LOAD RATING AND ASSESSMENT OF MASONRY ARCH BRIDGES

Abstract

The load rating and assessment of masonry arch bridges, many of which are 5 to 10 decades old, is the most prodigious task on date so as to maintain and check the adequacy of such bridges, for present and envisaged future loads. Such bridges are present in surprisingly large numbers on the railway and highway networks of many countries. With the increase in the axle loads plying on these bridges and owing to increasing age, many of such bridges have distressed. A considerable amount of effort has been made in recent years to develop a rational and pragmatic procedure to establish load ratings for such bridges. The problem of load rating can be overtly divided in two distinct parts for carrying out research. The first part can be related to the prediction of the load carrying capacity in terms of the axle load on an immaculate arch. The evaluation of the modifying factors those consider the variation of the actual arch from the ideal, required to arrive at the allowable axle load can be treated as the second part. The first part mainly involves application of principles of continuum mechanics to determine the collapse load for the given configuration, geometry and material properties of an arch bridge. This can be dealt with quite rationally. On the other hand the evaluation of the condition factor is quite subjective and depends to a great extent on the experience and the judgement of the inspector/evaluator. Since BRS has carried out such an evaluation some times back, their recommendations regarding these factors have been widely accepted by many countries The primary objectives of the present study is to scientifically develop a load rating procedure for masonry arch bridges through the development of an analytical model and to validate the same to predict the allowable axle loads to a reasonable accuracy. The present study consists of the following distinct aspects. > Determination and review of existing research information; > Conducting tests in laboratory to examine the basic properties of the materials used and to examine the typical forms of the construction to provide reliable input to the proposed analytical model; > Experimental determination of the moment-axial force interaction for indigenous brick masonry construction and to compare the same with the models available in the literature; > Development of a computer aided mechanism based 2D frame analysis to determine the ultimate collapse load of the arch bridges, incorporating the failure criteria appropriate to indigenous masonry; iii > Development of a suitable 2-D non-linear finite element model incorporating effect of non-linearity arising from material constitutive relationship of masonry, as well as from the cracking and the crushing of masonry, to predict the complete response of masonry arch bridges including the collapse load. > Tests to failure on two series of arches, with two specimens each, to study the complete response including the failure mechanism. The results of the laboratory studies and those from the published literature have been used to validate the proposed analysis models; mechanism based approach and non-linear finite element model; > The application of artificial neural network (ANN) for load rating based on the experimental field-tests on real bridges. In the thesis, emphasis is focussed on the determination of the load carrying capacity of the masonry arch bridges to arrive at a safe axle load that can be allowed on any given masonry arch bridge. A computerized mechanism based analysis procedure that takes into account the interaction of axial thrust and the moments has been proposed. Unlike other mechanism methods in vogue, in the proposed formulation, the four obvious hinge positions are not selected, but, instead the analysis procedure itself determines the hinge locations and the corresponding load factors. The analysis stops when the requisite number of hinges has formed to convert the structure to a mechanism. For defining the failure criteria for the hinge formation at any section in the proposed analytical procedure, the moment - axial force interaction was investigated in laboratory by carrying out flexural testing on prisms, subjected to different levels of axial precompression ranging from 0 to 60 percent of the uniaxial masonry crushing strength. The collapse load prediction is based on the premises that an arch bridge will fail by conversion to a mechanism with the formation of four hinges in a fixed arch and a hinge will form when the value of axial thrust and bending moment at the section are such that the resulting point lies on the failure envelope. The finite element method is being used extensively in all fields of the civil engineering. For successful applications of finite element method for analysis of any structure a true representation of the material characteristics in different ranges of stress is essential. A limited experimental study was undertaken to determine the masonry properties viz. compressive strength, tensile strength, shear strength, and effect of precompression on the shear strength. In order to establish a constitutive relation for masonry, two types of control specimens were tested in laboratory. On the basis of the experimental results a representative mathematical third degree polynomial has been best fitted to define the material constitutive relationship of masonry for use in the proposed analysis. A two-dimensional finite element IV computer program has been developed in Fortran code to facilitate a computer aided non linear analysis of masonry arch bridges. The analysis takes into account the material nonlinearity due to non-linear stress-strain relationship and cracking and crushing of the masonry. A limited comparison of 2-dimensional and 3-dimensional finite element analysis of masonry arch bridges in context to the load rating of masonry arch bridges have been carried out to ascertain the suitability and adequacy of 2-dimensional finite element analysis. The 3- dimensional analysis has been carried out using commercially available general-purpose finite element analysis software ANSYS. In order to validate the results of proposed mechanism based frame analysis model and 2-dimensional non-linear finite element model, tests to failure on series of arches were conducted in the laboratory. The test programme included construction and testing of two bare test arches having vault only and two full arches with spandrel walls and overlaying fill. A good agreement between the analytically predicted and the experimental test results has been observed, which suggest that the proposed analytical procedures are adequate for the present tasks. Using these two proposed analytical methods, some of the bridges tested on full scale by Transport Research Laboratory, U.K. and whose results are well documented, were analysed. The results have been compared with few other methods such as CTAP, ARCHIE, MINIPONT, ARCH, and MAFEA available in the literature to present an up to date status. The correlations obtained by the proposed methods gives better estimate of carrying capacity in some of the cases while for others the results are quite similar to those predicted by other analytical methods. On the whole the proposed methods are better than those available. It is revealed from the study that finite element method can provide a better estimate of carrying capacity in comparison to other mechanism-based methods but requires large input data. In addition to the above two analytical procedures, Artificial Neural Network has been just used as a tool for the prediction of axle load on the basis of the data existing in literature, from the results of full-scale tests conducted under the service loads. The numbers of available data sets on masonry arch bridges are very less as are required for the use of ANN. However, it should be remembered that, with only a limited number of test results available, such an exercise can not be regarded as a fully comprehensive evaluation. Hence, the proposed method should not be used independently. The back propagation algorithm has been used in this work. From different trials, a neural network with two hidden layers with fifteen nodes in each layer was found to be best. The network has been trained on eleven input parameters namely, square span of the bridge, skew span of the bridge, rise at the crown, thicknesses ofthe arch ring at crown and springing, skew angle, shape ofthe arch, material ofconstruction and maximum observed crown displacements. The applied test load, which is representative ofthe maximum axle load ofthe arch, has been kept as the output target value for the training purpose. The prediction error from the trained ANN can be further reduced with the availability of more field test data especially for indigenous conditions, used for the training of the network. The trained network has been applied to predict the axle load for the 8 bridges under this study. The comparison of predicted crown load corresponding to 1.25 mm deflection with finite element results is within a reasonable range. The trained network has further been used for prediction of load-deflection behaviour of a masonry arch bridge within the serviceable range of deflections. The trained network can hence, predict the axle load corresponding to an acceptance criteria based on the maximum permissible crown deflections for any given masonry arch bridge. An acceptance criterion for evaluation ofthe safe load for rating of existing masonry arch bridges recommended in Indian Roads Congress SP-9 can be suitably used. VI