论文部分内容阅读
由于部分稳定氧化锆具有优良的物理化学性能,在冶金及材料中有着重要的地位,稳定率是部分稳定氧化锆产品性能的一个重要指标。而部分稳定氧化锆的制备过程具有非线性、多变量、时变等特点,本文采用了支持向量机(SVM)及BP神经网络方法对部分稳定氧化锆的稳定率进行了预测。将热处理温度、保温时间、降温速率、淬火温度及升温速率5个指标(参数)作为模型输入量,部分稳定氧化锆的稳定率作为输出值,分别以48组实验数据作为学习样本,并建立模型,运用该模型预测了5组部分稳定氧化锆的稳定率。实验结果表明,2种模型均具有较好的预测能力,人工神经网络模型预测结果平均误差为1.48%,支持向量机模型预测结果平均误差为0.68%,并且支持向量机预测部分稳定氧化锆的稳定率精度更高,可在实际生产过程中推广应用。
Partially stabilized zirconia has an important physicochemical property and plays an important role in metallurgy and materials. Stability is an important indicator of the stability of partially stabilized zirconia. However, the preparation of partially stabilized zirconia is nonlinear, multivariable, time-varying and so on. In this paper, support vector machine (SVM) and BP neural network are used to predict the stability of partially stabilized zirconia. Five parameters (parameters), such as heat treatment temperature, holding time, cooling rate, quenching temperature and heating rate, were taken as model inputs and the stability of partially stabilized zirconia was taken as output value. 48 groups of experimental data were used as learning samples respectively, and model The model was used to predict the stability of five groups of partially stabilized zirconia. The experimental results show that the two models have good predictive ability, the average error of artificial neural network model prediction results is 1.48%, the average error of support vector machine model prediction results is 0.68%, and the support vector machine predicts the stability of some stabilized zirconia The rate of accuracy is higher, can be popularized and applied in the actual production process.