Online meta-learning approach for sensor fault diagnosis using limited data
Source abstract
Abstract
Abstract The accurate and timely diagnosis of sensor faults plays a critical role in ensuring the reliability and performance of structural health monitoring (SHM) systems. However, the challenge is detecting, locating, and estimating sensor faults in an online manner using limited training data. To resolve this problem, a novel approach for online sensor fault diagnosis is proposed for SHM. The proposed approach is based on meta-learning, which enables superior model generalization capabilities using limited data. The detection, localization, and estimation of typical sensor faults in an online manner can be achieved efficiently by the proposed approach. First, a one-dimensional convolutional neural network (1D CNN) is designed to detect and locate faulty sensors. The initial model parameters of the 1D CNN are optimized using a model-agnostic meta-learning training strategy. This strategy allows the acquisition of transferable prior knowledge, which can speed up the learning process on new sensor fault detection and localization tasks. The meta-learning strategy also enables efficient and accurate detection and localization of potential faulty sensors with limited data. After detecting and locating the faulty sensors, an online updating algorithm based on a dual Kalman filter is used to estimate the severity of sensor faults and structural states simultaneously. The proposed approach is demonstrated with simulated sensor faults that cover a numerical example and field measurements from the Canton Tower. The results show that the proposed approach is applicable for online sensor fault diagnosis in SHM.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Fault Detection and Control Systems · Anomaly Detection Techniques and Applications · Advanced Sensor and Control Systems
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (42)
Show source-provided references
1. Corrosion fatigue crack growth prediction of bridge suspender wires using Bayesian Gaussian process10.1016/j.ijfatigue.2022.107377 ↗
2. Numerical simulation on crack–inclusion interaction for rib-to-deck welded joints in orthotropic steel deck10.3390/met13081402 ↗
3. Anomaly detection for large span bridges during operational phase using structural health monitoring data10.1088/1361-665X/ab79b3 ↗
4. Hierarchical outlier detection approach for online distributed structural identification10.1002/stc.2623 ↗
5. Review on field monitoring of high-rise structures10.1002/stc.2629 ↗
6. Early warning for abnormal strains of continuous bridges using a proposed temperature–strain mapping model10.1088/1361-665X/ad0c01 ↗
7. Development of sensor validation methodologies for structural health monitoring: a comprehensive review10.1016/j.measurement.2017.05.064 ↗
8. Sensor fault management techniques for wireless smart sensor networks in structural health monitoring10.1002/stc.2362 ↗
9. The 1st international project competition for structural health monitoring (IPC-SHM, 2020): a summary and benchmark problem10.1177/14759217211006485 ↗
10. Detection, identification, and quantification of sensor fault in a sensor network10.1016/j.ymssp.2013.05.007 ↗
11. Detection of shifts in GPS measurements for a long-span bridge using CUSUM chart10.1142/S0219455416400241 ↗
12. Bayesian combination of weighted principal-component analysis for diagnosing sensor faults in structural monitoring systems10.1061/(ASCE)EM.1943-7889.0001309 ↗
13. ICA based sensors fault diagnosis: an audio separation application10.1007/s11277-021-08184-x ↗
14. Decentralized fault detection and isolation in wireless structural health monitoring systems using analytical redundancy10.1016/j.advengsoft.2014.02.005 ↗
15. A sensor fault detection strategy for structural health monitoring systems10.12989/sss.2017.20.1.043 ↗
16. Residual statistics-based current sensor fault diagnosis for smart battery management10.1109/JESTPE.2021.3131696 ↗
17. Machine learning algorithms and fault detection for improved belief function based decision fusion in wireless sensor networks10.3390/s19061334 ↗
18. Distributed adaptive diagnosis of sensor faults using structural response data10.1088/0964-1726/25/10/105019 ↗
19. A distributed sensor-fault detection and diagnosis framework using machine learning10.1016/j.ins.2020.08.068 ↗
20. 1D convolutional neural networks and applications: a survey10.1016/j.ymssp.2020.107398 ↗
21. Explainable 1D convolutional neural network for damage detection using Lamb wave10.1016/j.ymssp.2021.108220 ↗
22. Bayesian transfer learning with active querying for intelligent cross-machine fault prognosis under limited data10.1016/j.ymssp.2022.109628 ↗
23. Mgml: momentum group meta-learning for few-shot image classification10.1016/j.neucom.2022.10.012 ↗
24. Minet: meta-learning instance identifiers for video object detection10.1109/TIP.2021.3099409 ↗
25. A novel method based on meta-learning for bearing fault diagnosis with small sample learning under different working conditions10.1016/j.ymssp.2021.108765 ↗
26. Model-agnostic meta-learning for fast adaptation of deep networks
27. Sensor fault diagnosis for structural health monitoring based on statistical hypothesis test and missing variable approach10.1061/(ASCE)AS.1943-5525.0000572 ↗
28. Canonical correlation analysis based fault diagnosis method for structural monitoring sensor networks10.12989/sss.2016.17.6.1031 ↗
29. Sensor fault diagnosis with a probabilistic decision process10.1016/j.ymssp.2012.07.014 ↗
30. Sensor validation using minimum mean square error estimation10.1016/j.ymssp.2009.12.001 ↗
31. Probabilistic reconstruction for spatiotemporal sensor data integrated with Gaussian process regression10.1016/j.probengmech.2022.103264 ↗
32. A new bearing fault diagnosis method based on modified convolutional neural networks10.1016/j.cja.2019.07.011 ↗
33. Sensor fault detection, localization, and reconstruction for online structural identification10.1002/stc.2925 ↗
34. Batch normalization: accelerating deep network training by reducing internal covariate shift
35. A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals10.3390/s17020425 ↗
36. A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load10.1016/j.ymssp.2017.06.022 ↗
37. Random search for hyper-parameter optimization
38. Meta-transfer learning for few-shot learning
39. Prechelt · 2022 · 55
40. Statistical inference by confidence intervals: issues of interpretation and utilization10.1093/ptj/79.2.186 ↗
41. A comparative analysis of signal decomposition techniques for structural health monitoring on an experimental benchmark10.3390/s21051825 ↗
42. Development of a structural health monitoring benchmark problem for high rise slender structures10.4028/www.scientific.net/AST.56.489 ↗