During the last couple of decades, several design guidelines and codes relating to major nuclear components have been introduced and revised for risk-based design and evaluation. It is an important process to determine statistical parameters in practice, for instance, by using the traditional method of moments and Maximum Likelihood Method (MLM). Since appropriate estimation of the parameters is not easy due to mathematical complexity, a robust method adopting concepts of both Chi-Square test and genetic algorithm (GA) is proposed in this paper. The Chi-Square test is a useful technique to get the goodness-of-fit of distributions, which is represented in terms of error between observed frequencies and frequencies calculated by assumed probability density function (PDF) of certain statistical distribution. The GA is an efficient optimization algorithm to solve nonlinear optimum problems. Using the Chi-Square test, statistical parameters can be determined and transferred to an optimum problem, and then solved by the GA employing proper nonlinear objective function. Reliability of the proposed method is verified against fracture toughness test data sets of SA508 reactor pressure vessel material obtained from PCVN specimens at various temperatures. The large scatter of experimental data is examined in use of a distribution reported by Neville and Kennedy, Burr type III and XII distributions by Nadarajah and Kortz as well as well-known Weibull distribution. A systematic assessment is carried out by using the new method and its results are compared with corresponding ones derived from the traditional method. Pros and corns of the alternative distributions as well as technical findings from the statistical assessment are fully discussed to show applicability of them.

This content is only available via PDF.
You do not currently have access to this content.