基于机器学习的磷石膏多孔陶瓷材料性能预测  

Prediction of Phosphogypsum Porous Ceramic Materials Based on Machine Learning

作　　者：杨凯理 龚伟[1] 蔺万鹏 YANG Kaili;GONG Wei;LIN Wanpeng(School of Manufacturing Science and Engineering,Southwest University of Science and Technology,Mian yang 621000,China;School of Materials and Chemistry,Southwest University of Science and Technology,Mian yang 621000,China)

机构地区：[1]西南科技大学制造科学与工程学院,绵阳621000 [2]西南科技大学材料与化学学院,绵阳621000

出　　处：《中国陶瓷》2023年第8期23-29,共7页China Ceramics

基　　金：四川省国际科技合作项目(2021YFH0089)。

摘　　要：为探究磷石膏多孔陶瓷材料成分-制备工艺-性能之间定量关系进行研究。以磷石膏为发泡剂制备多孔陶瓷的实验数据,建立4种机器学习算法模型,对磷石膏多孔陶瓷材料数据集中的吸水率、体积密度、抗压强度、导热系数等4种性能参数进行预测,比较各种学习方法的预测结果,并对实验阶段确定的最佳配比进行验证预测。实验表明:SVR-RBF算法可以对抗压强度、导热系数性能进行有效预测,验证预测误差分别为5.402%和0.725%;LMBP算法可以对吸水率、抗压强度性能进行有效预测,验证预测误差分别为0.29%和2.964%。In order to explore the quantitative relationship between the composition,preparation process and performance of phosphogypsum porous ceramic materials,the quantitative relationship was studied.Experimental data on the preparation of porous ceramics using phosphogypsum as the blowing agent,four machine learning algorithm models were established,and four performance parameters such as water absorption,bulk density,compressive strength,and thermal conductivity in the dataset of phosphorus gypsum porous ceramic materials were predicted,the performance of various learning methods was compared,and the optimal ratio determined in the experimental stage was verified and predicted.Experiments show that the SVR-RBF algorithm can effectively predict the compressive strength and thermal conductivity performance,and the prediction error is 5.402%and 0.725%respectively.The LMBP algorithm can effectively predict the water absorption rate and compressive strength performance,and verify that the prediction error is 0.29%and 2.964%,respectively.

关 键 词：机器学习 性能预测 多孔陶瓷 支持向量机 神经网络 

分 类 号：TQ174.75[化学工程—陶瓷工业]