基于人工神经网络方法对Nb-Si基超高温合金微观组织与力学性能关系建模的研究

Nb-Si based alloys show great promise for application as the next generation turbine airfoil materials at the temperature of up to 1250℃. However, a major barrier to the development of the Nb-Si based alloys is that the relationships among the composition, processing, structure and properties have not been modeled quantitatively. So, Nb-Si based alloys are investigated traditionally by trial and error, which retards the application of the Nb-Si based alloys. It is difficult to model these relationships precisely using traditional modeling methods , because Nb-Si based alloys possess complicated in-situ composite microstructure and contain various phases. As one of the data-driven soft computational tools, Artificial Neural Network (ANN) has been found to be effective for the purpose of prediction, optimization or even exploring the hidden knowledge within the complex and highly nonlinear materials system. In this study, the database of microstructures, interface effects and mechanical properties of Nb-Si based alloys will be set up on the basis of nanoindentation and finite element method. Then, the models between the microstructure and mechanical properties of Nb-Si based alloys will be built up based the database by means of ANN. These models can predict properties of Nb-Si based alloys, guide the optimization of alloy compostion and microstructure. With the help of these ANN models, the various properties of Nb-Si based alloys can be balanced, which will finally promote the development and application of Nb-Si based alloys.

Nb-Si基超高温合金作为最具潜力的可在1250℃以上使用的航空发动机涡轮叶片用材料之一，由于其“成分-工艺-组织-性能”之间的定量关系尚不明确，传统的“炒菜”式研究方法严重迟滞了材料发展，至今尚未成功应用。Nb-Si基合金具有自生复合的多相组织特征，难以使用传统建模方法建立“微观组织-力学性能”模型。人工神经网络方法能够有效地对多元多相合金这样的相互作用复杂、高度非线性的系统做出预测和优化。鉴于此，本项目拟采用有限元、纳米压痕等模拟表征技术，对Nb-Si基合金的微观组织、界面作用和力学性能进行定量化表征，建立微观组织特征与力学性能数据库。在此基础上使用人工神经网络方法，建立Nb-Si基合金“微观组织-力学性能”定量模型，并使用该模型进行合金性能预测、指导合金成分设计和组织设计，实现Nb-Si基超高温合金室温韧性、高温强度和高温持久性能综合匹配，推动Nb-Si基超高温合金的发展与应用。

随着航空航天技术的发展，对航空发动机的要求越来越高。高推比航空涡轮发动机需要更高承温能力的材料，因此研制新型超高温合金迫在眉睫。Nb-Si基超高温合金由于熔点高、密度低、高温性能优良等特点受到了广泛的关注，但是其综合性能尤其是强韧化匹配性不足，严重迟滞了合金的应用。本项目通过研究微观组织对Nb-Si基超高温合金室温断裂韧性、塑性、高温强度和高温抗氧化性等性能的影响规律，归纳出影响力学性能的相关组织特征，并对相关组织进行定量化、数字化，采用有限元和纳米压痕等模拟表征技术，对Nb-Si基合金的显微组织、力学性能和界面作用进行定量表征，建立成分-组织-性能的数据库。研究和表征Nb-Si基合金的相界面性能，包括剪切强度、法向分离强度、断裂韧性等，提出能够反映固溶体和硅化物的界面作用表征参数，明确两相界面在载荷传递中的作用。研究了神经网络模型中隐藏层个数、隐藏层神经元个数、输入输出数量、变量传递方式等参数对模型准确度的影响顾虑，构建了多参量微观组织-多参量力学性能的人工神经网络模型。研究了显微组织特征对某种力学性能指标的影响规律，确定不同微观组织参量对力学性能指标的影响权重。