利用人工神经网络研究了不同化学成分NiAl合金超塑性变形的条件.建立了以化学成分、应变速率和温度等为输入参数的标准多层负反馈神经网络,利用挤压态NiAl及NiAl系合金数据库对网络进行了训练和测试. 研究了NiAl及添加P,Mo, Fe, Y, Ce, Nb, Cr和Hf元素的NiAl系合金在超塑性拉伸实验中,合金元素对超塑性延伸率的影响以及变形温度、应变速率等对延伸率的影响. 为了获得期望的结果,神经网络模型对输入参量、数据库和隐含层神经元个数进行了优化. 测试结果表明,神经网络的多元相关系数为0.93.利用模型预测并得出优化后的NiAl及NiAl系合金化学成分以及可以获得的最大延伸率以及最佳实验条件范围.

Chemical composition, grain size, and processing conditions such as temperature and strain rate have important influence on superplasticity of NiAl alloys, which would allow the optimization of these parameters in order to achieve the desired combination of properties. In this work, the optimal superplastic deformation conditions of NiAl alloys were studied by using artificial neural networks (ANN). The standard multilayer feedforward networks were trained and tested using comprehensive datasets from previous experimentally works on the as-extruded NiAl, NiAl-25Cr, NiAl-20Fe-Y(Ce), NiAl-30Fe-Y, NiAl-9Mo, NiAl-P alloys.Different effects are modeled, including the influence of the alloying elements on the superplastic, and the influence of deformation temperature, strain rate and grain size on the elongations during the superplastic tensile tests. The artificial neural network models are combined with computer programmers for optimization of the inputs in order to achieve desirable combination of outputs. Good performances of the neural networks  are achieved. Results of this research propose a range of strain rate and temperature within which the NiAl alloy possesses superplasticity with larger elongations, although the deformation temperature and strain rate of superplastic alloys alternately influence each other within the range. These models are convenient and powerful tools for practical applications in superplastic prediction in NiAl alloys.