基于GRNN神经网络优选高温合金铸件晶粒细化剂

Optimization of Grain Refiner for Superalloy Casting Based on GRNN

对镍基高温合金铸件研制出四组元化学细化剂A-B-C-D,针对细化剂的优选问题,采用GRNN神经网络模拟细化剂各组元含量与铸件晶粒尺寸间的非线性关系。研究发现:加入A-B-C-D新型复合细化剂可以明显细化铸件晶粒;细化剂的最佳加入量为0.11wt%A、0.23wt%B、0.14wt%C、0.17wt%D,即组元A-B-C-D的最佳质量配比约为1∶2.1∶1.3∶1.5。

A new chemical refiner composed of A-B-C-D for Ni-based superalloy casting was developed. According to the optimization of refiner, the generalized regression neural network （GRNN） was adopted to simulate the nonlinear connection between each component content of refiner and casting grain size. The results show that the grain size of ingots can be termed obviously by adding new composite refiner A-B-C-D; the combination of 0.1 l wt%A, 0.23wt%B, 0.14wt%C and 0.17wt%D is best as composite refiner the optimum proportion of component A-B-C-D is about 1 ： 2.1 ： 1.3 ： 1.5.