Batch 03 Agent 1 -- Literature Synthesis (Files 401-440)¶
Generated: 2026-04-17 | Source: literature_review/ positions 401-440 | All 2021 publications
Individual Paper Summaries¶
| # | Author(s) | Year | Title | Core Finding | Method | Tags |
|---|---|---|---|---|---|---|
| 1 | Guechi, Bordjiba | 2021 | Reliability Analysis of Stone Columns Bearing Capacity | FOSM, PEM, and Monte Carlo yield different reliability indices for stone column design; parameter sensitivity identifies reduced set of critical random variables | FOSM, PEM, Monte Carlo simulation, parametric sensitivity analysis | geotechnical, reliability, stone-columns, bearing-capacity |
| 2 | Gui et al. | 2021 | Uplift Resistance Capacity of Anchor Piles Used in Marine Aquaculture | Increasing tension angle and embedded depth significantly improves uplift capacity; dual anchor piles outperform single piles | Physical model experiments, CCD camera + load cell | marine-aquaculture, anchor-piles, uplift, model-test |
| 3 | He, Fenton, Griffiths | 2021 | Calibration of Resistance Factors for Bearing Resistance Design of Shallow Foundations Under Seismic and Wind Loading | Calibrated resistance factors for shallow foundations under combined seismic/wind loads using reliability-based design | Reliability-based design, resistance factor calibration | shallow-foundation, seismic, wind, LRFD, reliability |
| 4 | Huang et al. | 2021 | Responses of Soil Microbiome to Steel Corrosion | Iron-oxidizing, nitrifying, and denitrifying microorganisms enriched near corroding steel; MIC linked to biogeochemical cycling of Fe and N | Long-term outdoor soil burial experiments, 16S rRNA sequencing, co-occurrence network analysis | corrosion, MIC, soil-microbiome, steel |
| 5 | Ivanov et al. | 2021 | Offshore Wind Farm Footprint on Organic and Mineral Particle Flux to the Bottom | OWF foundations alter local hydrodynamics, modifying organic/mineral particle flux to seabed | Numerical modelling (hydrodynamic + sediment transport) | offshore-wind, environmental-impact, sediment-flux, ecology |
| 6 | Jang et al. | 2021 | Unsupervised Feature Learning for ECG Data Using Convolutional Variational Autoencoder | Conv-VAE learns clinically meaningful latent representations from ECG without labels | Convolutional variational autoencoder (Conv-VAE), unsupervised learning | autoencoder, VAE, ECG, unsupervised, deep-learning |
| 7 | Jin et al. | 2021 | Numerical Modeling of Soil-Pipe Interaction at Shallow Embedment in Clay by Hypoplastic Macroelement | Hypoplastic macroelement efficiently captures soil-pipe interaction for shallow pipelines in clay | Hypoplastic macroelement, numerical modeling | soil-pipe-interaction, clay, pipeline, macroelement |
| 8 | Kamariotis, Chatzi, Straub | 2021 | Value of Information from Vibration-Based SHM Extracted via Bayesian Model Updating | Bayesian decision analysis quantifies VoI of SHM systems; framework models full data-to-decision chain for deteriorating bridge | Bayesian model updating, preposterior decision analysis, OMA, life-cycle optimization | SHM, VoI, Bayesian, bridge, reliability |
| 9 | Kamei, Khan | 2021 | Current Challenges in Modelling Vibrational Fatigue and Fracture of Structures: A Review | Five key modelling challenges identified: vibration-based fatigue assessment, life estimation, crack propagation, crack quantification, and thermal-structural coupling | Literature review | fatigue, fracture, vibration, crack-propagation, review |
| 10 | Karniadakis et al. | 2021 | Physics-Informed Machine Learning | Comprehensive taxonomy of PIML approaches: observational/learning/inductive biases to embed physics into ML models | Review (Nature Reviews Physics) | PIML, PINN, physics-informed, deep-learning, review |
| 11 | Kim, Heo, Koo | 2021 | Analysis of Dynamic Response Characteristics for 5 MW Jacket-Type Fixed OWT | Under combined wind+wave, horizontal displacement dominated by wind; jacket leg shear force influenced by both; FFT reveals frequency-domain response characteristics | FAST simulation, FFT analysis | OWT, jacket, dynamic-response, wind-wave, FAST |
| 12 | Kim, Lee | 2021 | Model Tests for Tilting Control of Suction Bucket Foundation for OWT with Path Points | Tilting control within 5 deg achieved during installation and operation by combining positive/negative pressure in 3-cell suction bucket | 1:100 scale model tests, sand ground | suction-bucket, tilting-control, OWT, model-test |
| 13 | Kim, Ngo, Kim | 2021 | ULS Risk Assessment of Penta Pod Suction Bucket Support Structures for OWT Due to Scour | Scour risk assessed by combining scour hazard (empirical formula) with fragility curves based on allowable bearing capacity | Risk assessment: hazard x fragility, empirical scour models | OWT, suction-bucket, penta-pod, scour, risk-assessment |
| 14 | Kita et al. | 2021 | Earthquake-Induced Damage Identification in Historic Masonry Towers Combining OMA and IDA | DORI method combines OMA-based SHM with surrogate modelling and IDA for damage localization/quantification; validated on bell tower during 2016 Central Italy earthquakes | OMA, FE modelling, surrogate modelling, IDA, digital twin | SHM, masonry, earthquake, damage-identification, digital-twin |
| 15 | Kopf et al. | 2021 | Mixture-of-Experts VAE for Clustering from Similarity-Based Representations on Single Cell Data | MoE-Sim-VAE clusters single-cell data using similarity-based representations in a VAE framework | Mixture-of-experts VAE | VAE, single-cell, clustering, deep-learning |
| 16 | Kuester et al. | 2021 | 1D-Convolutional Autoencoder Based Hyperspectral Data Compression | 1D-Conv AE outperforms deep AE and NLPCA for hyperspectral compression at fixed compression ratio | 1D-convolutional autoencoder, SVM classification benchmark | autoencoder, hyperspectral, compression, remote-sensing |
| 17 | Larsen, Zhang, Hogsberg | 2021 | Vibration Damping of OWT by Optimally Calibrated Pendulum Absorber with Shunted EM Transducer | Pendulum absorber with electromagnetic shunt provides effective vibration reduction for OWT towers | Analytical/numerical optimization, pendulum damper design | OWT, vibration-damping, pendulum-absorber, structural-control |
| 18 | Lee, Fields | 2021 | An Overview of Wind-Energy-Production Prediction Bias, Losses, and Uncertainties | Long-term trend of declining overprediction bias (historically 3.5-4.5%); uncertainty remains prominent; IEC 61400-15 loss/uncertainty framework documented | Literature review, statistical analysis of industry EYA data | wind-energy, AEP, prediction-bias, uncertainty, losses |
| 19 | Liu et al. | 2021 | Exploring Autoencoder-Based Error-Bounded Compression for Scientific Data | AE-based error-bounded framework achieves 100-800% improvement in compression ratio vs SZ2.1/ZFP at high compression ratios | Convolutional autoencoder + SZ error-bounding framework | autoencoder, compression, scientific-data, error-bounded |
| 20 | Luo et al. | 2021 | A Topology-Preserving Dimensionality Reduction Method for scRNA-seq Using Graph Autoencoder | scGAE preserves global topological structure among cells, outperforming other deep learning methods for trajectory inference | Graph autoencoder, multitask learning | autoencoder, graph-AE, scRNA-seq, dimensionality-reduction |
| 21 | Madabhushi, Garcia-Torres | 2021 | Sustainable Measures for Protection of Structures Against Earthquake-Induced Liquefaction | Rubble brick earthquake drains effectively mitigate liquefaction-induced settlement; sustainable and economic ground improvement | Dynamic centrifuge testing, 3D FE analysis | liquefaction, centrifuge, drains, ground-improvement, sustainability |
| 22 | Mahfouz, Roser, Cheng | 2021 | Verification of SIMPACK-MoorDyn Coupling for 15 MW IEA-Wind FOWT Models | SIMPACK-MoorDyn coupling verified for 15 MW floating OWT reference models (Activefloat, WindCrete) | Multibody simulation (SIMPACK), mooring dynamics (MoorDyn) | FOWT, mooring, SIMPACK, MoorDyn, verification |
| 23 | Massaoudi et al. | 2021 | Deep Learning in Smart Grid Technology: A Review | DL enables decentralized intelligent energy management; federated learning and edge intelligence are key enablers | Systematic literature review, bibliometric analysis | deep-learning, smart-grid, federated-learning, review |
| 24 | Meixedo et al. | 2021 | Progressive Numerical Model Validation of Bowstring-Arch Railway Bridge Based on SHM | Progressive FE model validation using static, modal, and dynamic SHM data streams from a long-span bridge | FE modelling, ambient vibration test, SHM data integration | SHM, bridge, FE-model-validation, railway |
| 25 | Methratta | 2021 | Distance-Based Sampling Methods for Assessing Ecological Effects of OWFs | Synthesizes distance-based sampling designs (BACI, gradient) for fisheries resource studies near OWFs | Literature review, sampling design synthesis | offshore-wind, ecology, fisheries, sampling-design |
| 26 | Millan, Galindo, Alencar | 2021 | ANN for Predicting Bearing Capacity of Shallow Foundations on Rock Masses | ANN trained on FLAC numerical results accurately predicts bearing capacity using Hoek-Brown criterion inputs | ANN, FLAC numerical modelling, Hoek-Brown criterion | ANN, bearing-capacity, rock-mass, shallow-foundation |
| 27 | Musial et al. | 2021 | Offshore Wind Market Report: 2021 Edition | Comprehensive U.S. and global offshore wind market status; documents pipeline, technology trends, and cost trajectories | Industry market report (NREL/DOE) | offshore-wind, market-report, policy, cost |
| 28 | Nagai | 2021 | Evaluation of Dynamic Interaction Factor of Rectangular Piled Raft Foundation | Proposed formula for dynamic interaction factor considering foundation aspect ratio; simplified method matches FEM results | Numerical analysis (FEM), parametric study | piled-raft, dynamic-impedance, interaction-factor, seismic |
| 29 | Oestergaard et al. | 2021 | Intelligent Physical Exercise Training (IPET) in the Offshore Wind Industry | On-site supervised exercise during working hours achieves 95% compliance among wind technicians; home-administered drops to <20% | Within-subject feasibility study, 12-week intervention | offshore-wind, occupational-health, exercise, feasibility |
| 30 | Otter et al. | 2021 | A Review of Modelling Techniques for Floating Offshore Wind Turbines | Strong aero-hydro coupling and Froude-Reynolds scaling mismatch are key challenges; trend toward high-fidelity numerical methods | Review article | FOWT, modelling, aerodynamics, hydrodynamics, review |
| 31 | Parvizi et al. | 2021 | Regional Frequency Analysis of Drought in Karkheh River Basin Using L-Moments | SPI shows most severe droughts; hydrological drought follows meteorological drought with short delay; L-moments identify best regional distributions | L-moments, K-means clustering, regional frequency analysis | drought, L-moments, frequency-analysis, hydrology |
| 32 | Poggio et al. | 2021 | SoilGrids 2.0: Producing Soil Information for the Globe with Quantified Spatial Uncertainty | Global 250 m soil property maps using 240k observations + 400 covariates; spatial uncertainty highlights need for more high-latitude data | Machine learning (quantile regression forest), digital soil mapping | soil, global-mapping, machine-learning, uncertainty |
| 33 | Prakash, Muthukkumaran | 2021 | Estimation of Lateral Capacity of Rock Socketed Piles in Layered Soil-Rock Profile | 3D rock socketing yields ~18x lateral capacity vs non-socketed piles; embedment depth in soil also significant | Model pile experiments with instrumentation | rock-socketed-pile, lateral-capacity, model-test |
| 34 | Pulletz et al. | 2021 | Dynamic Relative Regional Strain Visualized by EIT in COVID-19 Patients | Novel EIT-based DRRS metric visualizes inhomogeneous lung strain in COVID-19; correlates with lung ultrasound scores | Electrical impedance tomography (EIT), clinical observational study | EIT, COVID-19, lung-strain, monitoring |
| 35 | Ramadan | 2021 | Numerical Analysis of Bearing Capacity of Loose Sand Overlying Clay | Bearing capacity of layered sand-over-clay depends on H/B ratio and undrained shear strength; design charts developed | PLAXIS 3D Tunnel, FEM parametric study | bearing-capacity, layered-soil, sand-over-clay, FEM |
| 36 | Rathore et al. | 2021 | The Role of AI, ML, and Big Data in Digital Twinning: A Systematic Literature Review | Proposed big-data-driven, AI-enriched reference architecture for DT systems; mapped current tools and deployment gaps | Systematic literature review, patent analysis | digital-twin, AI, ML, big-data, review |
| 37 | Ritto, Rochinha | 2021 | Digital Twin, Physics-Based Model, and ML Applied to Damage Detection in Structures | Physics-based model generates training data for ML classifier serving as real-time digital twin; interpretable + fast | Lumped-parameter model, ML classifiers (SVM, QDA), uncertainty quantification | digital-twin, damage-detection, ML, physics-based, structural-dynamics |
| 38 | Safdar, Newson, Shah | 2021 | CID Behavior of Fibre Reinforced Cemented Toyoura Sand in Triaxial Loading | Fibre+cement additives increase peak strength up to 243% at higher effective stresses; least effective in extension loading | Triaxial tests (CID), undercompaction moist tamping | geotechnical, fibre-reinforcement, cemented-sand, triaxial |
| 39 | Sarkar, Chakraborty | 2021 | Stability Analysis for Two-Layered Slopes by Strength Reduction Method | Stability charts for two-layered slopes; SRM avoids prior assumptions on failure surface unlike LEM | Finite element limit analysis (LB + UB), SRM | slope-stability, layered-soil, SRM, limit-analysis |
| 40 | Sause, Jasiuniene (eds.) | 2021 | Structural Health Monitoring Damage Detection Systems for Aerospace | Comprehensive treatment of SHM damage detection approaches for aerospace structures | Edited book/review | SHM, aerospace, damage-detection, NDE |
SYNTHESIS¶
CONSENSUS (areas of broad agreement)¶
-
Reliability and probabilistic methods are essential for geotechnical design. Multiple papers (Guechi, He, Guechi) confirm that deterministic safety factors inadequately capture parameter uncertainty. Monte Carlo simulation, FOSM, and LRFD-calibrated resistance factors are becoming standard practice for foundations and ground improvement.
-
SHM provides quantifiable decision support when embedded in Bayesian or digital-twin frameworks. Kamariotis demonstrates that Value of Information (VoI) from vibration-based SHM can be formally quantified, while Kita, Meixedo, and Ritto show that combining OMA data with FE models and ML classifiers enables damage detection through localization. All converge on the view that SHM's real value emerges only when linked to maintenance decisions.
-
Physics-informed and hybrid approaches outperform pure data-driven methods. Karniadakis's landmark review, Ritto's digital-twin framework, and the autoencoder papers collectively support embedding physical constraints (conservation laws, boundary conditions, monotonicity) into ML architectures to improve generalization and interpretability.
-
Autoencoders are versatile across domains. Six papers apply autoencoder variants (Conv-VAE, graph-AE, 1D-Conv-AE, MoE-VAE, error-bounded AE) to ECG, single-cell genomics, hyperspectral imagery, and scientific data compression. Consensus: latent space representations preserve task-relevant structure when architecture matches data geometry.
-
Offshore wind substructure design must account for scour, dynamic soil-structure interaction, and multi-hazard loading. Kim (jacket), Kim (suction bucket tilting), Kim (penta pod scour risk), and Larsen (vibration damping) all demonstrate that combined wind-wave-current loading and seabed degradation govern design adequacy.
DEBATES (contested or unresolved issues)¶
-
Deterministic vs. probabilistic foundation design. While reliability methods are advocated (Guechi, He), the industry still relies heavily on deterministic approaches. The optimal level of probabilistic sophistication for routine design remains debated, particularly for novel foundation types like suction buckets.
-
SHM data utilization: detection vs. full damage identification. Kita notes that vibration-based SHM is strong at detection but limited at localization without model integration. The degree to which data-driven methods alone can achieve Level 3+ damage identification (localization, quantification) without physics-based models remains an open question.
-
Autoencoder architecture selection. The six autoencoder papers each propose different architectures for different domains. Whether a unified architecture principle exists or domain-specific tuning is always required is not settled. Graph-AE (Luo) preserves topology; 1D-Conv (Kuester) captures spectral correlation; MoE-VAE (Kopf) handles multimodal distributions.
-
Digital twin fidelity requirements. Rathore and Ritto disagree implicitly on how much physics is needed. Rathore's systematic review emphasizes data-driven DTs powered by big data and AI, while Ritto argues that physics-based models provide essential interpretability and that ML should augment, not replace, physics.
GAPS (under-explored areas needing future research)¶
-
No papers address autoencoder-based representation learning for geotechnical or structural health monitoring data specifically. The autoencoder papers target biomedical and remote-sensing domains; applying these architectures to vibration signals, soil sensor data, or load-displacement curves for OWT foundations is unexplored in this batch.
-
Scour-SHM integration is missing. Kim (penta pod) assesses scour risk via fragility curves, but no paper links real-time scour monitoring data back to structural reliability updating or digital-twin frameworks for OWT foundations.
-
Multi-hazard probabilistic assessment for OWT foundations. Individual hazards (scour, seismic, wind-wave) are treated separately. A unified multi-hazard reliability framework that combines all loading sources with soil-structure interaction is absent.
-
Transfer learning and domain adaptation for SHM. Karniadakis discusses PIML broadly but no paper demonstrates transfer of learned damage features across different structure types or loading conditions.
-
Long-term field validation of centrifuge-derived design methods. Madabhushi validates rubble-brick drains via centrifuge and simplified FE, but field-scale long-term performance data is not presented.
METHODS (methodological patterns across the batch)¶
- Numerical simulation dominates: FAST, PLAXIS 3D, FLAC, FEM limit analysis, SIMPACK-MoorDyn coupling
- Physical model testing remains critical: centrifuge (Madabhushi), 1:100 scale (Kim suction bucket), instrumented model piles (Prakash, Gui)
- Machine learning integration is growing: ANN for bearing capacity (Millan), ML classifiers for damage detection (Ritto), autoencoders for representation learning (6 papers)
- Bayesian frameworks for decision-making: VoI analysis (Kamariotis), reliability-based design (He, Guechi)
- Review papers synthesize rapidly evolving fields: Karniadakis (PIML), Otter (FOWT modelling), Kamei (vibrational fatigue), Rathore (DT+AI), Massaoudi (DL in smart grids), Lee (wind energy prediction)
BENCHMARKS (reference values and datasets)¶
| Parameter | Value | Source |
|---|---|---|
| Wind energy P50 overprediction bias (US) | 3.5-4.5% | Lee & Fields 2021 |
| Suction bucket tilting control limit | 5 deg achievable | Kim & Lee 2021 |
| Rock-socketed pile lateral capacity gain (3D socket) | ~18x vs non-socketed | Prakash & Muthukkumaran 2021 |
| Fibre+cement peak strength improvement (high stress) | up to 243% | Safdar et al. 2021 |
| AE compression ratio improvement vs SZ2.1/ZFP | 100-800% | Liu et al. 2021 |
| SoilGrids 2.0 resolution | 250 m, 240k observations, 400 covariates | Poggio et al. 2021 |
| IPET offshore exercise compliance (supervised) | 95% | Oestergaard et al. 2021 |
| Karniadakis PIML review citations | 2,261+ | Karniadakis et al. 2021 |