Linear elastic methods of analysis are commonly used to assess the flexural strength of reinforced concrete beam-and-slab bridges. In certain cases, they will produce overly conservative estimates. Plastic and non-linear finite element (NLFE) analyses account for both plastic behaviour and global failure mechanisms. Consequently, they should predict ultimate load more accurately.
This thesis applies NLE, yield-line, and two linear elastic analyses to three reinforced concrete beam-and-slab deck models, two of which were constructed by the author and one by Hazell (Hazell, 1999). The decks built by the author had the same dimensions and reinforcement but were subjected to different double-axle load configurations. The model constructed by Hazell had heavier reinforcement, thinner slab sections, and was loaded by a single-axle arrangement. All three decks were designed to discourage the formation of shear and punching mechanisms.
An NLFE calibration study conducted on a previously tested slab deck (Collins, 1997) and a parameter sensitivity study were used to determine the optimal combination of parameter choices and values that were selected to analyse the three beam-and-slab models. The NLFE analysis was found to be highly sensitive to feasible variations in the mesh size, the types of elements selected (including integration points), the use or absence of numerical convergency aids, the definition of the tension stiffening curve, the tensile strength of concrete, and the stiffness of end restraints. Similarly, an exhaustive array of mechanisms and dimensions were analysed for the yield-line method in order to optimise the reliability of the ultimate flexural strength estimates.
The predictions of all the analyses were compared with the results of three experimental tests, one of which was conducted by Hazell (Hazell, 1999), and were all found to be consistently conservative. The yield-line analysis (COBRAS, 1999) yielded the most accurate prediction for the deck employing normal reinforcement. It was more conservative for the two lightly reinforced models. The discrepancies between observed and predicted ultimate load capacity were consistently due to an underestimation of the number of beams participating in the failure mechanism. This underestimation was attributed to the neglected strengthening of the slab segments due to transverse compressive membrane action.
The NLFE analysis undertaken using the ABAQUS program (ABAQUS, 1998) produced disappointing estimates of the ultimate flexural capacity for all three models. The errors were attributed to a consistent underestimation of the load at which yielding of the steel in the loaded beams initiated and the tendency of the NLFE program to consider failure to have occurred once acute flexural cracking has developed in the loaded beams. The experimental tests demonstrated that reinforced concrete beam-and-slab decks might be capable of carrying additional load beyond this point. Post-test NLFE analyses were also carried out using more rigorously defined tension stiffening curves as opposed to the default linear definition. This step improved the accuracy of the flexural capacity predictions with varying degrees of success.
The elastic analyses (ACI 318-95, 1995) (ACS, 1997) yielded estimates of ultimate live load that were more conservative than the best predictions of both alternate analyses for all three models.
The computerised yield-line method used in this thesis (COBRAS) is recommended
over the NLFE analysis (ABAQUS) for the assessment of beam-and-slab
decks in cases where elastic grillage methods are likely to yield overly
conservative estimates of flexural capacity. These cases include decks
that are unlikely to be fully loaded across their transverse width.
Grillage analysis is recommended for any loading arrangement if the deck
is suspected of not having sufficient ductility to enable the formation of
the complete failure mechanism predicted by the yield-line analysis.
[Cambridge University | CUED | Structures Group | Geotechnical Group]