基于反向传播人工神经网络对SiC氧化反应行为的预测研究

硅酸盐通报 ›› 2021, Vol. 40 ›› Issue (10): 3213-3218.

基于反向传播人工神经网络对SiC氧化反应行为的预测研究

赵春阳, 王恩会, 方志, 郭春雨, 段兴骏, 侯新梅

北京科技大学钢铁共性技术协同创新中心,北京 100083

收稿日期:2021-08-15 修回日期:2021-09-15 出版日期:2021-10-15 发布日期:2021-11-11
通讯作者: 王恩会,博士,副研究员。E-mail:wangenhui@ustb.edu.cn
作者简介:赵春阳(1997—),男,硕士研究生。主要从事人工神经网络在耐火材料设计中的应用研究。E-mail:g20199295@xs.ustb.edu.cn
基金资助:
国家自然科学基金(51904021,51974021)

Prediction of SiC Oxidation Reaction Behavior Based on Back Propagation Artificial Neural Network

ZHAO Chunyang, WANG Enhui, FANG Zhi, GUO Chunyu, DUAN Xingjun, HOU Xinmei

Collaborative Innovation Center of Steel Technology, University of Science and Technology Beijing, Beijing 100083, China

Received:2021-08-15 Revised:2021-09-15 Online:2021-10-15 Published:2021-11-11

摘要/Abstract

摘要： 以SiC为代表的非氧化物耐火原料作为高温结构材料重要组分,被广泛应用于冶金高温行业。在实际应用过程中,SiC的氧化行为加速了对应耐火材料的高温性能失效,导致其服役寿命大大缩短。因此明晰非氧化物耐火原料在高温环境下的氧化行为尤为重要,利用动力学模型分析氧化行为是目前最常用的手段。但动力学模型的建立往往需要大量的数据处理工作,且很难同时满足描述准确性高和模型参数简单两个条件。随着人工智能与大数据技术在材料领域的应用探索,反向传播人工神经网络(BP-ANN)有望在此方面取得突破。本文以典型非氧化物耐火原料SiC为例,通过建立神经网络,训练、预测SiC的氧化行为,预测结果与实验数据的相对误差均小于3%,用预测数据回归计算的反应活化能和反应速率常数与实验数据计算结果的相对误差低于4%,表明BP-ANN在研究非氧化物耐火原料的氧化行为方面具有巨大应用前景。

关键词: 耐火材料, 高温结构材料, 非氧化物材料, 氧化, SiC, 反向传播人工神经网络, 模型优化, 反应活化能

Abstract: Non-oxide refractory raw materials represented by SiC are widely used in the metallurgical high-temperature industry as important components of high-temperature structural materials. In practice, the oxidation behavior of SiC accelerates the failure of corresponding refractories at high temperatures, resulting in a significant reduction of their service life. Therefore, it is important to understand the oxidation behavior of non-oxide refractory raw materials at high temperature. Kinetic models are the most common means to analyze the oxidation behavior. However, the establishment of kinetic models often requires a lot of data processing work, and it is difficult to meet the two conditions of high descriptive accuracy and simple model parameters at the same time. With the exploration of artificial intelligence and big data technology in the field of materials, back propagation artificial neural network (BP-ANN) is expected to make a breakthrough in this area. In this paper, a typical non-oxide refractory raw material SiC was taken as an example, and the oxidation behavior of SiC was trained and predicted by building a neural network. The relative errors of the predicted results and experimental data are less than 3%, and the relative errors of the reaction activation energy and reaction rate constant calculated by regression with the predicted data are less than 4% with the results calculated by the experimental data. This shows that BP-ANN has great potential in studying the oxidation behavior of non-oxide refractory raw materials.

Key words: refractory, high-temperature structural material, non-oxide material, oxidation, SiC, back propagation artificial neural network, model optimization, reaction activation energy

中图分类号:

TQ175

赵春阳, 王恩会, 方志, 郭春雨, 段兴骏, 侯新梅. 基于反向传播人工神经网络对SiC氧化反应行为的预测研究[J]. 硅酸盐通报, 2021, 40(10): 3213-3218.

ZHAO Chunyang, WANG Enhui, FANG Zhi, GUO Chunyu, DUAN Xingjun, HOU Xinmei. Prediction of SiC Oxidation Reaction Behavior Based on Back Propagation Artificial Neural Network[J]. BULLETIN OF THE CHINESE CERAMIC SOCIETY, 2021, 40(10): 3213-3218.

参考文献

[1] 钟香崇,傅征蔚.冶金工业耐火材料技术发展趋向[J].钢铁,1964(8):1-6.
ZHONG X C, FU Z W. Trend of refractory technology development in metallurgical industry[J]. Gangtie, 1964(8): 1-6 (in Chinese).
[2] FERBER M K, LIN H T, KEISER J. Oxidation behavior of non-oxide ceramics in a high-pressure, high-temperature steam environment[M]//Mechanical, Thermal and Environmental Testing and Performance of Ceramic Composites and Components. ASTM International, 2020.
[3] WATARI K. High thermal conductivity non-oxide ceramics[J]. Journal of the Ceramic Society of Japan, 2001, 109(1265): S7-S16.
[4] LUTHRA K L. Oxidation-resistant fiber coatings for non-oxide ceramic composites[J]. Journal of the American Ceramic Society, 1997, 80(12): 3253-3257.
[5] KLEMM H, HERRMANN M, SCHUBERT C. High temperature oxidation and corrosion of silicon-based non-oxide ceramics[J]. Journal of Engineering for Gas Turbines and Power, 2000, 122(1): 13-18.
[6] OPILA E J, JACOBSON N S. SiO(g) formation from SiC in mixed oxidizing-reducing gases[J]. Oxidation of Metals, 1995, 44(5/6): 527-544.
[7] JANDER W. Reactions in solid states at room temperature: I announcement the rate of reaction in endothermic conversions[J]. Zeitschrift fur Anorganische und Allgemeine Chemie, 1927, 163(1/2): 1-30.
[8] CARTER R E. Kinetic model for solid-state reactions[J]. The Journal of Chemical Physics, 1961, 34(6): 2010-2015.
[9] DEAL B E, GROVE A S. General relationship for the thermal oxidation of silicon[J]. Journal of Applied Physics, 1965, 36(12): 3770-3778.
[10] CHOU K C. A kinetic model for oxidation of Si-Al-O-N materials[J]. Journal of the American Ceramic Society, 2006, 89(5): 1568-1576.
[11] ZHAO Y N, CHEN Y F, AI S G, et al. A diffusion, oxidation reaction and large viscoelastic deformation coupled model with applications to SiC fiber oxidation[J]. International Journal of Plasticity, 2019, 118: 173-189.
[12] WANG E H, HOU X M, CHEN Y F, et al. Progress in cognition of gas-solid interface reaction for non-oxide ceramics at high temperature[J]. Critical Reviews in Solid State and Materials Sciences, 2021, 46(3): 218-250.
[13] ESKANDARI-NADDAF H, KAZEMI R. ANN prediction of cement mortar compressive strength, influence of cement strength class[J]. Construction and Building Materials, 2017, 138: 1-11.
[14] LUCON P A, DONOVAN R P. An artificial neural network approach to multiphase continua constitutive modeling[J]. Composites Part B: Engineering, 2007, 38(7/8): 817-823.
[15] GUO D, LI L T, NAN C W, et al. Modeling and analysis of the electrical properties of PZT through neural networks[J]. Journal of the European Ceramic Society, 2003, 23(12): 2177-2181.
[16] CAI K, XIA J T, LI L T, et al. Analysis of the electrical properties of PZT by a BP artificial neural network[J]. Computational Materials Science, 2005, 34(2): 166-172.
[17] QUAN G Z, PAN J, WANG X. Prediction of the hot compressive deformation behavior for superalloy nimonic 80A by BP-ANN model[J]. Applied Sciences, 2016, 6(3): 66.
[18] WANG E H, HOU X M, CHEN J H, et al. Comparison of the reaction behavior of hexagonal silicon carbide powder in different atmospheres[J]. Metallurgical and Materials Transactions A, 2017, 48(10): 5122-5131.
[19] WANG S C. Artificial neural network[M]//Interdisciplinary Computing in Java Programming. Boston, MA: Springer US, 2003: 81-100.
[20] MORÉ J J. The Levenberg-Marquardt algorithm: implementation and theory[M]//Numerical Analysis. Springer Berlin Heidelberg, 1978: 105-116.