检索
E-mail
RSS
作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 开发研究与工程应用 • 上一篇    下一篇

LSSVM与HMM在航空发动机状态预测中的应用研究

崔建国 1,高波 1,蒋丽英 1,于明月 1,郑蔚 2   

  1. (1.沈阳航空航天大学 自动化学院,沈阳 110136; 2.故障诊断与健康管理技术航空科技重点实验室,上海 201601)
  • 收稿日期:2016-07-14 出版日期:2017-10-15 发布日期:2017-10-15
  • 作者简介:崔建国(1963—),男,教授、博士,主研方向为可视化仿真;高波,硕士研究生;蒋丽英,副教授;于明月,讲师;郑蔚,助理工程师。
  • 基金项目:
    航空科学基金(20153354005,20163354004);辽宁省自然科学基金(2014024003)。

Application Research of LSSVM and HMM in Aeroengine Condition Prediction

CUI Jianguo  1,GAO Bo  1,JIANG Liying  1,YU Mingyue  1,ZHENG Wei  2   

  1. (1.School of Automation,Shenyang Aerospace University,Shenyang 110136,China;2.Aviation Key Laboratory of Science and Technology on Fault Diagnosis and Health Management,Shanghai 201601,China)
  • Received:2016-07-14 Online:2017-10-15 Published:2017-10-15

PDF

316

被引次数

1

摘要： 传统单一的状态预测方法难以精确预测航空发动机状态的缺陷,而最小二乘支持向量机(LSSVM)具有较强的非线性预测能力和泛化能力,可以有效地对信号进行非线性预测,隐马尔科夫模型(HMM)有利于处理连续的动态信号,能够精确计算出似然度概率。提出一种结合LSSVM与HMM的状态预测方法。利用提升小波函数全阈值降噪法对采集的振动信号进行降噪,采用小波包分解提取有效的特征,选择不同状态下的特征量训练多个HMM模型,并通过此模型对未知信号特征量以及LSSVM预测的特征量进行状态监测,从而预测出发动机未来时刻的状态以及状态的退化趋势。实验结果表明,该方法的预测准确率达到92%以上,能够有效地预测航空发动机的状态情况。

关键词: 航空发动机, 最小二乘支持向量机, 隐马尔科夫模型, 状态预测, 振动信号, 降噪, 小波包分解

Abstract: The traditional single condition prediction methods cannot predict the condition of aeroengine accurately.As Least Squares Support Vector Machine(LSSVM) has strong nonlinearly prediction and generalization ability,the signal can be nonlinear predicted effectively.Hidden Markov Model(HMM) which is conducive to processing continuous dynamic signals,can calculate the likelihood probability characteristics accurately.Hence,a condition forecast method based on LSSVM and HMM is presented.It uses the lifting wavelet threshold denoising method to denoise the collected vibration signal,and the wavelet packet decomposition to extract the effective features.Multiple HMM models are trained by selecting the features of different conditions.Then the unknown signal feature and its corresponding prediction calculated by LSSVM are both applied to the trained HMM model for condition,thus predicting the condition of the future time and the degradation trend of the condition.Experimental results show that the monitoring accuracy of prediction reaches more than 92%,which verifies the effectiveness of the proposed method.

Key words: aeroengine, Least Squares Support Vector Machine(LSSVM), Hidden Markov Model(HMM), condition prediction, vibration signal, denoising, wavelet packet decomposition

中图分类号: