Prediction of normalized shear modulus and damping ratio for granular soils over a wide strain range using deep neural network modelling DOI

Wei‐Qiang Feng,

Meysam Bayat, Zohreh Mousavi

et al.

Georisk Assessment and Management of Risk for Engineered Systems and Geohazards, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 30

Published: Dec. 19, 2024

Dynamic properties, such as shear modulus and damping ratio, are critical for civil engineering applications essential accurate dynamic response analysis. This study introduces a novel Deep Neural Network (DNN) approach to predict the normalized (G/Gmax) ratio (D) of granular soils over wide strain range. Utilising comprehensive dataset from cyclic triaxial (CT) resonant column (RC) tests, we developed Feed-Forward (DFFNN) model. The model incorporates grading characteristics, strain, void mean effective confining pressure, consolidation stress specimen preparation method inputs. DFFNN demonstrated high accuracy with testing results 0.9830 G/Gmax 0.9396 D, outperforming traditional empirical models other intelligent techniques Shallow (SNN), Support Vector Regression (SVR), Gradient Boosting (GBR). data-driven offers robust adaptable predicting properties across diverse conditions.

Language: Английский

Shear Wave Velocity Prediction with Hyperparameter Optimization DOI Creative Commons
Gebrai̇l Bekdaş, Yaren Aydın, Ümit Işıkdağ

et al.

Information, Journal Year: 2025, Volume and Issue: 16(1), P. 60 - 60

Published: Jan. 16, 2025

Shear wave velocity (Vs) is an important soil parameter to be known for earthquake-resistant structural design and determining the dynamic properties of soils such as modulus elasticity shear modulus. Different Vs measurement methods are available. However, these methods, which costly labor intensive, have led search new Vs. This study aims predict (Vs (m/s)) using depth (m), cone resistance (qc) (MPa), sleeve friction (fs) (kPa), pore water pressure (u2) N, unit weight (kN/m3). Since varies with depth, regression studies were performed at depths up 30 m in this study. The dataset used open-source dataset, data from Taipei Basin. was extracted, a 494-line created. In study, HyperNetExplorer 2024V1, prediction based on shell (fs), (kN/m3) values could satisfactory results (R2 = 0.78, MSE 596.43). Satisfactory obtained Explainable Artificial Intelligence (XAI) models also used.

Language: Английский

Citations

1
Compound damage detection using wavelet transform and deep neural network trained on healthy and single damage states: Validation on a laboratory-scale offshore jacket model DOI
Wei-Qiang Feng, Zohreh Mousavi, Jian‐Fu Lin

et al.

Structural Health Monitoring, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 21, 2025

Structural health monitoring is vital for the early detection of damage, enabling effective life cycle management structures. Detecting compound where multiple types damage occur simultaneously in different sections a structure, particularly challenging, especially when some damages are subtle or minor. Existing methods typically treat as distinct category, separate from single types. This paper introduces novel approach to based solely on vibration responses, combining wavelet transform with deep convolutional neural network interference (MIDCNN). In this approach, MIDCNN trained using time-frequency data healthy and states, intentionally excluding training phase. During testing, model accurately distinguishes between healthy, untrained states output probabilities meet predefined conditions. The method validated laboratory-scale offshore jacket structure. results demonstrate method’s ability extract relevant features classify structural including single, damage.

Language: Английский

Citations

1
Parameters determination methods and project validation of hardening soil model with small strain stiffness based on finite element method DOI Creative Commons
Maiying Kong, Yun Qin, Chao Zhang

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 21, 2025

The hardening soil model with small strain stiffness is a valuable tool for predicting the deformation of support structures during excavation phase construction projects. parameters model, which are dependent on technically complex and costly tests or estimated through specific proportionate, may exhibit some discrepancy between analyzed results project monitoring data in certain aspects. In light findings from conducted analyses studies, new concepts reference in-situ overburden pressure void ratio proposed objective enabling determination essential required HSS utilization current geotechnical tests, high popularity economy. Additionally, article offers recommendations determining other parameters, providing summary systematic approach to necessary strain. Ultimately, comparison verification finite element method deep demonstrated that research outcomes exhibited sufficient numerical analysis accuracy practical applicability.

Language: Английский

Citations

0
Onsite intensity prediction for earthquake early warning with multimodal deep learning DOI
Jingbao Zhu, Shanyou Li, Qiang Ma

et al.

Soil Dynamics and Earthquake Engineering, Journal Year: 2025, Volume and Issue: 195, P. 109430 - 109430

Published: April 8, 2025

Language: Английский

Citations

0
Prediction of normalized shear modulus and damping ratio for granular soils over a wide strain range using deep neural network modelling DOI

Wei‐Qiang Feng,

Meysam Bayat, Zohreh Mousavi

et al.

Georisk Assessment and Management of Risk for Engineered Systems and Geohazards, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 30

Published: Dec. 19, 2024

Dynamic properties, such as shear modulus and damping ratio, are critical for civil engineering applications essential accurate dynamic response analysis. This study introduces a novel Deep Neural Network (DNN) approach to predict the normalized (G/Gmax) ratio (D) of granular soils over wide strain range. Utilising comprehensive dataset from cyclic triaxial (CT) resonant column (RC) tests, we developed Feed-Forward (DFFNN) model. The model incorporates grading characteristics, strain, void mean effective confining pressure, consolidation stress specimen preparation method inputs. DFFNN demonstrated high accuracy with testing results 0.9830 G/Gmax 0.9396 D, outperforming traditional empirical models other intelligent techniques Shallow (SNN), Support Vector Regression (SVR), Gradient Boosting (GBR). data-driven offers robust adaptable predicting properties across diverse conditions.

Language: Английский

Citations

0