyzed a posteriori and not assumed a priori.…”

Influence of Reinforcing Steel Corrosion on Life Cycle Reliability Assessment of Existing R.C. Buildings

“…In the following section, the general methodology for the evaluation of statistical parameters of material properties from secondary material tests databases, which was adopted in [6] for concrete classes and in [16] for reinforcing steel classes, is briefly recalled.…”

“…When all the distributions of the mixture belong to the normal family, the mixture model is said to be a Gaussian mixture model (GMM) [27]. As discussed in [6] and [16], they have been successfully adopted to identify material resistance classes in a whole database of test results, even if the origin of individual data is unknown.…”

“…To overcome the lack of information about the mechanical properties of buildings' structural materials in terms of pdf s and relevant statistical parameters, suitable methods for the analysis of secondary material properties databases are needed. Concerning this aspect, a very sound and efficient methodology to identify homogenous material classes in databases of raw test results commonly stored in testing laboratory's archives has been previously proposed in [6,16]. Starting from standard acceptance test results obtained at the time of the construction, this methodology, based on a Gaussian mixture model, briefly described in Section 3, allows identification, in an "objective" way, of the relevant clusters of data, as well as their statistical properties.…”

“…To realistically represent the actual pdfs of the relevant material properties, the stochastic models for steel and concrete classes are those derived from the cluster analyses of the historical test data of the Laboratory of the Department of Civil and Industrial Engineering of the University of Pisa illustrated in the already mentioned references [6,16].…”