Pub Date : 2026-02-13DOI: 10.1016/j.ymssp.2026.113953
Ziyang Zheng, Zhaoqiang Wang, Changhua Hu, Penghua Li, Jie Hou, Qian Xiang, Zhichao Feng, Can Li
The domain shift issue is often encountered in prognostics and health management (PHM) domain due to variable operating conditions, load fluctuations, environmental changes, etc. A common limitation lying in the existing transfer learning methods for remaining useful life (RUL) prediction is their reliance on a rigid offline training paradigm, which leads to distinct challenges for two commonly used transfer learning methods: the domain generalization (DG) methods suffer from unreliable performance, while unsupervised domain adaptation (UDA) methods are constrained by high- and costly-data requirements. To overcome these limitations, we propose a prototype-routed multi-source unsupervised domain adaptation framework via online fine-tuning (PR-OFT) for RUL prediction in this paper. Specifically, a prototype-routed degradation stage identification (PDSI) model is designed firstly, which integrates an efficient Mamba-based feature extractor with a novel prototype-guided supervised contrastive learning strategy to precisely identify the health state stage of target sample in real time. Based on the online identified stage, a candidate knowledge base is then dynamically constructed and enriched via a pseudo-domain augmentation strategy. Furthermore, the framework precisely routes and matches the optimal knowledge for the target sample, executing a one-time and distribution-aligned online fine-tuning to instantly generate a personalized predictor. The proposed PR-OFT framework follows a novel paradigm of generalizing at training and adapting at testing, i.e., it requires no exposure to target domain data during training and dynamically constructs personalized predictors for individual unlabeled target samples arriving in a data stream during inference. The proposed PR-OFT method is experimentally validated on two publicly available bearing datasets as well as a hard disk drive dataset, where the proposed method consistently outperformed several state-of-the-art methods, demonstrating its outstanding predictive performance and strong generalization capability.
{"title":"Prototype-routed multi-source unsupervised domain adaptation framework for RUL prediction via online fine-tuning","authors":"Ziyang Zheng, Zhaoqiang Wang, Changhua Hu, Penghua Li, Jie Hou, Qian Xiang, Zhichao Feng, Can Li","doi":"10.1016/j.ymssp.2026.113953","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.113953","url":null,"abstract":"The domain shift issue is often encountered in prognostics and health management (PHM) domain due to variable operating conditions, load fluctuations, environmental changes, etc. A common limitation lying in the existing transfer learning methods for remaining useful life (RUL) prediction is their reliance on a rigid offline training paradigm, which leads to distinct challenges for two commonly used transfer learning methods: the domain generalization (DG) methods suffer from unreliable performance, while unsupervised domain adaptation (UDA) methods are constrained by high- and costly-data requirements. To overcome these limitations, we propose a prototype-routed multi-source unsupervised domain adaptation framework via online fine-tuning (PR-OFT) for RUL prediction in this paper. Specifically, a prototype-routed degradation stage identification (PDSI) model is designed firstly, which integrates an efficient Mamba-based feature extractor with a novel prototype-guided supervised contrastive learning strategy to precisely identify the health state stage of target sample in real time. Based on the online identified stage, a candidate knowledge base is then dynamically constructed and enriched via a pseudo-domain augmentation strategy. Furthermore, the framework precisely routes and matches the optimal knowledge for the target sample, executing a one-time and distribution-aligned online fine-tuning to instantly generate a personalized predictor. The proposed PR-OFT framework follows a novel paradigm of generalizing at training and adapting at testing, i.e., it requires no exposure to target domain data during training and dynamically constructs personalized predictors for individual unlabeled target samples arriving in a data stream during inference. The proposed PR-OFT method is experimentally validated on two publicly available bearing datasets as well as a hard disk drive dataset, where the proposed method consistently outperformed several state-of-the-art methods, demonstrating its outstanding predictive performance and strong generalization capability.","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"23 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146208639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-12DOI: 10.1016/j.ymssp.2026.114020
Taro Yaoyama, Tatsuya Itoi, Jun Iyama
{"title":"Hierarchical Bayesian model updating using Dirichlet process mixtures for structural damage localization","authors":"Taro Yaoyama, Tatsuya Itoi, Jun Iyama","doi":"10.1016/j.ymssp.2026.114020","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.114020","url":null,"abstract":"","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"316 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146160985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-12DOI: 10.1016/j.ymssp.2026.114001
Weicheng Wang, Huan Wang, Xiao Cai, Min Xie
Under time-varying speeds, rotating machinery experiences pronounced dynamic effects, making this operational phase highly susceptible to faults. Anomaly detection under such non-stationary condition is challenging, as speed-induced distribution shifts in vibration signals often lead conventional methods to mistake normal variations for faults, resulting in false alarms. Although data driven methods aim to learn speed-invariant representations, its generalization to unseen speed profiles remains limited. To address this, we propose a physics-informed Order-Adaptive Subspace Scale Learning (OASSL) method that integrates order tracking grounded in the physical relationship between vibration harmonics and shaft rotation. This approach resamples time-domain signals into the angular domain to generate unified order-frequency features, considerably eliminating the influence of speed fluctuations. Furthermore, a novel multi-subspace sampling and subspace scale learning strategy is introduced within the network, which enhances the extraction of subtle fault signatures and improves robustness against varying operating conditions. Experimental results on time-varying speed datasets demonstrate that the proposed OASSL significantly outperforms existing methods in reducing false alarms and accurately identifying faults under complex speed variations.
{"title":"Order-adaptive subspace scale learning for unsupervised anomaly detection under time-varying rotational speed conditions","authors":"Weicheng Wang, Huan Wang, Xiao Cai, Min Xie","doi":"10.1016/j.ymssp.2026.114001","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.114001","url":null,"abstract":"Under time-varying speeds, rotating machinery experiences pronounced dynamic effects, making this operational phase highly susceptible to faults. Anomaly detection under such non-stationary condition is challenging, as speed-induced distribution shifts in vibration signals often lead conventional methods to mistake normal variations for faults, resulting in false alarms. Although data driven methods aim to learn speed-invariant representations, its generalization to unseen speed profiles remains limited. To address this, we propose a physics-informed Order-Adaptive Subspace Scale Learning (OASSL) method that integrates order tracking grounded in the physical relationship between vibration harmonics and shaft rotation. This approach resamples time-domain signals into the angular domain to generate unified order-frequency features, considerably eliminating the influence of speed fluctuations. Furthermore, a novel multi-subspace sampling and subspace scale learning strategy is introduced within the network, which enhances the extraction of subtle fault signatures and improves robustness against varying operating conditions. Experimental results on time-varying speed datasets demonstrate that the proposed OASSL significantly outperforms existing methods in reducing false alarms and accurately identifying faults under complex speed variations.","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"36 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146208806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-12DOI: 10.1016/j.ymssp.2026.113997
Qiyuan Zhu, Hongjun Xiang
{"title":"Inerter-enhanced piezoelectric energy harvesting for vehicle-induced bridge vibrations: Analytical modeling and optimal parameter design","authors":"Qiyuan Zhu, Hongjun Xiang","doi":"10.1016/j.ymssp.2026.113997","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.113997","url":null,"abstract":"","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"6 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146160988","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Structural displacement monitoring is essential for ensuring the safety and longevity of civil infrastructures. Millimeter-wave radar offers high-precision and non-contact measurements but suffers from phase-wrapping issues. Existing methods perform unwrapping alone, limiting performance under high-noise conditions. This study proposes an acceleration-aided Kalman filtering method for joint phase denoising and unwrapping in FMCW millimeter-wave radar displacement monitoring. The radar phase is modeled as a discrete-time constant-acceleration system, with measured acceleration incorporated to enhance dynamic tracking and suppress high-frequency noise. A predictive phase-correction step removes phase discontinuities, and recursive Kalman updates produce a continuous and noise-suppressed phase trajectory, which can be converted to displacement. Noise parameters of the Kalman filter are adaptively optimized, and the Kalman recursion is further expressed in convolution form, enabling analysis of minimum convergence time for robust phase unwrapping. The effectiveness of the proposed method was validated through experimental testing on a four-story building model and a 45 m long pedestrian bridge under multiple excitation scenarios. Under high-noise conditions, the existing methods performed poorly, whereas the proposed method completed phase unwrapping and denoising in both laboratory and field bridge tests, achieving errors of 2.1 rad and 0.2 rad, respectively.
{"title":"Acceleration-aided Kalman filtering for joint phase denoising and unwrapping in FMCW radar-based displacement monitoring","authors":"Zhanxiong Ma, Tongtong Zhang, Yang Zhu, Shuhan Lin, Jigu Lee, Hoon Sohn, Qiangqiang Zhang","doi":"10.1016/j.ymssp.2026.113991","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.113991","url":null,"abstract":"Structural displacement monitoring is essential for ensuring the safety and longevity of civil infrastructures. Millimeter-wave radar offers high-precision and non-contact measurements but suffers from phase-wrapping issues. Existing methods perform unwrapping alone, limiting performance under high-noise conditions. This study proposes an acceleration-aided Kalman filtering method for joint phase denoising and unwrapping in FMCW millimeter-wave radar displacement monitoring. The radar phase is modeled as a discrete-time constant-acceleration system, with measured acceleration incorporated to enhance dynamic tracking and suppress high-frequency noise. A predictive phase-correction step removes phase discontinuities, and recursive Kalman updates produce a continuous and noise-suppressed phase trajectory, which can be converted to displacement. Noise parameters of the Kalman filter are adaptively optimized, and the Kalman recursion is further expressed in convolution form, enabling analysis of minimum convergence time for robust phase unwrapping. The effectiveness of the proposed method was validated through experimental testing on a four-story building model and a 45 m long pedestrian bridge under multiple excitation scenarios. Under high-noise conditions, the existing methods performed poorly, whereas the proposed method completed phase unwrapping and denoising in both laboratory and field bridge tests, achieving errors of 2.1 rad and 0.2 rad, respectively.","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"22 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146208808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-11DOI: 10.1016/j.ymssp.2026.113996
Shuang Liang, Qingyu Meng, Chuanhai Chen, Zhifeng Liu, Bing Guo, Bin Shen, Kuo Liu, Hongyan Guo
{"title":"Dynamic stability enhancement of weak stiffness grinding system through microstructure-induced spatial force regulation","authors":"Shuang Liang, Qingyu Meng, Chuanhai Chen, Zhifeng Liu, Bing Guo, Bin Shen, Kuo Liu, Hongyan Guo","doi":"10.1016/j.ymssp.2026.113996","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.113996","url":null,"abstract":"","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"91 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146160996","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-11DOI: 10.1016/j.ymssp.2026.114013
Johannes D. Schmid, Sebastian F. Zettel, Steffen Marburg
{"title":"Physics-informed neural operators for predicting structural intensity from laser Doppler vibrometry measurements of plates","authors":"Johannes D. Schmid, Sebastian F. Zettel, Steffen Marburg","doi":"10.1016/j.ymssp.2026.114013","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.114013","url":null,"abstract":"","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"22 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146160995","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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 : 2026-02-11DOI: 10.1016/j.ymssp.2026.113998
Zhuhong Wang, Hang Zhou, Hanlong Liu
Accurate measurement of three-dimensional deformation behavior is critical for understanding material mechanical properties. However, traditional Digital Volume Correlation (DVC) methods are limited by discrete sub-volume discretization, lack of physical constraints, and low computational efficiency. Data-driven approaches cannot guarantee physical plausibility and depend on large quantities of densely sampled data. This study proposes a novel physics-informed deep learning method for DVC (PiNetDVC). The method takes spatial coordinates as inputs and simultaneously predicts displacement and strain fields through continuous function representation, overcoming spatial resolution limitations and data dependency. The strain field is directly incorporated as a network output, with strain–displacement compatibility enforced by comparing network-predicted strains with strains derived from displacement gradients. A unified loss function integrates image consistency constraints with physics-informed regularization. Validation on six scenarios demonstrates superior performance over traditional ALDVC, achieving accuracy improvements of 81%, 83%, and over 95% for rigid body translation, uniaxial tension, and shear band deformation, respectively. For complex deformation patterns such as sinusoidal and non-uniform star-shaped modes, errors are maintained at the order of 10-3. Stable accuracy is maintained under 20 dB noise, with robust performance across different architectures and loss configurations. PiNetDVC provides an effective solution for 3D deformation measurement in aerospace, mechanical, and civil engineering applications.
{"title":"Physics-informed neural networks based digital volume correlation for displacement and strain measurements","authors":"Zhuhong Wang, Hang Zhou, Hanlong Liu","doi":"10.1016/j.ymssp.2026.113998","DOIUrl":"10.1016/j.ymssp.2026.113998","url":null,"abstract":"<div><div>Accurate measurement of three-dimensional deformation behavior is critical for understanding material mechanical properties. However, traditional Digital Volume Correlation (DVC) methods are limited by discrete sub-volume discretization, lack of physical constraints, and low computational efficiency. Data-driven approaches cannot guarantee physical plausibility and depend on large quantities of densely sampled data. This study proposes a novel physics-informed deep learning method for DVC (PiNetDVC). The method takes spatial coordinates as inputs and simultaneously predicts displacement and strain fields through continuous function representation, overcoming spatial resolution limitations and data dependency. The strain field is directly incorporated as a network output, with strain–displacement compatibility enforced by comparing network-predicted strains with strains derived from displacement gradients. A unified loss function integrates image consistency constraints with physics-informed regularization. Validation on six scenarios demonstrates superior performance over traditional ALDVC, achieving accuracy improvements of 81%, 83%, and over 95% for rigid body translation, uniaxial tension, and shear band deformation, respectively. For complex deformation patterns such as sinusoidal and non-uniform star-shaped modes, errors are maintained at the order of 10<sup>-3</sup>. Stable accuracy is maintained under 20 dB noise, with robust performance across different architectures and loss configurations. PiNetDVC provides an effective solution for 3D deformation measurement in aerospace, mechanical, and civil engineering applications.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"248 ","pages":"Article 113998"},"PeriodicalIF":8.9,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146147128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Model updating of structures by combining reduced order modelling and deep reinforcement learning","authors":"Gianluca Bruno, Fabio Parisi, Sergio Ruggieri, Eleni Chatzi, Giuseppina Uva","doi":"10.1016/j.ymssp.2026.114002","DOIUrl":"https://doi.org/10.1016/j.ymssp.2026.114002","url":null,"abstract":"","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"3 1","pages":""},"PeriodicalIF":8.4,"publicationDate":"2026-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146160994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：