-
摘要:
针对航空发动机传感器的数据缺失问题，提出基于张量奇异值阈值(TSVT)的张量重构模型LRTC-PTNN，对航空发动机的传感器数据进行重构。LRTC-PTNN模型运用截断p-shrinkage范数的方式代替原始张量迹范数作为张量秩的凸包络，并根据TSVT的特性，计算了传感器之间的相关性，选取传感器截面作为重构精度最佳的数据输入方向，使用交替乘子法实现LRTC-PTNN算法。选取NASA提供的PHM2008数据集进行实验，对数据集进行标准化，并在重构后进行恢复，将多个时间序列个数相近的发动机传感器数据构建为高维张量的形式，设置2种传感器的数据缺失场景进行实验，结果表明：重构后数据的均方根误差和平均绝对百分比误差范围分别为2.10～13.13和0.32～1.49，LRTC-PTNN模型优于现有的基线模型，且在极端情况下有较强的鲁棒性。
-
关键词:
- 航空发动机 /
- 数据缺失 /
- 张量 /
- 截断p-shrinkage范数 /
- 交替乘子法
Abstract:To address the data loss problem of aero engine sensors, a tensor reconstruction model LRTC-PTNN based on tensor singular value threshold (TSVT) is proposed to reconstruct the sensor data of aircraft engines. LRTC-PTNN uses truncation p-shrinkage norm to replace the original tensor trace norm as the convex envelope of tensor rank. According to the characteristics of TSVT, the correlation between sensors is calculated, and the data input direction with the best reconstruction accuracy is selected. The LRTC-PTNN algorithm was finally implemented using the alternating direction method of multipliers. Using the PHM2008 dataset provided by NASA for experiments, the dataset was standardized and restored after reconstruction, and the multiple time series similar number of engine sensor data were constructed into the form of high-dimensional tensor, and the data deletion scenarios of the two sensors were set for experiments. The results showed that the RMSE and MAPE values of the reconstructed data were between 2.10%−13.13% and 0.32%−1.49%, respectively; the LRTC-PTNN model was better than the existing baseline model; in extreme cases, the model also has strong robustness.
-
图 3 RM情况下重构值与原始值对比
Figure 3. Reconstructed value compared with original value in case of RM
图 4 NM情况下重构值与原始值对比
Figure 4. Reconstructed value compared with original value in case of NM
表 1 RM情况下的数据重构精度对比
Table 1. Comparison of data reconstruction accuracy at RM
缺失率/% 均方根误差 平均绝对百分比误差 KNN LRMC SVT LRTC-PTNN KNN LRMC SVT LRTC-PTNN 20 2.51 3.61 7.96 2.37 0.152 0.80 1.82 0.37 30 4.30 3.98 8.36 2.47 0.26 0.91 1.87 0.38 40 13.71 5.22 9.38 2.92 1.39 1.04 1.91 0.44 50 28.68 7.87 12.69 3.83 3.63 1.32 2.17 0.56 60 38.02 15.49 21.89 6.17 4.93 1.77 3.03 0.82 70 41.67 30.62 65.08 13.13 5.22 2.68 7.27 1.49 
下载: 导出CSV
表 2 NM情况下的数据重构精度对比
Table 2. Comparison of data reconstruction accuracy at NM
缺失率/% 均方根误差 平均绝对百分比误差 KNN LRMC SVT LRTC-PTNN KNN LRMC SVT LRTC-PTNN 20 2.57 3.59 7.80 2.10 0.14 0.78 1.77 0.32 30 2.57 3.52 7.59 2.43 0.15 0.91 1.88 0.38 40 3.00 5.23 10.27 2.65 0.20 1.10 1.96 0.43 50 15.37 7.51 11.51 4.28 0.90 1.32 2.10 0.63 60 28.81 12.03 29.01 5.12 2.90 1.76 3.85 0.76 70 32.94 29.75 72.40 11.66 3.22 2.72 8.23 1.34
下载: 导出CSV
-
[1] HUANG Y, TANG Y F, VANZWIETEN J, et al. Reliable machine prognostic health management in the presence of missing data[J]. Concurrency and Computation Practice and Experience, 2020, 34(12): e5762. [2] 孙瑞谦, 缑林峰, 韩小宝, 等. 考虑性能退化的航空发动机故障诊断量化评估[J]. 推进技术, 2022, 43(8): 337-348. doi: 10.13675/j.cnki.tjjs.210247SUN R Q, GOU L F, HAN X B, et al. Quantitative evaluation of aeroengine fault diagnosis considering performance degradation[J]. Propulsion Technology, 2022, 43(8): 337-348(in Chinese). doi: 10.13675/j.cnki.tjjs.210247 [3] LI X Q, JIANG H K, LIU Y, et al. An integrated deep multiscale feature fusion network for aeroengine remaining useful life prediction with multisensor data[J]. Knowledge-based Systems, 2022, 235: 107652. doi: 10.1016/j.knosys.2021.107652 [4] 王昆, 郭迎清, 赵万里, 等. 基于SSAE和相似性匹配的航空发动机剩余寿命预测[J]. 北京航空航天大学学报, 2023, 49(10): 2817-2825.WANG K, GUO Y Q, ZHAO W L, et al. Residual life prediction of aero-engine based on SSAE and similarity matching[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 49(10): 2817-2825(in Chinese). [5] 孙浩, 郭迎清, 赵万里. 航空发动机传感器与执行机构信息重构算法[J]. 北京航空航天大学学报, 2020, 46(2): 331-339. doi: 10.13700/j.bh.1001-5965.2019.0240SUN H, GUO Y Q, ZHAO W L. Aircraft engine sensor and actuator information reconstruction algorithm[J]. Journal of Beijing University of Aeronautics and Astronautics, 2020, 46(2): 331-339(in Chinese) . doi: 10.13700/j.bh.1001-5965.2019.0240 [6] NI D H, LEONARD J D, GUIN A, et al. Multiple imputation scheme for overcoming the missing values and variability issues in ITS data[J]. Journal of Transportation Engineering, 2005, 131(12): 931-938. doi: 10.1061/(ASCE)0733-947X(2005)131:12(931) [7] CAI J F, CANDES E J, SHEN Z W, et al. A singular value thresholding algorithm for matrix completion[J]. SIAM Journal on Optimization, 2010, 20(4): 1956-1982 [8] LIU J, MUSIALSKI P, WONKA P, et al. Tensor completion for estimating missing values in visual data[C]//Proceedings of the IEEE International Conference on Computer Vision. Piscataway: IEEE Press, 2009: 2114-2121. [9] CHEN X Y, HE Z, SUN L. A Bayesian tensor decomposition approach for spatiotemporal traffic data imputation[J]. Transportation Research Part C:Emerging Technologies, 2018, 98: 73-84. [10] CHEN X Y, SUN L. Bayesian temporal factorization for multidimensional time series prediction[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(9): 4659-4673. [11] 周媛, 左洪福, 何军. 信息缺失的航空发动机传感器数据重构[J]. 北京航空航天大学学报, 2016, 42(5): 891-898. doi: 10.13700/j.bh.1001-5965.2015.0350ZHOU Y, ZUO H F, HE J. Information missing aircraft engine sensor data refactoring[J]. Journal of Beijing University of Aeronautics and Astronautics, 2016, 42(5): 891-898(in Chinese). doi: 10.13700/j.bh.1001-5965.2015.0350 [12] ELNOUR M, MESKIN N, AL-NAEMI M. Sensor data validation and fault diagnosis using auto-associative neural network for HVAC systems[J]. Journal of Building Engineering, 2020, 27: 100935. doi: 10.1016/j.jobe.2019.100935 [13] ZHAO Z, LIU Z, SUN Y, et al. WOS-ELM-based double redundancy fault diagnosis and reconstruction for aeroengine sensor[J]. Journal of Control Science and Engineering, 2017, 2017: 1982879. [14] LI Y F, SHANG K, HUANG Z H. A singular value p-shrinkage thresholding algorithm for low rank matrix recovery[J]. Computational Optimization & Applications, 2019, 73(2): 453-476. [15] LIU C, SHAN H, CHEN C. Tensor p-shrinkage nuclear norm for low-rank tensor completion[J]. Neurocomputing, 2020, 387: 255-267. doi: 10.1016/j.neucom.2020.01.009 [16] ZHANG D, HU Y, YE J, et al. Matrix completion by truncated nuclear norm regularization[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2012: 2192-2199. [17] ZHANG Z M, AERON S. Exact tensor completion using t-SVD[J]. IEEE Transactions on Signal Processing, 2017, 65(6): 1511-1526. doi: 10.1109/TSP.2016.2639466 -
下载:
点击查看大图
计量
- 文章访问数: 561
- HTML全文浏览量: 107
- PDF下载量: 9
- 被引次数: 0
