An efficient algorithm for architecture design of Bayesian neural network in structural model updating

Yin, Tao; Zhu, Hongping

doi:10.1111/mice.12492

Cited by 45 publications

(18 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, too small number of hidden neurons will make the ANN a poor-quality estimator. In the presented case, training an ANN with a higher number of neurons might result in higher accuracy, however, too large number of hidden neurons often also results in decrease of results quality as the output fluctuates in the area between training points [40,41].…”

Section: Discussionmentioning

confidence: 99%

Operational Load Monitoring of a Composite Panel Using Artificial Neural Networks

Mucha

Kuś

Viana

et al. 2020

Sensors

View full text Add to dashboard Cite

Operational Load Monitoring consists of the real-time reading and recording of the number and level of strains and stresses during load cycles withstood by a structure in its normal operating environment, in order to make more reliable predictions about its remaining lifetime in service. This is particularly important in aeronautical and aerospace industries, where it is very relevant to extend the components useful life without compromising flight safety. Sensors, like strain gauges, should be mounted on points of the structure where highest strains or stresses are expected. However, if the structure in its normal operating environment is subjected to variable exciting forces acting in different points over time, the number of places where data will have be acquired largely increases. The main idea presented in this paper is that instead of mounting a high number of sensors, an artificial neural network can be trained on the base of finite element simulations in order to estimate the state of the structure in its most stressed points based on data acquired just by a few sensors. The model should also be validated using experimental data to confirm proper predictions of the artificial neural network. An example with an omega-stiffened composite structural panel (a typical part used in aerospace applications) is provided. Artificial neural network was trained using a high-accuracy finite element model of the structure to process data from six strain gauges and return information about the state of the panel during different load cases. The trained neural network was tested in an experimental stand and the measurements confirmed the usefulness of presented approach.

show abstract

Section: Discussionmentioning

confidence: 99%

Operational Load Monitoring of a Composite Panel Using Artificial Neural Networks

Mucha

Kuś

Viana

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…To resolve the combinatorial optimization problem relevant to the optimal sensor configuration more effectively, a novel algorithm for handling this type of optimization problem is newly proposed in this paper. By introducing the concept of binary optimization strategy, the present algorithm as given in Algorithm 1 is developed by modifying a bound‐constrained version of the classical Nelder–Mead simplex method, 61 which is a commonly applied numerical routine to perform a direct search in a multidimensional real space without knowing the derivatives of objective function. Where, the mandatory constraint required for this particular combinatorial optimization problem, that is, the total number of optimally configured sensors should be equal to the available ones

\sum_{i = 1}^{N_{d}} δ_{i} = N_{o}

, is well satisfied.…”

Section: The Proposed Methodologymentioning

confidence: 99%

“…To fulfill this purpose, the concept of structural health monitoring (SHM) was proposed, and there has been great development in the SHM area utilizing dynamic measurements in the last few decades 1 . A large amount of methods has been proposed in this research area and applied to various types of basic structural members, laboratory models, and even real‐life structures, including trusses, 2–4 beams and frames, 5–11 plates and shells, 12–18 periodic structures, 19–23 hydraulic steel structures, 24–26 buildings, 27–33 and bridges 34–40 . However, in addition to the specific methods employed for solving the relevant problems as mentioned above, the success of vibration‐based SHM also highly depends on the quality of collected measurement data, which is ensured by the quantity and detailed layout of employed sensors; at present, the sensor configuration is still designed based on experience in most cases by considering a series of practical constraints.…”

Section: Introductionmentioning

confidence: 99%

Optimal sensor configuration for structural response prediction by a modified Nelder–Mead simplex method

Yin

2021

Struct Control Health Monit

Self Cite

View full text Add to dashboard Cite

Summary The optimal configuration of the sensor system is extremely critical to effectively control the test cost and ensure the quality of the collected data for structural health monitoring (SHM) system. Different from the traditional sensor configuration criterion, the information entropy measure aims to minimize the uncertainty of structural model parameters to be identified, which can ensure that the collected data provides the maximum amount of information for the model identification. However, most of the current researches within the framework of entropy measure are aimed at the model identification, while studies that consider response prediction are still rare. In this paper, by constructing the posterior predictive distribution function of the frequency response functions (FRFs) at the unmeasured positions of structures and also the corresponding information entropy representation, the optimal sensor configuration problem for the purpose of structural response prediction at the unmeasured position is studied. A modified bound‐constrained Nelder–Mead simplex method to handle the binary encoded optimization variables is also proposed, which can effectively solve the combinatorial optimization problem in the optimal sensor configuration. The methodology proposed is validated by a forced vibration study conducted for a laboratory rectangular cantilever plate. The obtained results show that the prediction error and uncertainty of the higher‐order modes of the FRFs at the unmeasured positions is generally larger than that of the lower ones. Also, by comparing the performance of worst and optimal sensor layouts, the importance of the optimal configuration for response prediction at unmeasured locations is clearly revealed especially for a small number of sensors.

show abstract

“…As computational efficiency is a major issue, and the large number of elements and parameters in cable-stayed bridge FE models make them difficult to update directly, the metamodels have been utilised to alleviate this problem. The response surface method [155,156], neural networks [157,158], Kriging model [159,160], and stochastic expansion methods [161,162] have been the focus of research in this area, yet few of these have been applied to cable-stayed bridges.…”

Section: Issues In Fe Modelling and Model Updatingmentioning

confidence: 99%

Latest Advances in Finite Element Modelling and Model Updating of Cable-Stayed Bridges

et al. 2022

View full text Add to dashboard Cite

As important links in the transport infrastructure system, cable-stayed bridges are among the most popular candidates for implementing structural health monitoring (SHM) technology. The primary aim of SHM for these bridges is to ensure their structural integrity and satisfactory performance by monitoring their behaviour over time. Finite element (FE) model updating is a well-recognised approach for SHM purposes, as an accurate model serves as a baseline reference for damage detection and long-term monitoring efforts. One of the many challenges is the development of the initial FE model that can accurately reflect the dynamic characteristics and the overall behaviour of a bridge. Given the size, slenderness, use of long cables, and high levels of structural redundancy, precise initial models of long-span cable-stayed bridges are desirable to better facilitate the model updating process and to improve the accuracy of the final updated model. To date, very few studies offer in-depth discussions on the modelling approaches for cable-stayed bridges and the methods used for model updating. As such, this article presents the latest advances in finite element modelling and model updating methods that have been widely adopted for cable-stayed bridges, through a critical literature review of existing research work. An overview of current SHM research is presented first, followed by a comprehensive review of finite element modelling of cable-stayed bridges, including modelling approaches of the deck girder and cables. A general overview of model updating methods is then given before reviewing the model updating applications to cable-stayed bridges. Finally, an evaluation of all available methods and assessment for future research outlook are presented to summarise the research achievements and current limitations in this field.

show abstract

An efficient algorithm for architecture design of Bayesian neural network in structural model updating

Cited by 45 publications

References 58 publications

Operational Load Monitoring of a Composite Panel Using Artificial Neural Networks

Operational Load Monitoring of a Composite Panel Using Artificial Neural Networks

Optimal sensor configuration for structural response prediction by a modified Nelder–Mead simplex method

Latest Advances in Finite Element Modelling and Model Updating of Cable-Stayed Bridges

Contact Info

Product

Resources

About