中文摘要：

考虑了液态金属冷却定向凝固过程中动态对流边界,建立了高温合金铸件温度场数学模型,采用三维元胞自动机（cellular automaton,CA）方法和KGT生长模型,建立了镍基高温合金凝固过程晶粒形核及生长的数学模型.采用宏观模型与微观模型双向同步耦合,实现了温度场和晶粒组织的数值模拟.进行了浇注实验,用冷却曲线和晶粒形貌验证了数学模型的准确性.对液态金属冷却定向凝固规律的研究表明,抽拉速率不仅对糊状区形状有重要影响,而且对晶粒的平行度以及枝晶组织的细密性也有很大的影响.抽拉速率过小时,糊状区上凸,晶粒组织易发散;抽拉速率过大时,糊状区下凹,晶粒组织汇聚,同时造成枝晶组织的粗化;适当的抽拉速率下能获得平坦的糊状区,提高晶粒的平行度,细化枝晶组织.

英文摘要：

Gas turbine plays an important role in energy and aviation, among which the turbine blades are the key components. Ni base superalloys are the preferred material to manufacture blades due to their high temperature strength, microstructural stability and corrosion resistance. As a conventional directional solidification method,high-rate solidification（HRS） is used to produce columnar grain and single crystal blades. However, there are several problems when HRS is scaled to cast industrial gas turbines（IGT） components. In recent years, several possible techniques are being proposed for large IGT blades. The liquid-metal cooling（LMC） is one of the best methods among them, which improves heat extraction by immersing the casting and the mold into a container of metal coolant with low melting temperature as they are withdrawn from the heating zone. Unfortunately, the trial and error method is time and money cost and lead to a long RD cycle. Therefore, numerical simulation plays an important role to optimize the process, and enhance the productivity in LMC directional solidification. In this work, mathematical models for dynamic heat radiation and convection boundary of LMC process are established to simulate the temperature fields. Cellular automaton（CA） method and KGT growth model are used to describe the nucleation and growth. The pouring experiments are carried out. The accuracy of the model is validated by the cooling curves and microstructure. Moreover, the liquid-metal cooling directional solidification process is discussed in more detail, including primary dendrite arm space（PDAS）, secondary dendrite arm space（SDAS）, mushy zone and microstructure, etc.. Simulation and experiment results are compared in the work. This study indicates that simulation and experimental results agree with each other well. The maximum error of temperature is less than 5 percent and the morphologies of grains are similar. The withdrawal rate has an important influence on the shape of mushy zone and dendritic str