|
In this study, an effective and novel method, termed M etamodel A ssisted H ybrid of P article S warm O ptimization with G enetic A lgorithm (MA-HPSOGA), is developed to identify unknown structural dynamic parameters. The method first constructs four popular metamodels to substitute the computationally expensive numerical analysis based on the Latin hypercube sampling method and probabilistic finite element analysis, and their accuracy is assessed by R-squared. Subsequently, a suitable and low-cost metamodel is selected in combination with a hybrid optimization strategy by incorporating Genetic Algorithm (GA) into Particle Swarm Optimization (PSO). Two examples with measured vibration response data and different levels of complexity are used to verify the effectiveness and practicality of the presented method. The results showed that polynomial chaos expansion assisted HPSOGA has the highest computational efficiency and accuracy in the four coupled methods. Besides, compared to the conventional iteration-based dynamic parameter identification methods, the presented method shows an overwhelming advantage in terms of computational efficiency. Furthermore, the performance of HPSOGA is compared with its sub-algorithms, showing that the hybrid strategy offers faster convergence and stronger robustness. Our findings reveal that the MA-HPSOGA may be used as a promising method for achieving high-efficiency model updating in large-scale complex structures. Introduction A mathematical model is a representation or an abstract interpretation of physical reality that is amenable to analysis and calculation. Numerical simulation allows us to calculate the solutions of these models on a computer, and therefore to simulate physical reality. The Finite Element Method (FEM), as one of the most mainstream methods for numerical simulations, has always played an extraordinarily important role in exploring scientific and engineering problems [1]. For example, during project feasibility studies, Finite Element Analysis (FEA) has been applied to large civil engineering structures to pre-assess structural safety, leading to better construction guidance and risk prevention. However, the structural finite element (FE) models have inherent and epistemic errors, leading to some deviation between the system response after FEA and the real system response. These errors can be divided into the following three categories: (1) idealistic assumptions made to describe the mechanical behavior of physical structures, (2) inherent errors introduced by numerical methods and (3) typical errors arising from incorrect assumptions regarding model parameters. The first two are related to the mathematical structure of the model, which is hard to eliminate, while the last typical error can be reduced by model updating [2,3]. Model updating is essentially an optimization search process for solving inverse problems, and it can be an important tool for calibrating FE models to mitigate modeling errors [4]. By modifying the modeling parameters of the initial FE model, the discrepancy between the model response and the measured system response is minimized, enabling the behavior of the updated FE model to be as close as possible to the corresponding real structure. Conventional model updating methods is achieved by a trial-and-error approach and sensitivity analysis approach [5,6], which are often time-consuming and may not be feasible in some cases. In recent years, with the rapid development of modern intelligent algorithms, numerous studies have focused on metaheuristic algorithms for FE model updating. Tran et al. [7] proposed a hybrid metaheuristic algorithm, which incorporates fully the advantages of Genetic Algorithm (GA) and Improved Cuckoo Search (ICS), and then updated the FE model of a large twin steel truss bridge based on vibration measurement data. Dinh et al. [8] developed a Multi-Objective Cuckoo Search (MOCS) algorithm for optimizing the objective function in finite element model updating, and subsequently implemented the identification of the location and extent of multi-damages in functionally graded material structures. Minh et al. [9] developed a new balance of Cuckoo Search (NB-CS) algorithm for model updating and damage identification of three shell models with different levels of complexity, based on co-simulation with SAP2000 software and MATLAB. Sang et al. [10] proposed a novel stochastic optimization algorithm, namely the Planetary Optimization Algorithm (POA), for predicting horizontal displacements of diaphragm wall in high-rise buildings, and the structural FE model was calibrated based on field monitoring data from the early stages of excavation. Wu et al. [11] proposed an improved Crow Search Algorithm based on Levy Flight (LFCSA) for FE model updating, and the superiority of the proposed algorithm was highlighted in comparison with the standard CSA and the PSO algorithm with Levy Flight (LFPSO). Although modern intelligent algorithms are widely used for model updating, the updating process requires numerous calls to the structural FE model for iterative optimization. Especially for performing structural dynamics or nonlinear analysis, such crude operations make it extremely inefficient to apply modern intelligent algorithms to complex structures. Fortunately, metamodeling techniques provide a bridge between the two. The technique treats the relationship between inputs and outputs as a ‘black box’ model, allowing to create a computationally inexpensive mathematical approximation to substitute the expensive FE models without requiring additional information about the system [12]. There are some popular metamodels such as Artificial Neural Network (ANN) [13], Polynomial Chaos Expansion (PCE) [14], Kriging [15] and Support Vector Regression (SVR) [16], etc. Zhou et al. [17] constructed a Multi-Response Gaussian Process (MRGP) metamodel to substitute the structural FE model, and then adopted PSO and Simulated Annealing Algorithm (SAA) to minimize the discrepancy between the predicted and measured structural frequency responses. Naranjo et al. [18] developed a new efficient collaborative algorithm, which combines Harmony Search and Active-Set Algorithm (HS-ASA) and ANN and Principal Component Analysis (PCA). The proposed algorithm was applied to address the model updating of a real steel footbridge. Wang et al. [19] proposed a multi-scale model updating method, which combined Kriging metamodel with Non-dominated Sorting Genetic Algorithm-II (NSGA-II) for identifying the unknown parameters in transmission tower structures. The results showed that the proposed method could improve the accuracy of the tower in both global and local structural responses. Xia et al. [20] combined Back-Propagation Neural Network (BPNN) with Gaussian-white-noise-Mutation Particle Swarm Optimization (GMPSO), proposing an effective model updating method for complex bridge structures. Most of the literature mentioned above usually only employs single surrogate model in conjunction with the optimization algorithm for model updating, lacking a comprehensive investigation in the computational accuracy and efficiency of surrogate models applied to parameter identification. As it is well known, no single surrogate model was found to be the most effective for all problems. Hence, four widely popular metamodels, are chosen in this study, namely ANN, PCE, Kriging and SVR, with the view of revealing their differential performance in assisting the same hybrid optimization strategy, attempting to expand the boundaries of the existing literature mostly focusing on single metamodel. This paper aims to provide a promising generic paradigm for the rapid implementation of FE model updating, and its innovations have been integrated as follows: • The differential performance of several popular metamodelling techniques applied to FE model updating are compared and mechanistic explanation for the differences is provided. • An easy-to-implement method for identifying structural dynamic parameters is constructed using the common surrogate models and the popular optimization algorithms, breaking through the limitations of conventional methods to quickly determine the parameters. • Compared with the conventional iteration-based methods, the proposed method achieves a qualitative leap in computational efficiency while improving the computational accuracy. The rest of this article is structured as follows. Section 2 briefly describes the methodology of the four metamodels and HPSOGA, and Section 3 presents the procedures and evaluation metrics for the MA-HPSOGA method. Next, Sections 4 and 5 use two models with varying complexity, aiming to demonstrate the effectiveness and practicality of the proposed method. Conclusions are given in Section 6 and some detailed results are presented in Appendix A. Section snippets Theoretical fundamentals This section will briefly introduce the underlying theory of four popular metamodels and HPSOGA. Proposed parameter identification method: MA-HPSOGA The procedure of the proposed MA-HPSOGA method to identify structural unknown parameters is described in detail below. Step 1: Building computational model and selecting input parameters: The first step in the work is to build a high-precision FE model that adequately reflects the structural properties, followed by setting the input parameters. Usually, due to some uncertainty exists in the modeling and measurement, it is difficult to obtain the exact values of these input parameters. Therefore, Validation of MA-HPSOGA This section validates the effectiveness of the MA-HPSOGA method for model updating by studying a cantilever aluminum plate. Application of MA-HPSOGA to small-scaled dam The previous section provides a detailed validation for the effectiveness of the MA-HPSOGA method applied to model updating. This section will take a small-scaled laboratory model as example, to further explore the applicability of the MA-HPSOGA method to practical engineering problems. Conclusions In the real world, almost all complex structures, such as bridges and dams, are fraught with numerous uncertainties. Their values of material parameters are often difficult to measure accurately due to the complex construction environment and processes, as well as the extremely long construction periods. Therefore, identifying the most accurate material parameters, making the response of the structural FE model as close as possible to the real structural system response, becomes a priority in CRediT authorship contribution statement Li YiFei: Investigation, Methodology, Validation, Writing – original draft. Cao MaoSen: Investigation, Methodology, Validation, Supervision, Writing – review & editing. Tran N. Hoa: Software, Investigation. S. Khatir: Software, Investigation. Hoang-Le Minh: Software, Investigation. Thanh SangTo: Software, Investigation. Thanh Cuong-Le: Validation, Writing – review & editing. Magd Abdel Wahab: Investigation, Methodology, Validation, Supervision, Writing – review & editing. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Acknowledgement The authors are grateful for the Fundamental Research Funds for the Central Universities : No. B220204002 ; the 2022 National Young Foreign Talents Program of China: No.QN2022143002L; the Jiangsu International Joint Research and Development Program (No. BZ2022010); the Nanjing International Joint Research and Development Program (No. 202112003). References (52) S. Ereiz et al. Review of finite element model updating methods for structural applications (2022) J.T. Wang et al. Frequency response function-based model updating using Kriging model Mech Syst Signal Process (2017) J.E. Mottershead et al. The sensitivity method in finite element model updating: a tutorial Mech Syst Signal Process (2011) P.G. Bakir et al. Sensitivity-based finite element model updating using constrained optimization with a trust region algorithm J Sound Vib (2007) H.L. Minh et al. Structural damage identification in thin-shell structures using a new technique combining finite element model updating and improved Cuckoo search algorithm Adv Eng Software (2022) Y.F. Li et al. A surrogate-assisted stochastic optimization inversion algorithm: parameter identification of dams Adv Eng Inf (2023) C. Wang et al. Artificial neural network combined with damage parameters to predict fretting fatigue crack initiation lifetime Tribol Int (2022) L. YiFei et al. Multi-parameter identification of concrete dam using polynomial chaos expansion and slime mould algorithm Comput Struct (2023) A. Roy et al. Support vector regression based metamodel by sequential adaptive sampling for reliability analysis of structures Reliab Eng Syst Saf (2020) K. Zhou et al. Structural model updating using adaptive multi-response Gaussian process meta-modeling Mech Syst Signal Process (2021) J. Naranjo-Pérez et al. A collaborative machine learning-optimization algorithm to improve the finite element model updating of civil engineering structures Eng Struct (2020) H. Tran-Ngoc et al. Efficient Artificial neural networks based on a hybrid metaheuristic optimization algorithm for damage detection in laminated composite structures (2021) G. Blatman et al. Adaptive sparse polynomial chaos expansion based on least angle regression J Comput Phys (2011) L. YiFei et al. Structure damage identification in dams using sparse polynomial chaos expansion combined with hybrid K-means clustering optimizer and genetic algorithm Eng Struct (2023) H.L. Minh et al. An enhancing particle swarm optimization algorithm (EHVPSO) for damage identification in 3D transmission tower Eng Struct (2021) H. Tran-Ngoc et al. An efficient approach to model updating for a multispan railway bridge using orthogonal diagonalization combined with improved particle swarm optimization J Sound Vib (2020) N. Fahimi et al. Dynamic modeling of flashover of polymer insulators under polluted conditions based on HGA-PSO algorithm Electric Power Syst Res (2022) B. Shi et al. A new surface fractal dimension for displacement mode shape-based damage identification of plate-type structures Mech Syst Signal Process (2018) P. Diaz et al. Sparse polynomial chaos expansions via compressed sensing and d -optimal design Comput Methods Appl Mech Eng (2018) M. Griebel et al. Numerical simulation in fluid dynamics: a practical introduction (1998) S. Sehgal et al. Structural dynamic model updating techniques: a state of the art review Arch Comput Meth Eng (2016) H. Tran-Ngoc et al. Finite element model updating of a multispan bridge with a hybrid metaheuristic search algorithm using experimental data from wireless triaxial sensors Eng Comput (2021) D. Dinh-Cong et al. An effective damage identification procedure using model updating technique and multi-objective optimization algorithm for structures made of functionally graded materials Eng Comput (2021) T. Sang-To et al. Forecasting of excavation problems for high-rise building in Vietnam using planet optimization algorithm Sci Rep (2021) H. Wu et al. Finite element model updating using crow search algorithm with Levy flight Int J Numer Methods Eng (2020) T. Santner et al. The design and analysis of computer experiments (2003) View more references Cited by (0) Recommended articles (6) Research article Linear and nonlinear multiscale latent variable-based generalized likelihood ratio for fault detection Data-Driven and Model-Based Methods for Fault Detection and Diagnosis, 2020, pp. 79-133 Show abstract The objective of this chapter is to enhance the quality of fault detection by suppressing the effect of these errors using wavelet-based multiscale representation of data, which is a powerful feature extraction tool. Multiscale representation of data has been used to improve the fault detection abilities of latent variable (LV) (like PLS, PCA, kernel PCA and kernel PLS). Thus, combining the advantages of multiscale representation with those of hypothesis testing should provide even further improvements in FD. To do that, multiscale latent variables (MSLV)-based on GLRT are proposed for fault detection. The advantages of MSLV-based GLRT methods are threefold: (i) the dynamical multiscale representation is proposed to extract accurate deterministic features and decorrelate autocorrelated measurements; (ii) the MSLV model evaluates the LV of the wavelet coefficients at each scale. Due to its multiscale nature, MSLV is appropriate for modeling of data that contain contributions from events whose behavior changes over time and frequency; (iii) the GLRT is a composite hypothesis testing method and is known to have better fault detection performance compared to conventional LV statistics. The fault detection performances are illustrated through two examples, one using synthetic data and the other using simulated Tennessee Eastman Process (TEP) data. The results demonstrate the effectiveness of the multiscale-based approaches over the conventional methods. Research article De la bioéthique, d’une approche individuelle à une approche collective impliquant toutes les disciplines dans l’analyse de ce champ de la connaissance Ethics, Medicine and Public Health, Volume 1, Issue 1, 2015, pp. 1-2 Research article Bending of Bi-directional inhomogeneous nanoplates using microstructure-dependent higher-order shear deformation theory Engineering Structures, Volume 291, 2023, Article 116230 Show abstract In contrast of experimental analysis in the mechanical behavior of the complicated structures, the progress of theoretical research has many difficulties, especially with regard to the modelling of structures. The main objective of this research is to analyze analytically the static problem of a new model of functionally graded materials, namely the “bi-coated FGM” plate. The materials are graded continuously through 2-directional by using a power law function. Two types of coated FG plates are investigated, Hardcore and Softcore FG plates. Based on the generalized field of displacement, a Quasi-3D higher-order shear deformation plate theory is proposed in this work by reducing the number of variables from six to five variables. The equilibrium equations are performed based on the principle of virtual work and solved using the Galerkin method to cover different boundary conditions. The nonlocal strain gradient theory is employed to capture the microstructure-dependent effect. The model is validated compared to recent relevant research interests. Numerical results are performed to present the impact of hardcore and softcore distributions, gradation indices, and boundary conditions on static bending deflection and stresses of bi-coated FGM plate. Research article PINN-FORM: A new physics-informed neural network for reliability analysis with partial differential equation Computer Methods in Applied Mechanics and Engineering, Volume 414, 2023, Article 116172 Show abstract The first-order reliability method (FORM) is commonly used in the field of structural reliability analysis, which transforms the reliability analysis problem into the solution of an optimization problem with equality constraint. However, when the limit state functions (LSFs) in mechanical and engineering problems are complex, particularly for implicit partial differential equations (PDEs), FORM encounters computation difficulty and incurs unbearable computational effort. In this study, the physics-informed neural network (PINN), which is a new branch of deep learning technology for addressing forward and inverse problems with PDEs, is applied as a black-box solution tool. For LSFs with implicit PDE expressions, PINN-FORM is constructed by combining PINN with FORM, which can avoid the calculation of the real structure response. Moreover, a loss function model with an optimization target item is established. Then, an adaptive weight strategy, which can balance the interplay between different parts of the loss function, is suggested to enhance the predictive accuracy. To demonstrate the effectiveness of PINN-FORM, five benchmark examples with LSFs expressed by implicit PDEs, including two-dimensional and three-dimensional problems, and steady state and transient state problems are tested. The results illustrate the proposed PINN-FORM not only is very accurate, but also can simultaneously predict the solutions of PDEs and reliability index within a single training process. Research article Data-driven shear strength predictions of recycled aggregate concrete beams with /without shear reinforcement by applying machine learning approaches Construction and Building Materials, Volume 387, 2023, Article 131604 Show abstract Recycled aggregate concrete (RAC) has gradually received attention in recent years for sustainable development. However, the adaptation of RAC for structural elements is becoming less popular due to the lack of proper design guidelines. Thus, to encourage the structural application of RAC, it is essential to develop a framework for predicting the shear capacity of RAC beams. The data-driven prediction has demonstrated excellent accuracy compared to traditional empirical equations. This study proposed a data-driven framework using machine learning. In this paper, 401 samples of different RAC beams are analysed, including 264 slender beams and 137 deep beams. Eight machine learning (ML) algorithms namely Linear Regression, K-nearest neighbour, Random Forest, Adaptive boosting, Gradient Boosting, Extreme gradient boosting (XGBoost), Category Boosting, and Light Gradient Boosting Machine were employed to develop the best ML-based framework. Having the best prediction output (R2 score: 0.95 in slender beam, 0.78 in deep beam), XGBoost is selected to further analyse the feature importance and parametric study with SHapley Additive exPlanations (SHAP). Simultaneously, the experiment values were compared with calculated values according to the codes in different regions, such as ACI-381-19, AS3600-2018, Eurocode 2, and GB50010-2010. The results of SHAP from XGBoost indicate that the replacement ratio of RCA has limited influence on normalized shear strength value, V u / b d , and it is applicable to adopt RAC without shear strength reduction if the replacement ratio of RA is lower than 40%. Nevertheless, the comparison of experiment value from the database and calculated value according to classical analytical methods demonstrates the necessity for adjustment to existing design codes against RAC replacement ratio. Some suggestions which could further improve the accuracy of results are provided at the end of this paper as well. Research article Adaptive Kriging-based Bayesian updating of model and reliability Structural Safety, Volume 104, 2023, Article 102362 Show abstract Bayesian updating is a powerful tool to reassess and calibrate models and their reliability as new observations emerge, and the Bayesian updating with structural reliability method (BUS) is an efficient approach that reformulates it as a structural reliability problem. However, the efficiency and accuracy of BUS depend on a constant c determined by the maximum of likelihood function. To efficiently complete Bayesian updating with new observations related to implicit performance function, a method that combines adaptive Kriging with Bayesian updating is proposed. The proposed method involves three stages. Firstly, an innovatively a dvanced e xpected i mprovement (AEI) learning function is proposed to train the Kriging model of the likelihood function for estimating c , in which the convergence criterion and the strategy of selecting new training point guarantee the accuracy and efficiency of estimating c . Secondly, a new learning function based on e xpectation and v ariance of c ontribution u ncertainty f unction (EVCUF) is proposed to adaptively train the Kriging model of the performance function constructed in BUS to extract posterior samples and complete Bayesian updating of model. By simultaneously taking the expectation and variance of the contribution of the candidate sample to improving accuracy of the Kriging model into consideration, the EVCUF learning function ensures the robust and efficient convergence of the Kriging model. Finally, based on the training points of the previous two stages, the traditional U learning function is employed to subsequentially update Kriging model of the performance function for classifying posterior samples and completing Bayesian updating of reliability. Additionally, a reduction strategy of the candidate sample pool is proposed to improve the efficiency of the proposed method. After demonstrating the basic principle and advantage of the proposed method, three examples are introduced to verify the efficiency and accuracy of the proposed method. View full text © 2023 Elsevier Ltd. All rights reserved. About ScienceDirect Remote access Shopping cart Advertise Contact and support Terms and conditions Privacy policy We use cookies to help provide and enhance our service and tailor content and ads. By continuing you agree to the use of cookies . Copyright © 2023 Elsevier B.V. or its licensors or contributors. ScienceDirect® is a registered trademark of Elsevier B.V. ScienceDirect® is a registered trademark of Elsevier B.V.
|
