摘要
针对现有倾斜仪故障记录的甄别与分类主要依靠人工经验,诊断过程烦琐及诊断精度不高的问题,充分利用信号分解技术的特征升维与机器学习模型的快速自动分类等优势,提出一种结合自适应噪声完备集成经验模态分解多尺度模糊熵和麻雀搜索算法(sparrow search algorithm,SSA)优化反向传播神经网络(back propagation neural network,BPNN)的VP型宽频带垂直摆倾斜仪故障自动诊断方法。该方法首先将故障数据归一化,利用自适应噪声完备集成经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)信号得到本征模态函数(intrinsic mode function,IMF)的多尺度模糊熵,然后基于SSA优化BPNN的权值与阈值,得到最佳权值和阈值,最后应用SSA-BPNN模型对倾斜仪故障特征数据进行辨识。实验表明:CEEMDAN多尺度模糊熵判据具有良好的倾斜仪故障特征区分效果;SSA-BP神经网络模型在准确率和召回率上,相比BP模型分别提高4.7581、6.3216个百分点,辨识过程更稳健,弥补了VP型倾斜仪在故障智能识别领域的空白。
Aiming at the problems that the identification and classification of existing inclinometer fault records mainly rely on manual experience,the diagnosis process is cumbersome and the diagnosis accuracy is not high,making full use of the advantages of signal decomposition technology,such as feature enhancement and fast automatic classification of machine learning models,a new method was proposed.Combined with complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and multi-scale fuzzy entropy,and sparrow search algorithm(SSA)to optimize the fault automatic diagnosis method of VP type broadband vertical pendulum inclinometer with back propogation neural network(BPNN).The method first normalizes the fault data,uses CEEMDAN to decompose the signal to obtain the multi-scale fuzzy entropy of the eigenmode function(IMF),and then optimizes the weights and thresholds of the BP neural network based on the sparrow search algorithm(SSA)to obtain the optimal Weights and thresholds,and finally the SSA-BPNN model is used to identify the fault characteristic data of the inclinometer.Experiments show that the CEEMDAN multi-scale fuzzy entropy criterion has a good effect of distinguishing fault features of the inclinometer.The SSA-BP neural network model is 4.7581 and 6.3216 percentage points higher than the BP model in terms of accuracy and recall,respectively.More robust,making up for the blank of the VP-type inclinometer in the field of intelligent fault identification.
作者
庞聪
马武刚
李查玮
江勇
王磊
廖成旺
PANG Cong;MA Wu-gang;LI Cha-wei;JIANG Yong;WANG Lei;LIAO Cheng-wang(Institute of Seismology,CEA,Wuhan 430071,China;Hubei Key Laboratory of Earthquake Early Warning,Wuhan 430071,China;Hubei Earthquake Administration,Wuhan 430071,China;Hebei Key Laboratory of Earthquake Dynamics,Sanhe 065201,China;Department of Electronic and Optical Engineering,Space Engineering University,Beijing 101416,China)
出处
《科学技术与工程》
北大核心
2022年第35期15612-15616,共5页
Science Technology and Engineering
基金
河北省地震动力学重点实验室开放基金(FZ202212)
湖北省自然科学基金(2019CFB768)
中国地震局地震研究所和应急管理部国家自然灾害防治研究院基本科研业务费专项(IS202236328,IS202226321)
中国地震局监测、预报、科研三结合课题项目(3JH-202201024)
地震科技星火计划攻关项目(XH15030)
武汉引力与固体潮国家野外科学观测研究站开放基金(WHYWZ202208)。
关键词
倾斜仪故障识别
自适应噪声完备集成经验模态分解
麻雀搜索算法
反向传播神经网络
多尺度模糊熵
inclinometer fault identification
complete ensemble empirical mode decomposition with adaptive noise
sparrow search algorithm
back propagation neural network
multi-scale fuzzy entropy
