In this paper, a feed forward neural network with back propagation artificial neural network (BP ANN) was developed to predict ultimate tensile strength (UTS) and optimize microstructure. The alloys were produced by directional solidification and heat treatment. The UTS was measured for ANN output. Five characteristic factors used for ANN input were abstracted and measured. As the result of this study, the ANN model with high accuracy and good generalization ability to predict UTS within the range of 343.5–1063.3 MPa was established and mutual verified with sensitivity analysis. Based on the optimized ANN model, a new way to design microstructure of Nb-Si alloy to obtain required UTS was proposed. With silicide design maps made by ANN model, the microstructure of the sample of 343.5 MPa was optimized and the UTS reached the target UTS (600 MPa) successfully.

在本文中，开发了具有反向传播人工神经网络（BP ANN）的前馈神经网络，以预测极限抗拉强度（UTS）并优化微观结构。合金是通过定向凝固和热处理生产的。对UTS进行了ANN输出的测量。对用于人工神经网络输入的五个特征因子进行了抽象和测量。这项研究的结果是，建立了具有高准确度和良好的泛化能力的ANN模型，可以在343.5–1063.3 MPa范围内预测UTS，并通过敏感性分析进行了相互验证。在优化的人工神经网络模型的基础上，提出了一种新的设计铌硅合金显微组织以获得所需UTS的方法。借助ANN模型制作的硅化物设计图，可以得出343样品的微观结构。