全文下载:
202603017.pdf
文章编号:1672-6987(2026)03-0139-09;DOI:10. 16351/j. 1672-6987. 2026. 03. 017
郭金玉,张俊杰,李 元* (沈阳化工大学 信息工程学院, 辽宁 沈阳 110142)
摘 要:传统非线性故障检测方法在使用核技巧时会出现计算量大、选择参数困难等问题,为 了更好的进行非线性过程的故障检测,提出一种新颖的基于 LPP 子空间和贝叶斯推理的非线 性故障监测方法。通过局部保持投影(LPP)方法将原始非线性空间近似为多个线性子空间, 为每个子空间中的变量定义一个新的子空间贡献指标。首先在每个子空间中生成故障检测 结果,然后通过贝叶斯推理策略转换为故障概率。将子空间检测结果与其故障概率相结合做 出最终的故障检测决策。数值示例和田纳西-伊斯曼(TE)基准过程的仿真结果都显示了所 提出方法具有低误报率和高故障检测率的特点。
关键词:故障检测;非线性过程;线性子空间;局部保持投影;贝叶斯推理
中图分类号:TP 277 文献标志码:A
引用格式:郭金玉,张俊杰,李元 . 基于 LPP 子空间和贝叶斯推理的非线性故障检测[J]. 青 岛科技大学学报(自然科学版),2026,47(3):139-147.
GUO Jinyu, ZHANG Junjie, LI Yuan. Nonlinear fault detection based on LPP subspace and Bayesian inference[J]. Journal of Qingdao University of Science and Technology(Natural Sci⁃ ence Edition),2026,47(3):139-147.
Nonlinear Fault Detection Based on LPP Subspace and Bayesian Inference
GUO Jinyu,ZHANG Junjie,LI Yuan (College of Information Engineering, Shenyang University of Chemical Technology, Shenyang 110142, China)
Abstract:Traditional nonlinear fault detection algorithms encounter problems such as high computational complexity and difficulty in selecting parameters when using kernel techniques. To better detect faults in nonlinear processes, a novel nonlinear fault detection method based on LPP subspace and Bayesian inference is proposed. The original nonlinear space is approxi⁃ mated into multiple linear subspaces through the locality preserving projections (LPP) method. A new subspace contribution index is defined for the variables in each subspace. In each sub⁃ space the fault detection results are first generated and then converted into fault probabilities via a Bayesian inference strategy. The subspace detection results are combined with their fault probabilities to make final fault detection decisions. Simulation results of a numerical example and the Tennessee-Eastman (TE) benchmark process both show that the proposed method has low false alarm rate and high fault detection rate.
Key words:fault detection; nonlinear processes; linear subspace; locality preserving projec⁃ tions; Bayesian inference
收稿日期:2025-05-22
基金项目:国家自然科学基金项目(62273242);辽宁省教育厅科学研究项目(JYTMS20231516).
作者简介:郭金玉(1975—),女,副教授 . * 通信联系人 .