论文部分内容阅读
利用人工神经网络技术研究了DD6合金表面贫Cr层厚度与真空热处理工艺参数之间的关系。建立了以热处理温度、保温时间、真空充气压力为输入变量的三层BP人工神经网络模型,预报了该合金不同工艺下表面贫Cr层的厚度。获得了DD6合金真空热处理过程中应予以规避的工艺区间。为了获得期望的预测结果,对神经网络模型的数据库、隐含层和神经元个数、运算函数进行了优化。预测结果表明:该模型的多元线性相关系数0.9996,网络预测值与样本值相似度较高。采用该方法能够较为准确的预测DD6合金不同真空热处理工艺下的表面贫Cr层厚度。
The relationship between the thickness of lean Cr layer on the surface of DD6 alloy and the parameters of vacuum heat treatment was studied by artificial neural network. A three-layer BP artificial neural network model with heat treatment temperature, holding time and vacuum inflation pressure as input variables was established to predict the thickness of surface-depleted Cr layer under different conditions. Obtained DD6 alloy vacuum heat treatment process should be circumvented the process of interval. In order to obtain the expected results, the neural network model database, hidden layer and neuron number, arithmetic function are optimized. The prediction results show that the multivariate linear correlation coefficient of the model is 0.9996, and the similarity between the network forecast value and the sample value is high. The method can predict the surface lean Cr layer thickness under different vacuum heat treatment processes of DD6 alloy more accurately.