Prediction of flexible pavement 3-D finite element responses using Bayesian neural networks

Okte, Egemen; Al-Qadi, Imad L.

doi:10.1080/10298436.2021.1993219

Cited by 11 publications

(5 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This suggested that these models had a reasonable balance between the calibration of model uncertainty and acceptable model fit, making them valid choices. While

can be tuned during hyperparameter optimisation, its higher computational cost due to retraining with different

often necessitates post-hoc tuning [ 19 , 21 ]. In this study, it was shown that performing post-hoc calibration is effective since the optimal

has a negligible effect on the predictive mean of the model.…”

Section: Discussionmentioning

confidence: 99%

“…Various methods have been developed for calibrating probabilistic ML models quantitatively, including techniques that average and maximise calibration errors, yielding a single scalar metric for both classifications [ 15 ], and, more recently, regression models [ 17 , 19 ]. Recent studies [ 20 , 21 ] have incorporated model calibration into their BNNs to ensure accurate uncertainty quantification. Okte and Al-Qadi [ 21 ] developed a 3D FE surrogate model using a dropout-based BNN and demonstrated how varying dropout rates influence model uncertainties through calibration curve plots.…”

Section: Introductionmentioning

confidence: 99%

“…Recent studies [ 20 , 21 ] have incorporated model calibration into their BNNs to ensure accurate uncertainty quantification. Okte and Al-Qadi [ 21 ] developed a 3D FE surrogate model using a dropout-based BNN and demonstrated how varying dropout rates influence model uncertainties through calibration curve plots. Nevertheless, it is imperative to emphasise the importance of both hyperparameter optimisation and model calibration in ensuring the accuracy and reliability of probabilistic ANNs for practical applications.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon

Vien,

Kuen,

Rose

et al. 2024

Sensors

View full text Add to dashboard Cite

This study aims to enhance diagnostic capabilities for optimising the performance of the anaerobic sewage treatment lagoon at Melbourne Water’s Western Treatment Plant (WTP) through a novel machine learning (ML)-based monitoring strategy. This strategy employs ML to make accurate probabilistic predictions of biogas performance by leveraging diverse real-life operational and inspection sensor and other measurement data for asset management, decision making, and structural health monitoring (SHM). The paper commences with data analysis and preprocessing of complex irregular datasets to facilitate efficient learning in an artificial neural network. Subsequently, a Bayesian mixture density neural network model incorporating an attention-based mechanism in bidirectional long short-term memory (BiLSTM) was developed. This probabilistic approach uses a distribution output layer based on the Gaussian mixture model and Monte Carlo (MC) dropout technique in estimating data and model uncertainties, respectively. Furthermore, systematic hyperparameter optimisation revealed that the optimised model achieved a negative log-likelihood (NLL) of 0.074, significantly outperforming other configurations. It achieved an accuracy approximately 9 times greater than the average model performance (NLL = 0.753) and 22 times greater than the worst performing model (NLL = 1.677). Key factors influencing the model’s accuracy, such as the input window size and the number of hidden units in the BiLSTM layer, were identified, while the number of neurons in the fully connected layer was found to have no significant impact on accuracy. Moreover, model calibration using the expected calibration error was performed to correct the model’s predictive uncertainty. The findings suggest that the inherent data significantly contribute to the overall uncertainty of the model, highlighting the need for more high-quality data to enhance learning. This study lays the groundwork for applying ML in transforming high-value assets into intelligent structures and has broader implications for ML in asset management, SHM applications, and renewable energy sectors.

show abstract

“…This suggested that these models had a reasonable balance between the calibration of model uncertainty and acceptable model fit, making them valid choices. While

can be tuned during hyperparameter optimisation, its higher computational cost due to retraining with different

often necessitates post-hoc tuning [ 19 , 21 ]. In this study, it was shown that performing post-hoc calibration is effective since the optimal

has a negligible effect on the predictive mean of the model.…”

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon

Vien,

Kuen,

Rose

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

“…Apart from LIME, other alternative model agnostic approaches are available, and SHapley Additive exPlanations (SHAP) is one of them Lundberg and Lee (2017) . However, our choice of LIME over SHAP was influenced by some of the following factors ( Okte and Al-Qadi, 2021 , Ribeiro et al, 2016a ):…”

Section: Methodsmentioning

confidence: 99%

Deep transfer learning approaches for Monkeypox disease diagnosis

Ahsan

Uddin

Ali

et al. 2023

Expert Systems with Applications

View full text Add to dashboard Cite

“…Numerical simulation makes it possible to investigate the characteristics of dynamic responses of asphalt pavement under vehicle loading in simple or multiple service environments [18][19][20][21][22], which can help to present the health status of asphalt pavement structure and to further predict the residual service life [23][24][25][26]. Sun et al [3] constructed a 3D FEM model for the asphalt pavement under multi-field coupling, which included hydromechanical coupling and thermal hydromechanical coupling, and the dynamic response characteristics of saturated structure were numerically analyzed.…”

Section: Structural Dynamic Response Analysismentioning

confidence: 99%

Review on Design, Characterization, and Prediction of Performance for Asphalt Materials and Asphalt Pavement Using Multi-Scale Numerical Simulation

Wang,

Wang

2024

Materials

View full text Add to dashboard Cite

Asphalt pavement, which is mainly made up of the asphalt mixture, exhibits complicated mechanical behaviors under the combined effects of moving vehicle loads and external service environments. Multi-scale numerical simulation can well characterize behaviors of asphalt materials and asphalt pavement, and the essential research progress is systematically summarized from an entire view. This paper reviews extensive research works concerning aspects of the design, characterization, and prediction of performance for asphalt materials and asphalt pavement based on multi-scale numerical simulation. Firstly, full-scale performance modeling on asphalt pavement is discussed from aspects of structural dynamic response, structural and material evaluation, and wheel–pavement interaction. The correlation between asphalt material properties and pavement performance is also analyzed, and so is the hydroplaning phenomenon. Macro- and mesoscale simulations on the mechanical property characterization of the asphalt mixture and its components are then investigated, while virtual proportion design for the asphalt mixture is introduced. Features of two-dimensional and three-dimensional microscale modeling on the asphalt mixture are summarized, followed by molecular dynamics simulation on asphalt binders, aggregates, and their interface, while nanoscale behavior modeling on asphalt binders is presented. Finally, aspects that need more attention concerning this study’s topic are discussed, and several suggestions for future investigations are also presented.

show abstract

Prediction of flexible pavement 3-D finite element responses using Bayesian neural networks

Cited by 11 publications

References 24 publications

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon

Optimisation and Calibration of Bayesian Neural Network for Probabilistic Prediction of Biogas Performance in an Anaerobic Lagoon

Deep transfer learning approaches for Monkeypox disease diagnosis

Review on Design, Characterization, and Prediction of Performance for Asphalt Materials and Asphalt Pavement Using Multi-Scale Numerical Simulation

Contact Info

Product

Resources

About