Lower-bound Axial Buckling Load Prediction For Thin-walled Cylindrical Shells Using Probabilistic Random Perturbation Load Approach

Thin-walled cylindrical shells are widely used in fields such as aerospace and chemical engineering by virtue of highly efficient load-carrying capacity.However,due to the high sensitivity to various imperfections,the actual axial load-carrying capacity of these structures is much lower than the theoretical value of perfect structures,and the dispersion is large.Therefore,the influence of initial defects should be fully considered in the design stage,and the lower-bound axial buckling load should be accurately predicted to ensure the safe operation of the structure.In this thesis,the Probability Random Perturbation Load Approach?PRPLA?and its improved method were used to predict the lower-bound buckling load of thin-walled cylindrical shells under uniform and partial axial compression.The research results can provide technical support for the design of similar engineering structures.The main works are summarized as follows:?1?Based on Back Propagation Neural Network?BPNN?,a geometric morphology reconstructing method for thin-walled cylindrical shells was established.Take the reconstruction performance indexes(i.e.,RMSE,R2 and w_max)as the evaluation criteria,the abilities to reconstruct the overall tendency of the surface morphology and to reserve crucial geometric imperfection signatures were compared among the proposed method,support vector machine and Fourier series.Meanwhile,buckling simulation studies for thin-walled cylinders under uniform axial compression were carried out,and the differences of load-displacement curves and buckling modes among models containing different BPNN-based reconstructed geometric morphologies were discussed.?2?PRPLA was established to predict the lower-bound buckling load for thin-walled cylinders under uniform axial compression.The defect types were classified systemically.The imperfection modeling method in which a single perturbation load with random position?Random SPLA?was superposed on the BPNN-based reconstructed geometric morphology to represent geometric imperfection while non-traditional imperfections were generated using Monte-Carlo simulation was developed.The ABAQUS secondary development program,including modeling automation and Monte-Carlo simulation,was written.Take the experimental data of three series of shells and the buckling experimental results of two new specimens as benchmarks,the advantages and disadvantages of multiple prediction methods were compared and analyzed.The research showed that PRPLA had high reliability and broad applicability.?3?PRPLA was improved and applied to the lower-bound buckling load prediction for shells under partial axial compression.Preliminary research found that the effects of partial axial compression load characteristics and geometric defect characteristics on the partial load-carrying capacity were randomly and without apparent regularity.PRPLA was not suitable for partial axial compression conditions,so PRPLA was improved as following:?1?The boundary condition was adjusted to partial compression,?2?Random SPLA was changed to the two perturbation load approach and?3?The effect of partial axial load movement was introduced.Therefore,PRPLA-P using for partial axial load conditions was established.The lower-bound buckling load prediction for two specimens with different partial load distribution angles was carried out.The differences between the prediction results and the experimental values were analyzed.The results showed that PRPLA-P had high prediction accuracy.