Mechanical Systems and Signal Processing最新文献

{"title":"Surrogate-aided Bayesian calibration with adaptive learning strategies","authors":"Alexandros A. Taflanidis ,&nbsp;B.S. Aakash ,&nbsp;Sang-ri Yi ,&nbsp;Joel P. Conte","doi":"10.1016/j.ymssp.2025.113014","DOIUrl":"10.1016/j.ymssp.2025.113014","url":null,"abstract":"<div><div>Bayesian inference or calibration for engineering systems requires sampling from the posterior distribution of the system model parameters. For applications with complex numerical models, the computational burden for this sampling, requiring a large number of calls to that model, can be prohibitive. For alleviating this burden, this work investigates an adaptive surrogate model implementation for approximating the numerical model predictions within the process of the Bayesian inference. Gaussian Process (GP) regression is adopted as surrogate modelling technique, while the formulation is embedded within a sequential Monte Carlo (MC) approach, using a series of intermediate auxiliary densities to efficiently derive samples from the posterior target density. The adaptive training of the GP is established by iteratively adding training points to inform the target posterior density approximation. At each iteration, a GP model is constructed using the current set of training points to approximate the system response and, consequently, the posterior distribution. This approximation is then used to facilitate sampling from the posterior. The iterative procedure terminates once the posterior density approximations from successive iterations exhibit sufficient similarity. Convergence is assessed using a combination of criteria, including the stability of the most probable parameter estimates and the similarity of the approximated posterior densities. If convergence is not established, the surrogate model is refined through an adaptive design of computer experiments (DoE) that uses the weighted integrated mean squared error as acquisition function. Through proper selection of the weights, exploitation and exploration strategies are promoted to improve, respectively, efficiency and robustness for the GP approximation. This weight selection combines information from both the target posterior density and the intermediate, auxiliary densities. Strategies are examined for improving computational efficiency for the DoE optimization and for the sequential Monte Carlo sampling once the surrogate model has been updated. Three illustrative examples are examined, considering both static and dynamic Bayesian calibration problems. Results demonstrate the computational efficiency offered through the adaptive GP formulation and provide insights for the best adaptation strategies and for the appropriate selection of response quantities to be adopted as outputs within the GP implementation.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113014"},"PeriodicalIF":7.9,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144662743","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":"Learnable wavelet-driven physically interpretable networks for bearing fault diagnosis under variable speed","authors":"Yi Qin ,&nbsp;Lijuan Zhao ,&nbsp;Yuejian Chen ,&nbsp;Dengyu Xiao ,&nbsp;Yongfang Mao","doi":"10.1016/j.ymssp.2025.113121","DOIUrl":"10.1016/j.ymssp.2025.113121","url":null,"abstract":"<div><div>Variable speed conditions pose great challenges for intelligent bearing fault diagnosis. The popular intelligent fault diagnosis models neglect the effect of rotational speed when extracting features, meantime the extracted features lack the physical meaning. To this end, a learnable wavelet-driven physically interpretable (LWPI) network is proposed for diagnosing bearing faults under variable speed. Firstly, an entropy-based local peak search (LPS) algorithm with an adaptive instantaneous frequency (IF) search range is designed to extract the rotational frequency ridge from vibration signals, and it can effectively suppress the noise and adjacent interference components. Based on the extracted rotational speed information, a learnable multi-wavelet filter layer is constructed to guide the model in adaptively mining features with physical meaning. Next, a convolution block is constructed to mine the high-dimensional features, followed by a linear dense layer designed for fault classification. Experiments on bearing fault diagnosis under variable speed conditions demonstrate that the proposed LWPI network consistently outperforms five advanced methods. Meanwhile, the efficacy of the entropy-based LPS algorithm and learnable multi-wavelet filter layer are respectively verified by the ablation experiments.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113121"},"PeriodicalIF":7.9,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144662744","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":"Residual vibration suppression of cooperative robot based on optimization surrogate model and input shaper control","authors":"Tianzuo Niu ,&nbsp;Rong Hu ,&nbsp;Shengen Wu ,&nbsp;Xiao-Jin Wan ,&nbsp;Tianshu Liu","doi":"10.1016/j.ymssp.2025.113120","DOIUrl":"10.1016/j.ymssp.2025.113120","url":null,"abstract":"<div><div>Under high-precision assembly conditions, it is necessary to consider the modal behavior of industrial robots. However, the modal characteristics of complex systems such as industrial robots are highly nonlinear, which means that it is usually cumbersome to evaluate these characteristics by mechanical experiments and finite element methods (FEM). A new Kriging, Bayesian variable selection technique &amp; Domain Adaptation (KRG-BVS-DA) surrogate model optimized by transfer learning domain adaptation is proposed to obtain a natural frequency model of arbitrary pose of an industrial robot. Firstly, the Latin Hypercube sampling (LHS) technique was used to generate 100 reliable training data sets for the surrogate model. Then, seven surrogate models with different optimization algorithms were evaluated to determine the optimal model for predicting the natural frequency. The KRG-BVS-DA surrogate model, optimized via transfer learning and domain adaptation, achieves the highest prediction accuracy of 0.969. In addition, the accuracy of the KRG-BVS-DA surrogate model under different numbers of training samples and the robustness of the KRG-BVS-DA surrogate model for the prediction effect of each order of natural frequency were studied. The prediction data set of KRG-BVS-DA prediction model was iteratively updated by adaptive weighting, and it was found that the prediction accuracy was the highest when iterated once, reaching 0.9728. However, as the number of iterations increased, the prediction accuracy gradually declined, leading to algorithm termination after three iterations. Furthermore, a gravity compensation-based controller was developed, and a simulation model of the robot’s residual vibration suppression control system was established. Through residual vibration analysis of three flexible joints, the effectiveness of the gravity compensation controller in achieving closed-loop control was demonstrated. An improved control scheme combining a negative input shaper with the gravity compensation controller was proposed. Simulation results show that the improved scheme has enhanced the suppression capability of the maximum amplitude by 81.32% and reduced the response delay by 14.03%, achieving better residual vibration suppression effects.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113120"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656992","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":"Autonomous ultrasonic imaging and crack localisation in concrete beams during the fracture process","authors":"Magdalena Rucka ,&nbsp;Monika Zielińska","doi":"10.1016/j.ymssp.2025.113105","DOIUrl":"10.1016/j.ymssp.2025.113105","url":null,"abstract":"<div><div>Monitoring the fracture process in concrete and detecting cracks at an early stage remains one of the major challenges in non-destructive testing of civil engineering structures. This paper presents a novel methodology for real-time fracture imaging in concrete structures using ultrasonic waves. The proposed approach introduces an innovative damage index that enables precise identification of both the location and onset time of crack initiation. Integrated with ultrasonic transmission tomography, this method forms the basis of an automated structural health monitoring system. To thoroughly assess the method performance, experimental validation was conducted on four concrete beams, three of plain concrete and one reinforced with a steel bar, each representing a distinct damage scenario. The accuracy and effectiveness of ultrasonic fracture imaging were verified using X-ray micro-computed tomography and digital image correlation.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113105"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657086","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":"Research on intelligent fault diagnosis of rotating machinery for edge computing platforms","authors":"Zhenli Duan ,&nbsp;Wenbo Zhang ,&nbsp;Haifeng Zhang ,&nbsp;Fengyuan Yang","doi":"10.1016/j.ymssp.2025.113101","DOIUrl":"10.1016/j.ymssp.2025.113101","url":null,"abstract":"<div><div>Fault diagnosis of edge computing devices can detect mechanical faults more quickly and avoid unnecessary economic losses. However, existing fault diagnosis model research focuses on high-performance computing platforms, and the models have disadvantages such as large network volume and parameter redundancy, which make it difficult to meet the real-time requirements of edge computing platforms. To address these issues this paper proposes a lightweight model based on the fusion of time and frequency domain features for intelligent fault diagnosis of edge devices. First, a fast Fourier transform block embedded inside the model is designed to obtain the frequency domain signals. Second, the Transformer and CNN feature extractor were designed to extract time and frequency domain features, respectively, based on the characteristics of the time and frequency domain signals, and some network model lightweighting methods are used in the design. Thirdly, the feature fusion block is designed so that the time domain and frequency domain features can be fused efficiently. Finally, the multi-domain features are classified by a classifier to output the fault diagnosis results. The experimental results show that the proposed method has excellent diagnostic accuracy and robustness on multiple datasets, and achieves more than 99.8 % diagnostic accuracy on top of the Raspberry Pi 4B, and the params of the model is only 10 K with an inference time of 68 ms, which is especially suitable for edge computing platforms with limited computational resources.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113101"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656993","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":"Dynamic load identification of metro bogie frame based on the strain signals and modal parameters","authors":"Peng Lu ,&nbsp;Linfei Han ,&nbsp;Xi Wang ,&nbsp;Di Yang ,&nbsp;Yu Hou","doi":"10.1016/j.ymssp.2025.113098","DOIUrl":"10.1016/j.ymssp.2025.113098","url":null,"abstract":"<div><div>The bogie frame is a critical component of metro vehicles, and its load boundary conditions directly impact health monitoring and structural fatigue life assessment. However, conventional quasi-static methods cannot accurately capture dynamic loads, particularly when excitation frequencies approach the frame’s natural frequencies. In this study, on-track tests were conducted, and a load identification method based on strain signals was proposed. To address localized stress effects at measurement points, a correction strategy using reference point stresses was introduced. Results indicate that elastic vibration during operation involves ten modal orders. The proposed method effectively identifies bogie frame loads under elastic vibration. Floating loads are essentially unaffected by elastic vibration, and the results of the proposed method and the quasi-static method are largely consistent. However, the quasi-static method tends to overestimate roll and torsional loads and underestimate transverse loads. Especially anisotropic motor loads influenced by third-order elastic vibration modes are significantly overestimated by the quasi-static approach.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113098"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657083","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":"Blind deconvolution based on logarithmic cyclic content for early weak fault detection in gear degradation","authors":"Jiahao Gao ,&nbsp;Youren Wang ,&nbsp;Jinglin Wang ,&nbsp;Yong Shen","doi":"10.1016/j.ymssp.2025.113081","DOIUrl":"10.1016/j.ymssp.2025.113081","url":null,"abstract":"<div><div>Blind deconvolution methods (BDMs) have proven effective for fault detection, as they can recover periodic impulses from fault signals masked by noise. However, the periodic impulse energy generated by early-stage gear faults is typically weak, making it challenging for traditional BDMs to extract fault-related features. Although the close relationship between micro-surface changes induced by gear faults and second-order cyclostationary characteristics provides a valuable strategy for early fault detection, the stochastic nature of gear fault evolution during natural degradation introduces significant interference, further obscuring weak fault characteristics and making early detection more challenging. To this end, we propose a novel method based on blind deconvolution of logarithmic cyclic content (BDLCC) for early detection of weak faults in gear degradation. First, a logarithmic cyclic content index based on the logarithmic envelope spectrum is designed as the objective function of BDM to characterize the fault characteristics. Next, we develop a filter coefficient estimation way based on a one-dimensional convolutional neural network, which updates the BDM’s inverse filter coefficients through gradient backpropagation to suppress interference components. A series of experiments using simulated and run-to-failure datasets validate the effectiveness of BDLCC in the early detection of gear faults during the natural degradation process.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113081"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657084","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":"Retrieving dispersion curve of thin-plate structure via non-diffuse ambient noise with transient excitations: Application to passive airfoil icing detection in wind tunnel","authors":"Qihang Qin,&nbsp;Xun Wang","doi":"10.1016/j.ymssp.2025.113085","DOIUrl":"10.1016/j.ymssp.2025.113085","url":null,"abstract":"<div><div>Under the diffuse field assumption that waves arrive from every direction with equal probability, the cross-correlation of ambient noise measured at two receivers converges to the system’s impulse response, which forms the foundation of passive structural health monitoring (SHM). In practice, however, transient excitations at unpredictable locations and times may appear to form a non-diffuse but strongly directional sound field such that the cross-correlation is largely contaminated. This paper thus proposes a signal processing scheme to both identify and suppress the contribution of transient excitations. As a result, the diffuse field is retrieved so that the robust passive SHM is restored. The proposed method is applied to a wind tunnel test where the flow-induced random guided wave signals are refined, by which the passive detection of airfoil icing is realized.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113085"},"PeriodicalIF":7.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657085","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":"The role of vibration-induced settling on the normal and tangential forces within a jointed structure","authors":"Arati Bhattu ,&nbsp;Yi Chun Lo ,&nbsp;Gianmarco Zara ,&nbsp;Patrick Hippold ,&nbsp;Daniel Fochler ,&nbsp;Johann Groß ,&nbsp;Maren Scheel ,&nbsp;Matthew R.W. Brake ,&nbsp;Malte Krack ,&nbsp;Erhan Ferhatoglu","doi":"10.1016/j.ymssp.2025.112994","DOIUrl":"10.1016/j.ymssp.2025.112994","url":null,"abstract":"<div><div>When testing the vibrations of jointed structures, it is common practice to perform repeated runs until the behavior stabilizes and only report the final results. The purpose of the present work is to understand the transient settling phase and to establish a correlation between residual traction (static tangential force) at the interface, preload values, and the vibration response of the jointed structure. An L-beam coupled with a cross-beam is considered, which was originally designed to study the effect of residual tractions on bolted joints, as non-unique residual tractions significantly affect the vibration response of friction-damped structures. The contact interfaces are deliberately positioned orthogonal to each other to achieve tangential-normal coupling. A side load mechanism is added to change tangential forces to reach different static equilibria in the system. It is observed that the static equilibrium varies significantly during the first few vibration tests following the assembly and is marginally affected by the designed mechanism. The settling phase is experimentally analyzed with regard to tangential friction forces, bolt preload values and amplitude-dependent modal parameters. The initial normal and tangential contact forces, observed immediately after assembly, are affected by the misalignment induced by manufacturing and wear. Subsequently, well-repeatable behavior is observed in terms of the amplitude-dependent frequency and damping ratio of the fundamental mode. The results of this research provide novel insights into the physics of how interfaces settle. Recommendations for avoiding bolt loosening are made based on the observations of this research.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 112994"},"PeriodicalIF":7.9,"publicationDate":"2025-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657079","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":"Application of the dynamic masking chirplet transform method in rotating Machinery fault diagnosis","authors":"Junxiang Wang ,&nbsp;Hongkun Li ,&nbsp;Bin Sun ,&nbsp;Chen Yang ,&nbsp;Jiayi Li ,&nbsp;Yefei Liu","doi":"10.1016/j.ymssp.2025.113116","DOIUrl":"10.1016/j.ymssp.2025.113116","url":null,"abstract":"<div><div>The chirplet transform (CT), as a mainstream time–frequency analysis (TFA) tool, is widely used in processing non-stationary vibration signals in rotating machinery. However, the traditional CT and its variants have limited reliability in handling complex signal modulation and decoupling due to their constrained time–frequency (TF) atom matching mechanism. To overcome this limitation, this paper proposes a novel TFA method called dynamic masking chirplet transform (DMCT). This method first constructs a time–frequency masking operator (TFMO) to hierarchically and adaptively modulate frequency-modulated components in different TF regions. Second, a dynamic component resolution strategy is developed, which involves dynamically iterating TFMO gradient analysis intervals to select masking time–frequency representation (MTFR) that match the analysis intervals. Finally, a TF inversion algorithm is used to achieve adaptive decoupling of multi-component signals. Validation is carried out on multi-component numerical signals, bearing fault experiments, and measured data from industrial planetary gearboxes. The results show that the proposed method achieves higher TF resolution and better noise robustness than some advanced TFA methods.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"237 ","pages":"Article 113116"},"PeriodicalIF":7.9,"publicationDate":"2025-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657081","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}