A Lightweight SHM Framework Based on Adaptive Multisensor Fusion Network and Multigeneration Knowledge Distillation
The DOI matches its Crossref record; 53 of the 55 references it cites were found in a registry or an index.
Abstract
Structural health monitoring (SHM) technology is of great importance to ensure the long-term and reliable operation of engineering structures. The monitoring systems for large structures are characterized by large numbers of measurement points and complex signal. In recent years, deep learning has been widely used in many fields, such as intelligent fault diagnosis and health monitoring due to its excellent feature extraction and nonlinear fitting ability. However, the existing deep networks for condition monitoring have some drawbacks. First, the multisensor information fusion is rarely considered, which is quite common in SHM systems. Second, most of the proposed models by current research are with large parameters and high complexity. To solve the problems mentioned above, we propose a lightweight SHM framework suitable for the multisensor monitoring tasks. In other words, the adaptive multisensor fusion network (ASF-Net) is proposed to dynamically mix with multichannel input samples and extract the efficient feature on different scales. Moreover, a joint loss function is established for network compression by transferring the knowledge from bulky teacher network to lightweight student network. Finally, to reduce the performance degradation during model compression, a multigeneration knowledge distillation (MGKD) strategy is proposed. The experiments on two SHM datasets show that the proposed method can achieve excellent classification accuracy, and model performance degradation of less than 2% can be obtained with at most 40.24 times less parameters and 128.82 times less computation, which has a great prospect on resource-constrained platforms.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“In the studied literature, vibration-based SHM is mainly used in crack [109], brace damage [110,111,112], bolt loosening [113,114], and stiffness reduction [115,116] detection.”
From Deep Learning for Structural Health Monitoring: Data, Algorithms, Applications, Challenges, and Trends (2023) · cites this paper as [114]
Citations
Cites
MobileNetV2: Inverted Residuals and Linear Bottlenecks
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00474 · Cited by 28,755 · cited 1×
“To further illustrate the advantages of the proposed network compared with other neural networks, several commonly used networks are selected for comparison, including lightweight networks MobileNet [48], ShuffleNet [49], SqueezeNet [50], Xcep-tion [42], and our previously proposed WDCNN [51].”
MethodXception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380 · cited 3×
“For Xception on the QUGS dataset, the Flops and Param are 2.82 and 3.56 times larger with 6.54% lower result.”
BackgroundMethodInfluentialDistilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204 · cited 1×
“Hinton et al. [33] introduced temperature parameters to control the softening degree of soft targets, which effectively accelerated the training convergence speed of student network.”
BackgroundShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices
Zhang, X., Zhou, X., Lin, M., & Sun, J. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00716 · Cited by 9,982
SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016) · arXiv · doi:10.48550/arxiv.1602.07360 · Cited by 6,201 · cited 2×
“To further illustrate the advantages of the proposed network compared with other neural networks, several commonly used networks are selected for comparison, including lightweight networks MobileNet [48], ShuffleNet [49], SqueezeNet [50], Xcep-tion [42], and our previously proposed WDCNN [51].”
BackgroundLarge-Scale Machine Learning with Stochastic Gradient Descent
Bottou, L. (2010) · Proceedings of COMPSTAT'2010 · doi:10.1007/978-3-7908-2604-3_16 · Cited by 5,795 · cited 1×
“The stochastic gradient descent (SGD) [47] is applied for the network parameters optimization.”
MethodFitNets: Hints for Thin Deep Nets
Romero, A., Ballas, N., Kahou, S. E., Chassang, A., Gatta, C., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.6550 · Cited by 2,859 · cited 1×
“…less parameters and 3) Comparison With Other Distillation Methods: In order to verify the superiority of the proposed distillation strategy, we compare it with other KD methods, which are L1 [52], L2 [53], FSP matrix [35], attention transfer (AT) [34], and relation knowledge transfer (RKT) [54].”
MethodZagoruyko, S., & Komodakis, N. (2016) · arXiv · doi:10.48550/arxiv.1612.03928 · Cited by 2,027 · cited 2×
“…less parameters and 3) Comparison With Other Distillation Methods: In order to verify the superiority of the proposed distillation strategy, we compare it with other KD methods, which are L1 [52], L2 [53], FSP matrix [35], attention transfer (AT) [34], and relation knowledge transfer (RKT) [54].”
BackgroundMethodThe discrete wavelet transform: wedding the a trous and Mallat algorithms
Shensa, M. J. (1992) · IEEE Transactions on Signal Processing · doi:10.1109/78.157290 · Cited by 1,994 · cited 1×
“Using (3) and (4) to make the inner product with (5), the decomposition coefficient can be given as Equations (6) and (7) are called Mallat algorithm [40] as well.”
BackgroundZhang, W., Peng, G., Li, C., Chen, Y., & Zhang, Z. (2017) · Sensors · doi:10.3390/s17020425 · Cited by 1,925 · cited 1×
“To further illustrate the advantages of the proposed network compared with other neural networks, several commonly used networks are selected for comparison, including lightweight networks MobileNet [48], ShuffleNet [49], SqueezeNet [50], Xcep-tion [42], and our previously proposed WDCNN [51].”
MethodSelf-citation
Research works printed but not linked: 1
- Chen, Learning efficient object detection models with knowledge distillation, 30, 1
Cited by
Jia, J., & Li, Y. (2023) · Sensors · doi:10.3390/s23218824 · Cited by 182
“In the studied literature, vibration-based SHM is mainly used in crack [109], brace damage [110–112], bolt loosening [113,114], and stiffness reduction [115,116] detection.”
BackgroundMethodA Systematic Review of Multisensor Information Fusion for Equipment Fault Diagnosis
Lin, T., Ren, Z., Zhu, L., Zhu, Y., Feng, K., Ding, W., Yan, K. L., & Beer, M. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3529577 · Cited by 119
Few-shot learning for structural health diagnosis of civil infrastructure
Xu, Y., Fan, Y., Bao, Y., & Li, H. (2024) · Advanced Engineering Informatics · doi:10.1016/j.aei.2024.102650 · Cited by 68
Guo, W. J., Li, X., & Shen, Z. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.123083 · Cited by 54
Xu, Y., Yan, X., Sun, B., Feng, K., Kou, L., Chen, Y., Li, Y., Chen, H., Tian, E., Ni, Q., et al. (2023) · IEEE Transactions on Transportation Electrification · doi:10.1109/tte.2023.3313986 · Cited by 42
“Li et al. [25] put forward a multigeneration KD algorithm to reduce the performance degradation of a CNN model and promote the model to achieve better results in machine fault classification tasks.”
MethodA Lightweight Gear Fault Diagnosis Method Based on Attention Mechanism and Multilayer Fusion Network
Wang, M., Yang, Y., Wei, L., & Li, Y. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3330231 · Cited by 32
“Li et al. [25] proposed an adaptive multisensor fusion network, and for the problem of a large number of model parameters and calculation amounts, the multigenerational knowledge distillation (MGKD) strategy is proposed, which reduces the risk of model performance degradation by integrating the…”
BackgroundLiu, F., Zhang, Z., & Li, S. (2023) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2023.3332075 · Cited by 18
“Duetothe tremendous success of deep learning meth-ods, particularly convolutional neural networks (CNNs), for tasks such as voice recognition [4], fault diagnosis [5], [6], and computer vision [7], an increasing number of researchers have applied deep learning methods to estimate 6-”
BackgroundSelf-citationZhang, F., Fu, Y., Shen, K., Zhang, Y., & Wang, J. (2025) · Engineering Structures · doi:10.1016/j.engstruct.2025.120406 · Cited by 17
Li, S., Xu, Y., Feng, K., Wang, Y., Sun, B., Yan, X., Sheng, X., Zhang, K., Zheng, J., & Ni, Q. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3301863 · Cited by 16
“[25] used a multigeneration KD method to transfer the knowledge from a bulky teacher network to a lightweight structure, thus achieving”
MethodSelf-citationAdaptive Learning Against Muscle Fatigue for A-Mode Ultrasound-Based Gesture Recognition
Zeng, J., Sheng, Y., Zhou, Z., Yang, Y., Wang, J., Xie, Q., & Liu, H. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3301862 · Cited by 15
“knowledge distillation technology [43], [44], [45], [46] can aid in the compression of the neural network.”
BackgroundZhang, Z., Zhang, C., Chen, L., Li, H., & Han, P. (2023) · Measurement Science and Technology · doi:10.1088/1361-6501/ace7eb · Cited by 11
“The currently prevalent lightweight strategies consist primarily of pruning [28, 29], knowledge distillation [30, 31], and lightweight module design [32, 33].”
Wan, W., Chen, J., & Xie, J. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3316793 · Cited by 11
“Li et al. [17] employed a joint loss function and a multigeneration KD strategy to compress a bulky teacher model into a lightweight student network, minimizing model performance degradation.”
2022 Index IEEE Transactions on Instrumentation and Measurement Vol. 71
(2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3238219 · Cited by 6
Zhang, Z., Zhang, C., Zhang, X., Gao, S., & Li, H. (2023) · Measurement Science and Technology · doi:10.1088/1361-6501/ad0939 · Cited by 4
“These strategies encompass techniques such as pruning [35, 36], knowledge distillation [37, 38], and lightweight design [39, 40], which collectively offer effective solutions.”
A shallow 2D-CNN network for crack detection in concrete structures
Honarjoo, A., & Darvishan, E. (2024) · International Journal of Structural Integrity · doi:10.1108/ijsi-08-2023-0082 · Cited by 4
You, M., Ren, B., Han, X., & Zhou, H. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3417596 · Cited by 3
“In addition, Li et al. [22] employed an adaptive multisensor fusion network with a multigeneration knowledge distillation strategy for efficient feature extraction and network compression.”
BackgroundStructural Plastic Damage Warning and Real-Time Sensing System Based on Cointegration Theory
Gao, Q., Huo, J., Wang, Y., Wang, X., & Wang, C. (2024) · Sensors · doi:10.3390/s24185961 · Cited by 2
A lightweighted planetary gearbox fault diagnosis method based on multisensor fusion
Chen, R., Zeng, C., Puzhou, W., Zhou, Q., & Cai, H. (2025) · Measurement Science and Technology · doi:10.1088/1361-6501/ae035a · Cited by 2
“Li et al [22] proposed a knowledge distillation strategy, which realizes the compression model and integrates knowledge from multiple networks.”
Deng, C., Li, Y., Dai, J., Xiong, F., & Zeng, Y. (2025) · Measurement · doi:10.1016/j.measurement.2025.119123 · Cited by 1
Liu, J., Zhang, W., Sun, L., & Li, Y. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.115515 · Cited by 1
Lightweight comment review model based on large-scale knowledge compression
Chen, R., Chen, Z., Wu, Y., & Wu, Z. (2026) · IEEE Conference on Telecommunications, Optics and Computer Science (TOCS) · doi:10.1117/12.3107485
“…that utilizes a large pre-trained model as a teacher to instruct a small student model to learn key knowledge, and the student model not only learns hard labeled information, but also absorbs probability distributions and intermediate layer feature representations output by the teacher model [9].”
A physical reservoir involving frequency virtual nodes for structural damage detection
Takashima, K., Watanabe, N., Hiraga, M., Miura, N., & MASUDA, A. (2026) · Scientific Reports · doi:10.1038/s41598-026-56840-8
Hu, C., Chen, C., Si, X., Yue, J., Chen, M., & He, C. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7389570
Adarsh, S., Panda, J., & Chaudhuri, S. R. (2026) · Journal of Civil Structural Health Monitoring · doi:10.1007/s13349-026-01134-w
“Some standard techniques used for multi-sensor data fusion in SHM include classical and Bayesian estimation theory, Dempster-Shafer theory [16–18], state estimation, machine learning [19 ], and fuzzy logic.”
Metadata
- Authors
- Li, S., Liu, F., Peng, G., Cheng, F., Zhao, B., & Ji, M.
- Affiliations
- Harbin Institute of Technology (CN)
- Venue
- IEEE Transactions on Instrumentation and Measurement
- Published
- 2022
- Volume
- 71
- Pages
- 1-19
- ISSN
- 0018-9456 (ISSN-L)
- 1557-9662
- OpenAlex
- W4312517768
- Cited by
- 18
- References
- 55
- Citations by year
- 2026: 2 · 2025: 3 · 2024: 3 · 2023: 10
- FWCI
- 1.80
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Instrumentation and Measurement (publishedVersion)
- Funders
- National Natural Science Foundation of China, China (51875138, 52275099)
- Topics
- Physical Sciences › Engineering › Civil and Structural Engineering › Structural Health Monitoring Techniques
- Physical Sciences › Engineering › Control and Systems Engineering › Machine Fault Diagnosis Techniques
- Physical Sciences › Engineering › Civil and Structural Engineering › Infrastructure Maintenance and Monitoring
- Keywords
- structural health monitoring, lightweight neural networks, sensor fusion, knowledge distillation, model compression, joint loss function, resource-constrained platforms, fault diagnosis
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.