| Description |
viii, 90 leaves : illustrations (some colored) ; 29 cm |
| Summary |
"Described herein is the development of an efficient capability to analyze the statistics of structural responses given a description of the statistics of the input, or design, variables. In the application of interest, inputs and outputs are related through a very large finite element model. The finite element model is approximated here by a second-order hyper-surface. Developed for the quadratic relationship are expressions for response means, standard deviations and correlation coefficients. The analysis allows correlated and uncorrelated input variables, and admissible input variable distributions are normal, log-normal and Weibull. In addition to characterizations of response means, standard deviations and correlation coefficients, response cumulative distribution functions can also be produced for the quadratic input-output relationships, since an existing software interface was rewritten to permit the use of a quality statistical analysis tool for the purpose. Included in the thesis are several examples to illustrate the accuracy of the new analysis relative to the accuracy of a linear approach. Demonstrated is the superiority of the new analysis for the estimation of response means and of cumulative distribution functions"--Abstract, leaf iii. |
|
