论文部分内容阅读
针对Cu基非晶合金材料受时间、条件及制备成本的限制,导致获得的热力学性能数据较少,影响新材料的研发问题.提出一种在小样本数据情况下,仍然具有较好泛化能力的Cu基非晶合金热力学性能软测量方法,为新材料配方的优化提供模型参考.通过注入噪声,提出一种小样本数据的扩充方法,增加样本的多样性.考虑样本分布函数的未知,在准则函数中引入信息论的微分熵,建立最大熵准则的神经网络反向传播理论,获得具有较高泛化能力的数学模型.仿真分析表明:该方法可对三元Cu基非晶合金的小样本数据,建立其热稳定性及玻璃形成能力与材料配方之间的非线性关系,模型精度较高.
Due to the limitation of time, condition and preparation cost of Cu-based amorphous alloy materials, the obtained thermodynamic performance data is little, which affects the research and development of new materials. A new method is proposed which can still have good generalization ability in the case of small sample data Based soft-sensing method for the thermodynamic properties of Cu-based amorphous alloys, which provides a model reference for optimizing the formulation of new materials. By injecting noise, a method of extending the data of small samples is proposed to increase the sample diversity. Considering the unknown sample distribution function, The differential entropy of information theory is introduced into the criterion function, and the neural network reverse propagation theory of maximum entropy criterion is established to obtain the mathematical model with higher generalization ability. The simulation analysis shows that this method can measure the small sample of ternary Cu-based amorphous alloy Data to establish the non-linear relationship between the thermal stability and glass forming ability and the material formula, the model accuracy is high.