摘要
为了对滚动轴承的工作状态及故障类别进行准确的诊断,本文采用长短时记忆(LSTM)神经网络作为分类器对滚动轴承数据集进行分类诊断。首先,从滚动轴承原始运行振动信号中提取时域和频域特征参数,组成具有高维特征参数的数据集;使用核主成分分析(KPCA)方法对高维特征集进行降维处理,选取重要性程度高的特征构成输入特征向量。然后,针对LSTM神经网络在滚动轴承故障诊断中存在的超参数难以确定的问题,提出一种基于自适应t分布策略的麻雀搜索算法优化LSTM神经网络的故障诊断方法(tSSA–LSTM)。最后,使用凯斯西储大学滚动轴承数据中心的数据进行故障诊断精度测试、泛化性能测试及噪声环境下故障诊断性能测试等多个仿真实验,并将本文提出的诊断模型与麻雀搜索算法优化长短时记忆神经网络(SSA–LSTM)、遗传算法优化长短时记忆神经网络(GA–LSTM)、粒子群算法优化长短时记忆神经网络(PSO–LSTM)及传统LSTM诊断模型进行对比。结果表明:tSSA可以更有效地对LSTM的隐含层神经元数量、周期次数、学习率等超参数进行合理优化;所提方法的平均诊断准确率达到98.86%,交叉验证平均诊断结果为98.57%;所提方法在噪声干扰下的故障诊断准确率也优于对比方法。因此,本文提出的tSSA–LSTM模型不仅可以更精准地诊断滚动轴承故障状态,而且具有更强的泛化能力及抗干扰能力,有效地提高了滚动轴承故障诊断的性能。
In order to precisely diagnose the operational status and fault categories of rolling bearings,this paper employs a long short-term memory(LSTM)neural network as a classifier to categorize and diagnose the rolling bearing dataset.Initially,time-domain and frequency-do-main characteristic parameters are derived from the raw vibration signals of rolling bearings,creating a dataset with high-dimensional characterist-ic parameters.The Kernel Principal Component Analysis method is utilized to reduce the dimensionality of the high-dimensional characteristic set.The features with a high degree of significance are then selected to compose the input characteristic vector.Addressing the challenge of de-termining hyperparameters for the LSTM neural network in rolling bearing fault diagnosis research,a fault diagnosis model(tSSA-LSTM)is pro-posed based on the adaptive t-distribution sparrow search algorithm(tSSA)to optimize the LSTM neural network.Finally,several simulation ex-periments,including fault diagnosis accuracy tests,generalization performance tests,and fault diagnosis performance tests in noisy environments,are conducted using data from the Rolling Bearings Data Center at Case Western Reserve University.The proposed diagnostic model is experi-mentally compared with SSA-LSTM,GA-LSTM,PSO-LSTM,and traditional LSTM diagnostic models.The experimental results demonstrate that tSSA can more effectively optimize the hyperparameters of LSTM,such as the number of neurons in the hidden layer,the number of cycles,and the learning rate.The proposed method achieves an average diagnostic accuracy of 98.86%and a cross-validation result of 98.57%.Further-more,the fault diagnosis accuracy of the proposed method under noisy interference surpasses that of the compared methods.Consequently,the tSSA-LSTM model not only accurately diagnoses the state of rolling bearing faults but also exhibits stronger generalization and anti-interference capabilities,effectively enhancing the performance of rolling bearing fault diagnosis.
作者
周玉
房倩
裴泽宣
白磊
ZHOU Yu;FANG Qian;PEI Zexuan;BAI Lei(School of Electrical Eng.,North China Univ.of Water Resources and Electric Power,Zhengzhou 450045,China)
出处
《工程科学与技术》
EI
CAS
CSCD
北大核心
2024年第2期289-298,共10页
Advanced Engineering Sciences
基金
国家自然科学基金项目(U1504622,31671580)
河南省高等学校青年骨干教师培养计划项目(2018GGJS079)。
关键词
麻雀搜索算法
故障诊断
长短时记忆神经网络
特征提取
滚动轴承
sparrow search algorithm
fault diagnosis
long short-term memory neural network
feature extraction
rolling bearing
