摘要
针对传统HMM方法故障检测的准确率不高,以及带钢热连轧过程数据的非线性和混合高斯性问题,提出一种利用WT和PCA改进HMM的故障检测新方法。首先,采用小波变换对轧制数据进行去噪处理,并使用PCA将数据的维度降低、数据相关性减小,可以有效减少模型训练的迭代次数,并且能够提升故障检测的准确率;然后,利用期望最大化算法结合观测序列训练得到WT-PCA-HMM故障检测模型;最后,通过模型得出精轧工艺数据的对数似然值即可实现故障检测。结果表明:与传统HMM方法相比,WT-PCA-HMM的故障检测方法不仅能够降低8.1%的误报率,而且减少50%的模型训练迭代次数,为故障的检测提供了新方法。
A new method of fault detection was proposed to improve the HMM based on WT and PCA to solve the problems of low accuracy of the traditional HMM method and the nonlinearity and mixed Gaussianity of the hot strip rolling process data.Firstly,wavelet transform was used to denoise the rolling data,and PCA was used to reduce the dimensionality and correlation of the data,which can effectively reduce the number of iterations for model training and improve the accuracy of fault detection.Then,the WT-PCA-HMM fault detection model was obtained by using the expectation maximization algorithm combined with the training of observed sequence training.Finally,the logarithmic likelihood values of the finishing process data was derived from the model to achieve the fault detection.The results show that the WT-PCA-HMM fault detection method can not only reduce the false alarm rate by 8.1%compared with the traditional HMM method,but also reduce the number of model training iterations by 50%,which provides a new method for the fault detection.
作者
张瑞成
崔凯鑫
梁卫征
Zhang Ruicheng;Cui Kaixin;Liang Weizheng(College of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China)
出处
《锻压技术》
CAS
CSCD
北大核心
2023年第3期126-131,共6页
Forging & Stamping Technology
基金
河北省自然科学基金资助项目(F2018209201)
唐山市科技局科技计划项目(22130213G)
河北省省属高校基本科研业务费资助项目(JQN2021021)。
关键词
带钢
热连轧
故障检测
小波变换
主成分分析
隐马尔科夫模型
strip steel
hot rolling
fault detection
wavelet transform
principal component analysis
hidden Markov model