Pub Date : 2026-05-05Epub Date: 2026-03-07DOI: 10.1016/j.measurement.2026.121077
Jia Li , Zhilei Yuan , Pinghua Xu , Wenhui Shi , Lan Yao
Characterizing large-strain, anisotropic fabric deformation is challenging due to complex texture evolution and localized strain gradients. This study proposes a Fabric-Adaptive Digital Image Correlation (FA-DIC) framework to address these issues. FA-DIC integrates three key innovations: a hierarchical multi-scale strategy to resolve texture-related ambiguities, an adaptive regularization driven by strain gradients to preserve local features while suppressing noise, and an anisotropy-aware strain computation that incorporates fabric principal directions to correct isotropic bias. Validated through synthetic and experimental uniaxial tension tests on woven and knitted fabrics, FA-DIC demonstrates superior performance over reference methods, delivering more consistent full-field strain maps with reduced displacement errors. The framework provides a reliable and robust approach for the mechanical characterization of soft, deformable materials.
{"title":"A multi-scale digital image correlation framework for large-strain fabric deformation measurement","authors":"Jia Li , Zhilei Yuan , Pinghua Xu , Wenhui Shi , Lan Yao","doi":"10.1016/j.measurement.2026.121077","DOIUrl":"10.1016/j.measurement.2026.121077","url":null,"abstract":"<div><div>Characterizing large-strain, anisotropic fabric deformation is challenging due to complex texture evolution and localized strain gradients. This study proposes a Fabric-Adaptive Digital Image Correlation (FA-DIC) framework to address these issues. FA-DIC integrates three key innovations: a hierarchical multi-scale strategy to resolve texture-related ambiguities, an adaptive regularization driven by strain gradients to preserve local features while suppressing noise, and an anisotropy-aware strain computation that incorporates fabric principal directions to correct isotropic bias. Validated through synthetic and experimental uniaxial tension tests on woven and knitted fabrics, FA-DIC demonstrates superior performance over reference methods, delivering more consistent full-field strain maps with reduced displacement errors. The framework provides a reliable and robust approach for the mechanical characterization of soft, deformable materials.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121077"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388167","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-02-09DOI: 10.1016/j.measurement.2026.120570
Hisham Alghamdi , Ghulam Hafeez , Safeer Ullah , Ahmed S. Alsafran , Baheej Alghamdi , Muhyaddin Rawa , Sultan Alghamdi , Habib Kraiem
Accurate short-term electricity demand forecasting is essential for reliable grid operation, economic dispatch, and energy planning under increasingly variable consumption patterns. Conventional deep learning approaches often employ monolithic stacked architectures that offer limited transparency and may struggle to generalize across different forecasting horizons and demand profiles. To address these limitations, this paper proposes a cascaded Factored Conditional Deep Learning framework for short-term load forecasting. The proposed framework structurally decomposes the forecasting task into sequential learning stages, where probabilistic feature extraction, temporal dependency modeling, and output refinement are handled by functionally distinct yet interconnected parts. Unlike conventional deep belief networks, the Factored Conditional Deep Belief Network (FCDBN) introduces factorized conditional interactions combined with Gibbs sampling—based refinement, enabling more stable learning, improved generalization, and enhanced interpretability of temporal dynamics. The effectiveness of the proposed approach is validated using real electricity demand data from Western European power systems, including Austria and Sweden, covering five years (January 2015–August 2025) with 15-minute and hourly resolutions, respectively. Comprehensive evaluations are conducted for day-ahead, week-ahead, and month-ahead forecasting horizons using root mean square error (RMSE), mean absolute percentage error (MAPE), and the Pearson correlation coefficient metrics for accuracy and network training time and number of epochs for convergence rate. Experimental results demonstrate that the proposed framework consistently outperforms benchmark models: artificial neural network (ANN), conditional restricted Boltzmann machine (CRBM), and long short-term memory (LSTM) in terms of forecasting accuracy, and computational efficiency.
{"title":"A cascaded factored conditional deep learning framework for electricity demand forecasting in smart grids","authors":"Hisham Alghamdi , Ghulam Hafeez , Safeer Ullah , Ahmed S. Alsafran , Baheej Alghamdi , Muhyaddin Rawa , Sultan Alghamdi , Habib Kraiem","doi":"10.1016/j.measurement.2026.120570","DOIUrl":"10.1016/j.measurement.2026.120570","url":null,"abstract":"<div><div>Accurate short-term electricity demand forecasting is essential for reliable grid operation, economic dispatch, and energy planning under increasingly variable consumption patterns. Conventional deep learning approaches often employ monolithic stacked architectures that offer limited transparency and may struggle to generalize across different forecasting horizons and demand profiles. To address these limitations, this paper proposes a cascaded Factored Conditional Deep Learning framework for short-term load forecasting. The proposed framework structurally decomposes the forecasting task into sequential learning stages, where probabilistic feature extraction, temporal dependency modeling, and output refinement are handled by functionally distinct yet interconnected parts. Unlike conventional deep belief networks, the Factored Conditional Deep Belief Network (FCDBN) introduces factorized conditional interactions combined with Gibbs sampling—based refinement, enabling more stable learning, improved generalization, and enhanced interpretability of temporal dynamics. The effectiveness of the proposed approach is validated using real electricity demand data from Western European power systems, including Austria and Sweden, covering five years (January 2015–August 2025) with 15-minute and hourly resolutions, respectively. Comprehensive evaluations are conducted for day-ahead, week-ahead, and month-ahead forecasting horizons using root mean square error (RMSE), mean absolute percentage error (MAPE), and the Pearson correlation coefficient metrics for accuracy and network training time and number of epochs for convergence rate. Experimental results demonstrate that the proposed framework consistently outperforms benchmark models: artificial neural network (ANN), conditional restricted Boltzmann machine (CRBM), and long short-term memory (LSTM) in terms of forecasting accuracy, and computational efficiency.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 120570"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388279","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-03-05DOI: 10.1016/j.measurement.2026.120974
Yan Zhou , Rui-Pin Chen , Linsheng Chen , Wen Zhang , Huaping Gong , Changyu Shen , Yang Zhang , Wenjun Zhou
The ability to real-time sense the dynamic reactions is particularly critical in a wide range of applications, such as material formation process and biochemistry development. In this paper, a fiber-optic sensing platform based on a tilted fiber Bragg grating (TFBG) with multiresonance combs is proposed to real-time monitor the spectral dynamics in particle suspension. Utilizing the dense comb-like resonance spectrum of the TFBG, dynamic temporal evolution is fully observed by tracking the responses in liquid, which mainly progresses through the first phase dominated by bulk refractive index (RI), then transitions to the second phase modulated by the gradual scattering effect, and eventually to an equilibrium state. This method enables detection of bulk refractive index changes and particle induced scattering effects, where the subtle changes are successfully captured continuously in an evolving suspension. This work highlights the unique capability of multiresonant fiber grating to resolve real-time monitoring in particulate suspensions, paving the way for applications in process control under complex liquid environments.
{"title":"Real-time monitoring of spectral dynamics through multiresonant fiber grating in particle suspension","authors":"Yan Zhou , Rui-Pin Chen , Linsheng Chen , Wen Zhang , Huaping Gong , Changyu Shen , Yang Zhang , Wenjun Zhou","doi":"10.1016/j.measurement.2026.120974","DOIUrl":"10.1016/j.measurement.2026.120974","url":null,"abstract":"<div><div>The ability to real-time sense the dynamic reactions is particularly critical in a wide range of applications, such as material formation process and biochemistry development. In this paper, a fiber-optic sensing platform based on a tilted fiber Bragg grating (TFBG) with multiresonance combs is proposed to real-time monitor the spectral dynamics in particle suspension. Utilizing the dense comb-like resonance spectrum of the TFBG, dynamic temporal evolution is fully observed by tracking the responses in liquid, which mainly progresses through the first phase dominated by bulk refractive index (RI), then transitions to the second phase modulated by the gradual scattering effect, and eventually to an equilibrium state. This method enables detection of bulk refractive index changes and particle induced scattering effects, where the subtle changes are successfully captured continuously in an evolving suspension. This work highlights the unique capability of multiresonant fiber grating to resolve real-time monitoring in particulate suspensions, paving the way for applications in process control under complex liquid environments.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 120974"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388352","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-03-05DOI: 10.1016/j.measurement.2026.121033
Ambuj , P. Jayraj , Agnibha Basak , Rajendra Machavaram
Low-cost optical (fluorescence) dissolved oxygen (DO) sensors exhibit significant nonlinear biases that vary with temperature and pressure, thereby limiting their standalone accuracy. This work proposes a hybrid framework that integrates Gaussian Process Regression (GPR) for probabilistic bias correction with an Extended Kalman Filter (EKF) for accurate real-time DO estimation using low-cost optical DO sensors. The framework was validated using 3,788 laboratory samples under varied temperature, pressure, and DO conditions. Performance evaluation on an independent 994-sample test dataset against a high-precision reference sensor yielded an RMSE of 0.162 mg/L (RRMSE ≈ 2.1%) and an MAE of 0.101 mg/L, corresponding to an 84.5% reduction in RMSE relative to raw low-cost sensor readings. Statistical consistency was confirmed with a mean NEES of 0.96 and white innovations. The proposed approach effectively tracked rapid deoxygenation transients and executed in < 4 ms per update, enabling real-time embedded deployment for scalable, near-reference-grade DO monitoring with reliable uncertainty awareness.
{"title":"A hybrid machine learning and Extended Kalman filtering framework for sensor fusion of low-cost and high-precision dissolved oxygen sensors under environmental variability","authors":"Ambuj , P. Jayraj , Agnibha Basak , Rajendra Machavaram","doi":"10.1016/j.measurement.2026.121033","DOIUrl":"10.1016/j.measurement.2026.121033","url":null,"abstract":"<div><div>Low-cost optical (fluorescence) dissolved oxygen (DO) sensors exhibit significant nonlinear biases that vary with temperature and pressure, thereby limiting their standalone accuracy. This work proposes a hybrid framework that integrates Gaussian Process Regression (GPR) for probabilistic bias correction with an Extended Kalman Filter (EKF) for accurate real-time DO estimation using low-cost optical DO sensors. The framework was validated using 3,788 laboratory samples under varied temperature, pressure, and DO conditions. Performance evaluation on an independent 994-sample test dataset against a high-precision reference sensor yielded an RMSE of 0.162 mg/L (RRMSE ≈ 2.1%) and an MAE of 0.101 mg/L, corresponding to an 84.5% reduction in RMSE relative to raw low-cost sensor readings. Statistical consistency was confirmed with a mean NEES of 0.96 and white innovations. The proposed approach effectively tracked rapid deoxygenation transients and executed in < 4 ms per update, enabling real-time embedded deployment for scalable, near-reference-grade DO monitoring with reliable uncertainty awareness.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121033"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388355","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the advancement of industrial intelligence, hydraulic systems are typically equipped with multiple sensor types to monitor critical operating parameters. However, multi-source signals exhibit significant differences in modal characteristics, distribution patterns, and structural correlations, making it challenging for traditional fusion-based diagnostic methods to effectively capture cross-modal associations and thereby limiting diagnostic performance. To address this, this paper proposes a multi-source signal fusion diagnostic method based on graph neural networks. First, three graph construction methods—including neighborhood similarity—establish topological relationships. Second, a three-layer graph convolutional framework with residual structures is designed to enhance structural information representation across different levels through multi-scale graph-level feature fusion. Finally, a lightweight gated attention module is introduced to select key discriminative features, thereby improving the effectiveness of multi-source fusion representation. Experimental results demonstrate that this method achieves over 98% accuracy in hydraulic pump fault diagnosis and maintains approximately 80% accuracy under high-intensity noise conditions, exhibiting robust performance and significant application potential.
{"title":"Fault diagnosis method for hydraulic pumps based on multi-source signal fusion using graph neural networks","authors":"Yonghui Zhao , Anqi Jiang , Wanlu Jiang , Enyu Tang , Mengda Zhang , Shuaiyang Ma","doi":"10.1016/j.measurement.2026.121020","DOIUrl":"10.1016/j.measurement.2026.121020","url":null,"abstract":"<div><div>With the advancement of industrial intelligence, hydraulic systems are typically equipped with multiple sensor types to monitor critical operating parameters. However, multi-source signals exhibit significant differences in modal characteristics, distribution patterns, and structural correlations, making it challenging for traditional fusion-based diagnostic methods to effectively capture cross-modal associations and thereby limiting diagnostic performance. To address this, this paper proposes a multi-source signal fusion diagnostic method based on graph neural networks. First, three graph construction methods—including neighborhood similarity—establish topological relationships. Second, a three-layer graph convolutional framework with residual structures is designed to enhance structural information representation across different levels through multi-scale graph-level feature fusion. Finally, a lightweight gated attention module is introduced to select key discriminative features, thereby improving the effectiveness of multi-source fusion representation. Experimental results demonstrate that this method achieves over 98% accuracy in hydraulic pump fault diagnosis and maintains approximately 80% accuracy under high-intensity noise conditions, exhibiting robust performance and significant application potential.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121020"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147387727","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-02-28DOI: 10.1016/j.measurement.2026.120994
Łukasz Drużyński , Grzegorz Dombek , Andrzej Książkiewicz
This paper investigates the impact of current waveform distortion on apparent power losses in miniature circuit breakers (MCBs). The increasing penetration of nonlinear loads in low-voltage installations results in current waveforms with significant harmonic content, which may substantially affect the electrical and thermal behaviour of protective devices. An experimental methodology for frequency-resolved assessment of power losses in MCB current paths is presented, based on synchronized measurements of RMS voltage drop and RMS current under controlled harmonic excitation. Measurements were performed on MCBs with different rated currents, subjected to individual current harmonics up to the 25th order (1250 Hz) and to composite distorted waveforms representative of industrial and office installations. The results show a clear frequency-dependent increase in apparent power losses. Depending on the breaker rating, the measured losses increase by approximately 58–64% when comparing operation at the fundamental frequency (50 Hz) with higher-order harmonics within the investigated range. Devices with higher rated currents exhibit a steeper growth of losses with increasing harmonic order. The obtained results indicate that harmonic-rich currents significantly increase the thermal loading of MCB current paths, even when the RMS current value is maintained at a constant level. The study emphasizes the importance of accounting for frequency-dependent impedance effects and waveform distortion when evaluating power losses and thermal performance of miniature circuit breakers in modern low-voltage power systems.
{"title":"Frequency-Resolved Measurement of Power losses in Miniature Circuit Breakers (MCBs) under harmonic current","authors":"Łukasz Drużyński , Grzegorz Dombek , Andrzej Książkiewicz","doi":"10.1016/j.measurement.2026.120994","DOIUrl":"10.1016/j.measurement.2026.120994","url":null,"abstract":"<div><div>This paper investigates the impact of current waveform distortion on apparent power losses in miniature circuit breakers (MCBs). The increasing penetration of nonlinear loads in low-voltage installations results in current waveforms with significant harmonic content, which may substantially affect the electrical and thermal behaviour of protective devices. An experimental methodology for frequency-resolved assessment of power losses in MCB current paths is presented, based on synchronized measurements of RMS voltage drop and RMS current under controlled harmonic excitation. Measurements were performed on MCBs with different rated currents, subjected to individual current harmonics up to the 25th order (1250 Hz) and to composite distorted waveforms representative of industrial and office installations. The results show a clear frequency-dependent increase in apparent power losses. Depending on the breaker rating, the measured losses increase by approximately 58–64% when comparing operation at the fundamental frequency (50 Hz) with higher-order harmonics within the investigated range. Devices with higher rated currents exhibit a steeper growth of losses with increasing harmonic order. The obtained results indicate that harmonic-rich currents significantly increase the thermal loading of MCB current paths, even when the RMS current value is maintained at a constant level. The study emphasizes the importance of accounting for frequency-dependent impedance effects and waveform distortion when evaluating power losses and thermal performance of miniature circuit breakers in modern low-voltage power systems.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 120994"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147387988","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-03-10DOI: 10.1016/j.measurement.2026.121114
Caohongtai Liu, Tongxin Yan, Yuting Zhang
The expansion of the transportation industry has not only facilitated the flow of people and goods, but also brought security and regulatory challenges. To this end, we propose DEFSR-Net, a novel X-ray contraband detection network with a dual backbone and detection transformer structure. The network integrates image processing and object detection through multi-task joint learning. To enhance the visibility of key features in X-ray images, we adopt a dual-branch structure for collaborative learning. First, we use the Real-ESRGAN super-resolution enhancement dataset and apply a decolorization algorithm to highlight color information. Then, an improved edge enhancement module is used to emphasize edge features and a branched backbone is combined to capture various feature types. Then, we introduce an edge-guided feature fusion module to merge features from different stages of the dual backbone, thereby effectively enhancing multi-scale feature representation and edge receptive field. To address the class imbalance problem, we use Unified-IoU for weight distribution and an annealing strategy to balance training. Extensive experiments on the EDS and CLCXray dataset confirm that DEFSR-Net is suitable for real-time deployment and has high accuracy.
{"title":"DEFSR-Net: A joint learning network of super-resolution enhanced dual-branch edge features for X-ray contraband security detection in energy dispersive spectrometer environment","authors":"Caohongtai Liu, Tongxin Yan, Yuting Zhang","doi":"10.1016/j.measurement.2026.121114","DOIUrl":"10.1016/j.measurement.2026.121114","url":null,"abstract":"<div><div>The expansion of the transportation industry has not only facilitated the flow of people and goods, but also brought security and regulatory challenges. To this end, we propose DEFSR-Net, a novel X-ray contraband detection network with a dual backbone and detection transformer structure. The network integrates image processing and object detection through multi-task joint learning. To enhance the visibility of key features in X-ray images, we adopt a dual-branch structure for collaborative learning. First, we use the Real-ESRGAN super-resolution enhancement dataset and apply a decolorization algorithm to highlight color information. Then, an improved edge enhancement module is used to emphasize edge features and a branched backbone is combined to capture various feature types. Then, we introduce an edge-guided feature fusion module to merge features from different stages of the dual backbone, thereby effectively enhancing multi-scale feature representation and edge receptive field. To address the class imbalance problem, we use Unified-IoU for weight distribution and an annealing strategy to balance training. Extensive experiments on the EDS and CLCXray dataset confirm that DEFSR-Net is suitable for real-time deployment and has high accuracy.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121114"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388098","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-03-07DOI: 10.1016/j.measurement.2026.121078
Zeming Zhang , Chuanyang Li , Changhua Hu , Jianhui Hu , Meng Zhao , Mingzhe Leng , Zhaoqiang Wang , Xinyi Wan , Ziyang Zheng
Electric servo mechanisms are critical actuators in aerospace equipment, where emerging fault types continuously appear under varying operating conditions. Few-shot class-incremental diagnosis is therefore constrained by two coupled challenges: highly nonstationary vibration responses and extremely limited labelled samples, both of which weaken temporal representation learning and aggravate catastrophic forgetting in conventional models. To address these issues, a Liquid-Transformer Temporal Self-Supervised Network (LTTSNet) is proposed for continual recognition of both base and emerging fault classes. Its backbone integrates a Liquid Neural Network (LNN) with a Lightweight Transformer (LTransformer), in which the LNN captures local transient dynamics through learnable time constants, whereas the LTransformer models long-range cross-cycle dependencies. In the base stage, a simple framework for contrastive learning of visual representations is employed to learn invariant representations from scarce unlabelled signals by contrasting augmented views. Pseudo-labels are generated via nearest-neighbour clustering under a cosine-similarity threshold and are jointly trained with labelled samples, thereby introducing latent new-class information before incremental updates. In the incremental stage, attention-weighted prototype estimation and one-step gradient prototype distillation are jointly employed to refine new-class prototypes. The backbone is kept frozen, and only the classifier head is updated, enabling rapid adaptation to new classes while preserving old-class discrimination. Experiments on a laboratory electric servo mechanism fault dataset and the Case Western Reserve University bearing dataset demonstrate that LTTSNet delivers significantly improves overall accuracy, new-class recognition, and forgetting suppression under cross-condition few-shot settings with both single and compound faults.
{"title":"Liquid-transformer temporal self-supervised network for few-shot class-incremental fault diagnosis of servo mechanisms","authors":"Zeming Zhang , Chuanyang Li , Changhua Hu , Jianhui Hu , Meng Zhao , Mingzhe Leng , Zhaoqiang Wang , Xinyi Wan , Ziyang Zheng","doi":"10.1016/j.measurement.2026.121078","DOIUrl":"10.1016/j.measurement.2026.121078","url":null,"abstract":"<div><div>Electric servo mechanisms are critical actuators in aerospace equipment, where emerging fault types continuously appear under varying operating conditions. Few-shot class-incremental diagnosis is therefore constrained by two coupled challenges: highly nonstationary vibration responses and extremely limited labelled samples, both of which weaken temporal representation learning and aggravate catastrophic forgetting in conventional models. To address these issues, a Liquid-Transformer Temporal Self-Supervised Network (LTTSNet) is proposed for continual recognition of both base and emerging fault classes. Its backbone integrates a Liquid Neural Network (LNN) with a Lightweight Transformer (LTransformer), in which the LNN captures local transient dynamics through learnable time constants, whereas the LTransformer models long-range cross-cycle dependencies. In the base stage, a simple framework for contrastive learning of visual representations is employed to learn invariant representations from scarce unlabelled signals by contrasting augmented views. Pseudo-labels are generated via nearest-neighbour clustering under a cosine-similarity threshold and are jointly trained with labelled samples, thereby introducing latent new-class information before incremental updates. In the incremental stage, attention-weighted prototype estimation and one-step gradient prototype distillation are jointly employed to refine new-class prototypes. The backbone is kept frozen, and only the classifier head is updated, enabling rapid adaptation to new classes while preserving old-class discrimination. Experiments on a laboratory electric servo mechanism fault dataset and the Case Western Reserve University bearing dataset demonstrate that LTTSNet delivers significantly improves overall accuracy, new-class recognition, and forgetting suppression under cross-condition few-shot settings with both single and compound faults.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121078"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388102","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The objective of this study is to extract sequence components in a distribution system utilizing hardware-in-loop (HIL) for accurate fault detection and classification. The proposed system focuses on real-time monitoring using fixed-point arithmetic on a low-cost TMS320F28379D launch-xl DSP microcontroller, enabling quick computation and fault detection based on the sequence parameters. This study proposes a Fourier-based modified extraction technique for a three-phase induction motor that serves as a prototype of a distribution system. The real-time sequence component extraction of the voltage and current signals in the transmission lines was carried out using a Fourier-based extraction technique. The amplitude and phase of all the sequence components were successfully extracted with variable sampling frequencies. The system performance was successfully tested for both symmetrical and unsymmetrical faults in transmission system within three to five power cycles, as per the IEEE C37.103 standard for overcurrent protection in transmission lines. This research effectively illustrated real-time sequence extraction, enabling a rapid reaction to imbalances in the system. Reliable sequence component extraction is made possible by the use of HIL, which makes it easier to track changes in real time. Optimization of the Fourier-based extraction algorithm improves the overall execution speed and reduces the computational burden and memory utilization of the DSP. The algorithm can be deployed on low-cost target DSP platforms for HIL testing. Furthermore, the system is easily scalable and adaptable, with minimal changes to meet the requirements of changing physical conditions.
{"title":"Real-time sequence components extraction for an unbalanced distribution system for fault detection","authors":"Sushil Karvekar, Jayesh Kharat, Harshwardhan Khot, Prathmesh Gadkari","doi":"10.1016/j.measurement.2026.121048","DOIUrl":"10.1016/j.measurement.2026.121048","url":null,"abstract":"<div><div>The objective of this study is to extract sequence components in a distribution system utilizing hardware-in-loop <strong>(</strong>HIL) for accurate fault detection and classification. The proposed system focuses on real-time monitoring using fixed-point arithmetic on a low-cost TMS320F28379D launch-xl DSP microcontroller, enabling quick computation and fault detection based on the sequence parameters. This study proposes a Fourier-based modified extraction technique for a three-phase induction motor that serves as a prototype of a distribution system. The real-time sequence component extraction of the voltage and current signals in the transmission lines was carried out using a Fourier-based extraction technique. The amplitude and phase of all the sequence components were successfully extracted with variable sampling frequencies. The system performance was successfully tested for both symmetrical and unsymmetrical faults in transmission system within three to five power cycles, as per the IEEE C37.103 standard for overcurrent protection in transmission lines. This research effectively illustrated real-time sequence extraction, enabling a rapid reaction to imbalances in the system. Reliable sequence component extraction is made possible by the use of HIL, which makes it easier to track changes in real time. Optimization of the Fourier-based extraction algorithm improves the overall execution speed and reduces the computational burden and memory utilization of the DSP. The algorithm can be deployed on low-cost target DSP platforms for HIL testing. Furthermore, the system is easily scalable and adaptable, with minimal changes to meet the requirements of changing physical conditions.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121048"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-05-05Epub Date: 2026-03-06DOI: 10.1016/j.measurement.2026.121065
Zhengxiang He , Xingliang Xu , Pingan Peng , Liguan Wang , Suchuan Tian
The axial load of rock bolts serves as a vital indicator for monitoring this mechanical state. However, existing axial load measurement methods rely primarily on installing sensors on bolts, resulting in limited monitoring approaches, high costs, and poor feedback timeliness. Therefore, this paper proposes a noncontact load inversion method for rock bolts based on the integration of deep vision and deep learning. Innovatively, we establish a constitutive model that links the surface deformation of anchor plates to the axial load of bolts through deep learning. By incorporating a multiscale patch embedding block and a gated residual attention mechanism, we enhance the Vision Transformer (ViT) model, developing a multiscale gated vision transformer for load inversion computation. The proposed method was validated through laboratory experiments and field tests. In laboratory, it achieved a coefficient of determination (R2) of 0.97 for axial load prediction, outperforming the Gated Transformer (0.94), ViT (0.95), ResNet (0.92), and CNN (0.95). During the field tests, the model attained an R2 value of 0.96. Additionally, we analyzed the impact of the measurement offset at the anchor plate on the axial load inversion accuracy. The results demonstrate that the proposed noncontact method efficiently inverts the axial load of rock bolts.
{"title":"Noncontact inversion method for anchored rock bolt axial load based on deep vision–deep learning fusion","authors":"Zhengxiang He , Xingliang Xu , Pingan Peng , Liguan Wang , Suchuan Tian","doi":"10.1016/j.measurement.2026.121065","DOIUrl":"10.1016/j.measurement.2026.121065","url":null,"abstract":"<div><div>The axial load of rock bolts serves as a vital indicator for monitoring this mechanical state. However, existing axial load measurement methods rely primarily on installing sensors on bolts, resulting in limited monitoring approaches, high costs, and poor feedback timeliness. Therefore, this paper proposes a noncontact load inversion method for rock bolts based on the integration of deep vision and deep learning. Innovatively, we establish a constitutive model that links the surface deformation of anchor plates to the axial load of bolts through deep learning. By incorporating a multiscale patch embedding block and a gated residual attention mechanism, we enhance the Vision Transformer (ViT) model, developing a multiscale gated vision transformer for load inversion computation. The proposed method was validated through laboratory experiments and field tests. In laboratory, it achieved a coefficient of determination (R<sup>2</sup>) of 0.97 for axial load prediction, outperforming the Gated Transformer (0.94), ViT (0.95), ResNet (0.92), and CNN (0.95). During the field tests, the model attained an R<sup>2</sup> value of 0.96. Additionally, we analyzed the impact of the measurement offset at the anchor plate on the axial load inversion accuracy. The results demonstrate that the proposed noncontact method efficiently inverts the axial load of rock bolts.</div></div>","PeriodicalId":18349,"journal":{"name":"Measurement","volume":"272 ","pages":"Article 121065"},"PeriodicalIF":5.6,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147388109","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号