Analytical Models for Calculating the Response of Temporary Soil-Filled Walls Subjected to Blast Loading

The aims of the thesis were to study the response of temporary soil-filled walls both experimentally and numerically, and to develop an efficient and accurate analytical model to predict 2-D planar response from blast loading which could be used to efficiently calculate a pressure-impulse (P--I) curve. An explicit finite element (FE) formulation was constructed using LS-Dyna software, and two analytical models were also derived and presented: a Rigid-Body Rotation model as a preliminary model, and the Rigid-Body Hybrid model as the proposed model of this thesis. Seven full-scale experiments which consisted of blast loading simple free-standing soil-filled Hesco Bastion (HB) walls are presented. Soil densities and moisture contents were measured in the field, and soil properties were obtained from triaxial tests of the samples collected and prepared to match field conditions. All models used some or all of the derived soil properties pertinent to the experiments as input, and whenever possible, the recorded pressure--time histories in the experiments were assumed as the loadings. The results of both analytical models and the FE models were compared with experimental results. In addition, the results were compared with one another within parametric studies concerning sensitivity of model responses to soil properties and different height-to-width ratio walls. The models were also used to calculate and compare P--I curves.;The FE models were found to be in excellent agreement in both the post-experiment deformation and the displacement--time histories for the seven experiments (most results within 5%). The Rigid-Body Rotation model was found to be in reasonable agreement with the post-experimental deformation in cases where the wall did not critically overturn but sustained moderate rotations. However, comparison with the experimentally derived displacement--time histories showed that it underestimated displacement--time histories and thus it possessed too much resistance. Apart from comparison of an experimental result where the soil-fill in the wall possessed sizable cohesion, the response of the Rigid-Body Hybrid model was in very good agreement with the experiments overall (within 10%). A general recommendation for the model development follows that a sliding interface should be included in the model to capture sliding behaviour arising from use of soil-fills with substantial cohesion. A soil sensitivity study was conducted and overall very good agreement was reached between the Rigid-Body Hybrid model in comparison with the FE model in its ability to capture differences in displacement--time histories from differences in soil parameters. P--I curves were developed using the analytical and FE models for the three different wall configurations studied in the experiments. The results demonstrated that the proposed Rigid-Body Hybrid model is useful for calculating a P--I curve for a HB wall efficiently and yielded very accurate results (within 5% for the impulse asymptotes). To establish the limitations of both analytical models, an aspect ratio study was conducted where the rotation of the analytical models were compared to that of the FE models for walls of different height-to-empty width ratios, across a range of impulsive loadings. Comparison with the FE model for different height-to-width ratios of walls showed that the Rigid-Body Hybrid model was within 10% for all rotation angles and predictions of critical overturning impulse for height-to-width ratios of walls H/wa ≥ 1.43. For walls with H/wa =1.29 the Rigid-Body Hybrid model was only in agreement for rotation angles from 18 to 21 degrees. In view of this narrow range of accuracy, use of the model for walls H/wa ≤ 1.29 in its existing form is not recommended. Consideration of sidewall folding and contact with the ground is proposed to improve its accuracy for walls H/w a ≤ 1.29. Apart from this, overall the Rigid-Body Hybrid model is on average within 10 % experimental results and FE model results. Recommendations are provided to address minor deficiencies within the model and to expand its range of application.