基于连续型HMM和PSO-SVM的滚动轴承剩余寿命预测
|
摘要
为了准确识别轴承当前所处的退化状态,并进一步精确有效地预测其剩余寿命,提出一种基于连续型隐马尔可夫模型(CHMM)与PSO-SVM相结合的预测方法。首先,提取轴承振动全寿命周期信号的时域、频域、时频域的特征,并构建特征空间;然后,利用CHMM将轴承全寿命周期划分若干个退化阶段,并通过选取不同阶段的特征训练样本,采用PSO-SVM进行预测模型的训练,分别得到不同阶段的剩余寿命预测模型;最后,运用滚动轴承全寿命数据对所提方法进行测试,并与全寿命周期数据CHMM分区段后的SVM模型和未分区段的PSO-SVM模型的预测方法作对比。实验结果表明该方法能有效提高预测精度,具有一定的实用性。
In order to accurately identify the degradation state of the bearing and predict its residual life accurately and effectively, a prediction method based on Continuous Hidden Markov Model(CHMM) and Particle Swarm Optimization and Support Vecotr Machine(PSO-SVM) was proposed. Firstly, characteristics of the time domain, frequency domain and time-frequency domain of the bearing life cycle signal were extracted, and the feature space was constructed. Then, CHMM was used to divide the bearing life cycle into several degradation stages, and with the training samples of different stages selected, PSO-SVM was used to train and obtain the residual life prediction models of different stages. Finally, experiments were conducted on the full-life cycle data of rolling bearings with the comparison with pretection methods of segmented SVM model with full-life cycle data CHMM and un-segmented PSO-SVM model. The experimental results show that the proposed method can effectively improve the prediction accuracy and has certain practicability.
In order to accurately identify the degradation state of the bearing and predict its residual life accurately and effectively, a prediction method based on Continuous Hidden Markov Model(CHMM) and Particle Swarm Optimization and Support Vecotr Machine(PSO-SVM) was proposed. Firstly, characteristics of the time domain, frequency domain and time-frequency domain of the bearing life cycle signal were extracted, and the feature space was constructed. Then, CHMM was used to divide the bearing life cycle into several degradation stages, and with the training samples of different stages selected, PSO-SVM was used to train and obtain the residual life prediction models of different stages. Finally, experiments were conducted on the full-life cycle data of rolling bearings with the comparison with pretection methods of segmented SVM model with full-life cycle data CHMM and un-segmented PSO-SVM model. The experimental results show that the proposed method can effectively improve the prediction accuracy and has certain practicability.
引文
[1] 赵锦剑,杨光永,周安然,等.旋转机械振动信号的Kalman滤波及故障诊断[J].仪表技术与传感器,2014(5):80-83.
[2] 赵占飞,梁伟,郭晓燕.基于支持向量机模型的滚动轴承运行状态预测研究[J].石油矿场机械,2016,45(3):16-20.
[3] 夏新涛,叶亮,孙立明,等.滚动轴承性能保持可靠性预测[J].轴承,2016(6):28-34.
[4] 韩昕锋,任立坤.基于神经网络的轴承故障预测模型[J].海军航空工程学院学报,2015,30(3):281-285.
[5] 汪庆年,饶利强,龚文军,等.基于PSO优化的BP神经网络在电动机绝缘剩余寿命预测中的应用[J].水电能源科学,2015,33(12):161-164.
[6] 梁浩,杨光宇.基于连续隐马尔可夫的语音识别模型[J].无线互联科技,2013(6):56-57.
[7] 洪文,黄凤岗,苏菡.基于连续隐马尔科夫模型的步态识别[J].应用科技,2005,32(2):50-52.
[8] LEE J,QIU H,YU G,et al.Bearing data set [DB/OL].NASA Ames Prognostics Data Repository,2007[2015- 05- 04].http://ti.arc.nasa.gov/tech/dash/pcoe/prognostic-data-repository.
[9] 董绍江.基于优化支持向量机的空间滚动轴承寿命预测方法研究[D].重庆:重庆大学,2012:61-86.
[10] 陈英强,刘亚萌,蒋劲,等.基于EMD的自适应滤波方法在水泵故障诊断中的应用[J].机械工程与技术,2015,4(4):270-277.
[11] 范庚,马登武.基于EMD和RVM-AR的航空发动机磨损故障预测模型[J].计算机测量与控制.2013,21(7):1746-1749.
[12] 赵春临,郑崇勋,赵敏.基于脑电功率谱-连续隐马尔科夫链的精神疲劳分级模型[J].西安交通大学学报.2007,41(12):1474-1478.
[13] 申中杰,陈雪峰,何正嘉,等.基于相对特征和多变量支持向量机的滚动轴承剩余寿命预测[J].机械工程学报,2013,49(2):183-188.
[14] BISHOP C M.Pattern Recognition and Machine Learning (Information Science and Statistics)[M].Berlin:Springer,2006:237-241.
[15] 卢亚洲,王学文,杨兆建,等.PSO-SVM在提升机制动系统故障诊断中的应用[J].机械设计与制造,2018(6):239-242.
[2] 赵占飞,梁伟,郭晓燕.基于支持向量机模型的滚动轴承运行状态预测研究[J].石油矿场机械,2016,45(3):16-20.
[3] 夏新涛,叶亮,孙立明,等.滚动轴承性能保持可靠性预测[J].轴承,2016(6):28-34.
[4] 韩昕锋,任立坤.基于神经网络的轴承故障预测模型[J].海军航空工程学院学报,2015,30(3):281-285.
[5] 汪庆年,饶利强,龚文军,等.基于PSO优化的BP神经网络在电动机绝缘剩余寿命预测中的应用[J].水电能源科学,2015,33(12):161-164.
[6] 梁浩,杨光宇.基于连续隐马尔可夫的语音识别模型[J].无线互联科技,2013(6):56-57.
[7] 洪文,黄凤岗,苏菡.基于连续隐马尔科夫模型的步态识别[J].应用科技,2005,32(2):50-52.
[8] LEE J,QIU H,YU G,et al.Bearing data set [DB/OL].NASA Ames Prognostics Data Repository,2007[2015- 05- 04].http://ti.arc.nasa.gov/tech/dash/pcoe/prognostic-data-repository.
[9] 董绍江.基于优化支持向量机的空间滚动轴承寿命预测方法研究[D].重庆:重庆大学,2012:61-86.
[10] 陈英强,刘亚萌,蒋劲,等.基于EMD的自适应滤波方法在水泵故障诊断中的应用[J].机械工程与技术,2015,4(4):270-277.
[11] 范庚,马登武.基于EMD和RVM-AR的航空发动机磨损故障预测模型[J].计算机测量与控制.2013,21(7):1746-1749.
[12] 赵春临,郑崇勋,赵敏.基于脑电功率谱-连续隐马尔科夫链的精神疲劳分级模型[J].西安交通大学学报.2007,41(12):1474-1478.
[13] 申中杰,陈雪峰,何正嘉,等.基于相对特征和多变量支持向量机的滚动轴承剩余寿命预测[J].机械工程学报,2013,49(2):183-188.
[14] BISHOP C M.Pattern Recognition and Machine Learning (Information Science and Statistics)[M].Berlin:Springer,2006:237-241.
[15] 卢亚洲,王学文,杨兆建,等.PSO-SVM在提升机制动系统故障诊断中的应用[J].机械设计与制造,2018(6):239-242.