[1]董宝阳,解 晖,许兆宋,等.含输入噪声的多类别高斯过程液体火箭发动机故障诊断[J].南京师大学报(自然科学版),2025,48(05):75-84.[doi:10.3969/j.issn.1001-4616.2025.05.009]
　Dong Baoyang,Xie Hui,Xu Zhaosong,et al.Fault Diagnosis of Multi Class Gaussian Process Liquid Rocket Engine with Input Noise[J].Journal of Nanjing Normal University(Natural Science Edition),2025,48(05):75-84.[doi:10.3969/j.issn.1001-4616.2025.05.009]
点击复制

含输入噪声的多类别高斯过程液体火箭发动机故障诊断()

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
48
期数:
2025年05期
页码:
75-84
栏目:
计算机科学与技术
出版日期:
2025-10-20

文章信息/Info

Title:
Fault Diagnosis of Multi Class Gaussian Process Liquid Rocket Engine with Input Noise
文章编号:
1001-4616(2025)05-0075-10
作者:
董宝阳1; 解 晖2; 许兆宋2; 刘久富2
(1.郑州理工职业学院信息工程学院,河南 郑州 451100)
(2.南京航空航天大学自动化学院,江苏 南京 211106)
Author(s):
Dong Baoyang1; Xie Hui2; Xu Zhaosong2; Liu Jiufu2
(1.School of Information Engineering, Zhengzhou Institute of Technology, Zhengzhou 451100, China)
(2.School of Automation, Nanjing University of Aeronautics and Astronautics, Nanjing 210096, China)
关键词:
多类别高斯过程; 变分推断; 液体火箭发动机; 输入噪声
Keywords:
multi class Gaussian process; variational inference; liquid rocket engine; input noise
分类号:
TP181
DOI:
10.3969/j.issn.1001-4616.2025.05.009
文献标志码:
A
摘要:
针对传统多类别高斯过程分类算法往往忽略数据受到的噪声污染导致预测准确性降低的问题,提出一种基于变分推断优化算法的含输入噪声的多类别高斯过程分类算法. 以多类别高斯过程模型作为底层分类器,在传统模型上引入加性高斯噪声,使用变分推断方法优化改进后的模型,近似模型隐变量的后验分布,并据此进行新的预测. 将含输入噪声的多类别高斯过程分类方法应用到液体火箭发动机的故障分类问题中,实验证明,与传统多类别高斯过程分类算法相比,提出的算法在预测精度上有一定提高,负似然对数指标有效降低,改进后的模型与真实后验分布更接近.
Abstract:
A multi class Gaussian process classification algorithm with input noise based on variational inference optimization algorithm is proposed to address the problem that traditional multi class Gaussian process classification algorithms often ignore the noise pollution caused by data and reduce prediction accuracy. The multi class Gaussian process model is used as the underlying classifier, additive Gaussian noise is introduced into the traditional model, and the variational inference method is used to optimize the improved model, approximate the posterior distribution of the model's potential variables, and make new predictions accordingly. The multi class Gaussian process classification method containing input noise is applied to the fault classification problem of liquid rocket engines. Experiments have shown that compared with the traditional multi class Gaussian process classification algorithm, the proposed algorithm is better at predicting. There is a certain improvement in accuracy, the negative logarithmic likelihood index is effectively reduced, and the improved model is closer to the true posterior distribution.

参考文献/References:

[1]GRANDINI M,BAGLI E,VISANI G. Metrics for multi-class classification:an overview[J]. arXiv.2020. 2008.05756.
[2]BISHOP C M,NASRABADI N M. Pattern recognition and machine learning[M]. New York:Springer,2006.
[3]何志昆,刘光斌,赵曦晶,等. 高斯过程回归方法综述[J]. 控制与决策,2013,28(8):1121-1129,1137.
[4]SCHULZ E,SPEEKENBRINK M,KRAUSE A. A tutorial on Gaussian process regression:modelling,exploring,and exploiting functions[J]. Journal of mathematical psychology,2018,85:1-16.
[5]DERINGER V L,BARTóK A P,BERNSTEIN N,et al. Gaussian process regression for materials and molecules[J]. Chemical reviews,2021,121(16):10073-10141.
[6]KUSS M,RASMUSSEN C E,HERBRICH R. Assessing approximate inference for binary gaussian process classification[J]. Journal of machine learning research,2005,6(10):1679-1704.
[7]张炳晖,张研,王伟,等. 基于高斯过程二元分类模型的溶洞规模预测方法[J]. 中国岩溶,2020,39(02):259-263.
[8]黄维,童莹,曹雪虹. 基于高斯过程的多分类器[J]. 计算机工程与设计,2016,37(04):1021-1026.
[9]MCHUTCHON A,RASMUSSEN C. Gaussian process training with input noise[C]//Proceedings of the 25th International Conference on Neural Information Proceeding Systems,Granada Spain,Curran Associates Inc,2011:1341-1349.
[10]XUE Y,LIU Y,JI C,et al. System identification of ship dynamic model based on Gaussian process regression with input noise[J]. Ocean engineering,2020,216:107862.
[11]VILLACAMPA-CALVO C,ZALDÍVAR B,GARRIDO-MERCHÁN E C,et al. Multi-class gaussian process classification with noisy inputs[J]. The journal of machine learning research,2021,22(1):1696-1747.
[12]SNELSON E,GHAHRAMANI Z. Sparse Gaussian processes using pseudo-inputs[C]//Proceedings of the 19th International Conference on Neural Information Proceeding Systems. Vancourer British Columbia Canada,MIT Press,2005:1257-1264.
[13]TITSIAS M,LAWRENCE N D. Bayesian Gaussian process latent variable model[C]//Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics. Sardinia Italy,PMLR,2010:844-851.
[14]BLEI D M,KUCUKELBIR A,MCAULIFFE J D. Variational inference:A review for statisticians[J]. Journal of the American statistical association,2017,112(518):859-877.
[15]ZHANG C,BÜTEPAGE J,KJELLSTRÖM H,et al. Advances in variational inference[J]. IEEE transactions on pattern analysis and machine intelligence,2018,41(8):2008-2026.
[16]TITSIAS M. Variational learning of inducing variables in sparse Gaussian processes[C]//The 12th International Conference on Artificial intelligence and statistics. Florida USA,PMLR,2009:567-574.
[17]LEIBFRIED F,DUTORDOIR V,JOHN S T,et al. A tutorial on sparse Gaussian processes and variational inference[J]. arXiv,2020,2012.13962.
[18]WYNNE G,WILD V. Variational gaussian processes:A functional analysis view[C]//Proceedings of the 25th International Conference on Artificial Intelligence and Statistics. Virtual Conference,PMLR,2022:4955-4971.
[19]DAMIANOU A,LAWRENCE N D. Semi-described and semi-supervised learning with Gaussian processes[J]. arXiv,2015:1509.01168.
[20]ZHANG Y,LIU W,CHEN Z,et al. On the properties of kullback-leibler divergence between multivariate Gaussian distributions[J]. arXiv,2021,2102.05485.
[21]Trefethen L N. Exactness of quadrature formulas[J]. Siam review,2022,64(1):132-150.

备注/Memo

备注/Memo:
收稿日期:2025-01-06.
基金项目:国家自然科学基金资助项目(61473144)
通讯作者:刘久富,博士,研究方向:故障诊断、深度学习. E-mail:liujiufu4@126.com
更新日期/Last Update: 2025-10-20