The chemical etching method determining intergranular corrosion of 30Cr2MoV steam turbine rotor steels is developed,and intergranular corrosion groove widths were obtained. It contained statistical test in Normal distribution,Lognormal distribution,Weibull distribution of groove width. Statistical test method based on Quantile—Quantile Plot and K-S test. Statistical results indicate that intergranular corrosion groove width follow Lognormal distribution and Weibull distribution, not follow Normal distribution. Then Temper embrittlement is researched through calculating mathematical expectation. Analysis on the effects of intergranular corrosion groove width,chemical ingredients,crystal granularity and other factors on fracture appearance transition temperature by means of multiple linear regression and BP neural network . The results indicate that precision of prediction of BP neural network is higher than precision of prediction of multiple linear regression. Levenberg-Marquardt algorithm of BP neural network is much faster than the tradition BP algorithms and improved BP algorithms.
|