Batch 06 Agent 4 -- Literature Synthesis (Files 1121-1160)¶
Individual Paper Summaries¶
| # | Author(s) | Year | Title | Core Finding | Method | Tags |
|---|---|---|---|---|---|---|
| 1 | Wu et al. | 2019 | Numerical simulation of spudcan-soil interaction using improved SPH | Improved SPH captures excess pore water pressure during spudcan penetration in clay; validated against centrifuge data | Smoothed particle hydrodynamics (SPH), centrifuge validation | geotechnics, large-deformation, spudcan, numerical |
| 2 | Aguilera et al. | 2025 | Hydro-elastic coupling effect on spar-type FOWT dynamic response | Rigid floater assumption causes 37% eigenfrequency error; adding floater flexibility + added mass reduces error to 5% | In-situ sensor data, hydro-servo-aero-elastic modelling (Zefyros turbine) | FOWT, spar, eigenfrequency, hydro-elastic |
| 3 | Arivin et al. | 2026 | Autoencoder-based anomaly detection for turbofan engine sensors | Autoencoder with MSE reconstruction detects ~1% anomalies in unlabeled C-MAPSS FD001 data at 95th percentile threshold | Autoencoder, MSE reconstruction error | anomaly-detection, autoencoder, sensor-data |
| 4 | Berdugo et al. | 2023 | SeaFEM-OpenFAST coupled tool for floating offshore multi-wind turbines | Coupled aero-hydro-servo-elastic framework demonstrated on W2Power dual-turbine platform | SeaFEM + OpenFAST coupling, time-domain FEM | FOWT, multi-turbine, coupled-simulation |
| 5 | Bernardes et al. | 2024 | Tailings characterization: laser granulometry with ML | ML demonstrates compatibility between laser and conventional granulometry for iron ore tailings classification | Machine learning, laser granulometry vs. sieve-hydrometer | tailings, granulometry, ML, mining |
| 6 | Bull et al. | 2025 | Probabilistic digital-twin-informed risk-based inspection planning for OWT | Ensemble of probabilistic digital twins (not single most-likely) improves risk-based inspection decisions via SHM-OMA integration | Bayesian updating, OMA, probabilistic digital twin, RBI | SHM, digital-twin, OWT, inspection |
| 7 | Cerfontaine et al. | 2023 | Silent piling for offshore jacket foundations in sand: DEM and centrifuge | Screw piles can be installed by rotary jacking at low reaction force; DEM matches centrifuge data for advancement ratio effects | DEM, geotechnical centrifuge | piling, screw-pile, noise-mitigation, offshore |
| 8 | Cho | 2024 | Optimization of semisubmersible FOWT for Oregon coast via OpenFAST | 30% increase in cylinder spacing reduces stress factors by 1.2-3.2% for OC4 semisubmersible | OpenFAST simulation, parametric optimization | FOWT, semisubmersible, optimization |
| 9 | Drangsfeldt | 2023 | Vibration-based SHM of a laboratory-scale wind turbine blade | Combined implicit-explicit EOV mitigation (PCA + Bayesian regression) shows superior damage detection using non-physical DSFs | VAR models, PCA, Bayesian regression, Mahalanobis distance | SHM, wind-turbine-blade, EOV, damage-detection |
| 10 | Escala | 2018 | The principle of similitude in biology: from allometry to dimensionally homogeneous laws | Buckingham Pi theorem unifies metabolic rate allometry across mammals, birds, invertebrates into a single dimensionally homogeneous formula | Dimensional analysis, Pi theorem | similitude, dimensional-analysis, allometry |
| 11 | Guan & Wang | 2024 | Data-driven simulation of multivariate cross-correlated geotechnical random fields from sparse measurements | Joint sparse representation directly generates 2D cross-correlated random fields from sparse data without explicit correlation parameters | Sparse representation, random field simulation | geotechnics, random-field, sparse-data, reliability |
| 12 | Gunasekaran et al. | 2025 | Real-time soil fertility analysis and crop prediction using ML/DL | ML and DL models integrating soil metrics and meteorological data enable real-time soil fertility prediction for sustainable agriculture | Machine learning, deep learning | soil, agriculture, ML, prediction |
| 13 | Gomez et al. | XXXX | Vibration-based early detection of blade ice accumulation using Extended Isolation Forest | EIF on triaxial accelerometer data successfully detects ice accumulation on blades using healthy-condition training only | Extended Isolation Forest, vibration analysis, unsupervised | SHM, ice-detection, blade, anomaly-detection |
| 14 | Henneberg & Richards | 2022 | Implementing a rapid deployment bridge scour monitoring system in Colorado | Sonar-based rapid-deploy scour monitoring system tested on two Colorado bridges; practical deployment lessons documented | Sonar instrumentation, field monitoring | scour, bridge, monitoring, field-deployment |
| 15 | Jatoliya et al. | 2024 | Scour depth prediction using ML for offshore tripod foundations | ANN-PSO achieves R2=0.99 for scour depth prediction around tripods, outperforming standalone ANN and ANFIS | ANN, ANFIS, ANN-PSO, 99 experimental data points | scour, tripod, offshore, ML |
| 16 | Khatti & Kontoni | 2025 | Assessment of bearing capacity of concrete piles using bio/swarm-optimized ANN | ABC-optimized ANN achieves >95% accuracy for pile bearing capacity in alluvial soils (194 data points) | ANN with PSO/HHO/GWO/GA/ABC optimization | pile, bearing-capacity, ANN, optimization |
| 17 | Lee & Lin | XXXX | Deep learning autoencoder for outlier detection in PK concentration-time data | Autoencoder handles variable-length time-series PK data for outlier detection, overcoming CWRES and visual inspection limitations | Autoencoder, pharmacokinetics | autoencoder, outlier-detection, pharmacokinetics |
| 18 | Lima et al. | 2021 | Operational modal analysis and structural health monitoring | OMA enables periodic modal parameter identification under ambient excitation for SHM of civil structures; recent advances in uncertainty quantification | OMA, stochastic subspace identification | OMA, SHM, modal-analysis, review |
| 19 | Liu & Chertkov | 2024 | Anomalous response of FOWT to wind and waves | MCMC simulation (10,000 trials) differentiates short- and long-correlated anomalies in FOWT response; wind dominates long-term pitch anomalies | MCMC, reduced-order Betti model, TLP FOWT | FOWT, extreme-events, anomaly, MCMC |
| 20 | Liu et al. | 2017 | Nonlinear FEA of limit state and pressure of dented X60 pipeline | Parametric FEA yields engineering-applicable limit pressure prediction model for dented X60 pipe | 3D nonlinear FEM (ABAQUS), regression | pipeline, dent, FEA, limit-pressure |
| 21 | Macias-Amador et al. | 2025 | Risk assessment model for OWF decommissioning | Process-risk delays can exceed 20% of project duration; availability-related events are most significant discrete risks | Monte Carlo simulation, quantitative risk analysis (ISO 31000) | decommissioning, OWF, risk, Monte-Carlo |
| 22 | Michalewicz et al. | 2025 | INFUSSE: integrating protein sequence embeddings with structure via graph-based DL | Combined sequence (LLM) + structure (GCN) model improves B-factor prediction for antibodies, especially at disordered regions | Graph convolutional network, protein LLM embeddings | deep-learning, protein, graph-network |
| 23 | Molla & Gorems | 2022 | Impact of land use on spatial variability of soil physicochemical properties (Ethiopia) | Natural forest soils have highest organic carbon, total nitrogen, and microbial biomass vs. plantation and agricultural land | Kriging interpolation, GIS, field sampling | soil, land-use, GIS, spatial-variability |
| 24 | Marin-Moreno et al. | 2024 | Physics-informed ML to define seismic velocities and porosity from CPT data | DNN predicts shear wave velocity from CPT with MAE=55 m/s; enables centimetre-scale porosity and Vp resolution | Deep neural network, dynamic poroelasticity | offshore-wind, CPT, DNN, site-investigation |
| 25 | Otoo et al. | 2023 | Numerical investigation of scour around monopile foundation | Finer sediment increases scour susceptibility; flow velocity and period significantly affect scour depth around monopiles | 3D RANS with RNG k-epsilon, sediment transport | scour, monopile, CFD, sediment |
| 26 | Polyakova | 2023 | Diagnosis of railway bridge scour by natural vibration frequencies | Vibration-based scour monitoring for 30 railway bridge piers; FE models verified against field measurements via Tensor MS system | Field vibration measurement, FEM, Tensor MS | scour, bridge, vibration, railway |
| 27 | Rahman | 2025 | GIS-based allowable bearing capacity thematic maps for Bogura District | SPT-based bearing capacity maps from 255 boreholes show clay-dominated shallow depths with BC<73 kN/m2 at 1.5m | SPT, GIS thematic mapping | bearing-capacity, GIS, SPT, Bangladesh |
| 28 | Schmidt et al. | 2025 | Kriging meta-models for damage equivalent load assessment of idling OWT | Kriging with 2000 training points approximates idling fatigue loads acceptably; two extra input parameters needed vs. normal operation | Kriging surrogate modelling | OWT, fatigue, Kriging, meta-model, idling |
| 29 | Schmidt et al. | 2026 | Lifetime reassessment of OWT using Kriging meta-models for different operating conditions | Kriging meta-models maintain high accuracy while drastically reducing computational cost vs. ~1M aeroelastic simulations | Kriging surrogate, lifetime reassessment, IEC 61400-3 comparison | OWT, lifetime-extension, Kriging, fatigue |
| 30 | Shafiee et al. | 2024 | Stability of subsea tunnels using FELA and ANFIS | ANFIS outperforms multiple linear regression for predicting required internal pressure of subsea tunnels in Tresca material | Finite element limit analysis, ANFIS | tunnel, subsea, stability, ANFIS |
| 31 | Stuyts | 2024 | ML tools for offshore site investigations (plenary lecture) | Supervised and unsupervised ML can improve geotechnical parameter selection; no consensus on best practices yet; LLM potential discussed | ML overview, CPT databases, offshore wind farm data | site-investigation, ML, geotechnics, offshore-wind |
| 32 | Belhaouate et al. | 2025 | ANN-based prediction of load-bearing capacity in earthen construction (Morocco) | Shallow ANN architectures outperform deeper ones for predicting earthen construction load-bearing capacity | ANN architecture comparison | earthen-construction, ANN, bearing-capacity |
| 33 | Dieng et al. | 2022 | Heat transfer in dynamic frequency regime: equivalent dynamic impedance of kapok material | Characterization of kapok thermal insulation via 3D dynamic frequency heat transfer model; Bode diagrams of thermal impedance | 3D heat equation, electrical-thermal analogy | thermal-insulation, kapok, heat-transfer |
| 34 | Mokashi & Hirpurkar | 2019 | Hydraulic scaling and similitude from model to prototype | Shield's parameter used to determine prototype sediment diameter (d50=41.43mm) from undistorted flume model (d50=0.828mm) | Shield's parameter, similitude theory | similitude, hydraulic-scaling, sediment |
| 35 | Usowicz & Lipiec | 2021 | Spatial variability of saturated hydraulic conductivity at commune scale | SHC linked to sand content, organic carbon, bulk density via semivariogram and cross-semivariogram analysis at 140 km2 scale | Geostatistics, kriging, cross-semivariogram | soil, hydraulic-conductivity, geostatistics |
| 36 | Wang et al. | 2022 | Blade damage detection of small wind turbine via fluid-heat-solid coupling | Fluid-heat-solid coupled model in COMSOL reduces max error by 7.46% vs. natural convection approximation for IR blade inspection | COMSOL multiphysics, infrared thermography | blade, damage-detection, IR, COMSOL |
| 37 | Wang et al. | 2025 | Finite-frequency LPV H-infinity control for disturbed wind turbine | FF-domain LPV H-infinity control reduces conservatism of MPPT control under wind disturbances vs. entire-frequency domain | LPV model, H-infinity control, gain scheduling, LMI | wind-turbine, control, LPV, MPPT |
| 38 | Li, Wu & Hu | 2022 | BBO-MLP neural network to predict CBR of stabilized pond ash | BBO-MLP1 achieves R2=0.9977 for CBR prediction of lime/lime-sludge stabilized pond ash | Biogeography-based optimization + MLP | CBR, pond-ash, neural-network, stabilization |
| 39 | Zhanfang & Tuo | XXXX | Enhancing wind turbine blade damage detection with YOLO-Wind | Enhanced YOLOv8n with DWConv, MBConv, ECA achieves 83.9% mAP@0.5, +2.3% over baseline on DTU blade dataset | YOLOv8n, computer vision, depthwise separable convolutions | blade, damage-detection, YOLO, computer-vision |
| 40 | Zhao et al. | XXXX | DFIG impedance reshaping via dynamic rotor current compensation for mid-frequency stability | Rotor current compensation reshapes DFIG impedance to suppress PLL-induced negative resistance in mid-frequency band under weak grid | Impedance modelling, MIMO, LMI | DFIG, wind-power, stability, impedance |
CONSENSUS¶
-
ML/AI as universal surrogate: Across geotechnics, offshore wind, scour prediction, and structural health monitoring, machine learning models (ANN, ANFIS, Kriging, autoencoders) consistently outperform traditional empirical and regression methods. Papers 15, 16, 28, 29, 30, 31, 32, 38 all report R2 > 0.95 or equivalent improvements.
-
Vibration-based SHM is maturing: Multiple studies (6, 9, 13, 18, 26) confirm that modal parameters extracted from operational vibration data (OMA) reliably detect damage, scour, and ice accumulation. The field consensus is that unsupervised or one-class approaches trained only on healthy data are practical and scalable.
-
Scour remains a critical offshore risk: Papers 14, 15, 25, 26 converge on scour as a dominant threat to both offshore wind and bridge foundations. Flow velocity, sediment size, and wave-current interaction are universally recognized as governing parameters.
-
Coupled multi-physics modelling is essential for FOWT: Studies 2, 4, 8, 19 agree that rigid-body or single-discipline assumptions introduce significant errors (up to 37% in eigenfrequency). Full hydro-servo-aero-elastic coupling is now considered mandatory for reliable FOWT design.
-
Kriging/surrogate models enable lifetime reassessment: The Schmidt pair (28, 29) establishes that Kriging meta-models can replace millions of aeroelastic simulations for fatigue assessment with acceptable accuracy, covering both normal operation and idling states.
DEBATES¶
-
Optimal ML architecture depth: Belhaouate et al. (32) find shallow ANNs outperform deeper ones for earthen construction, while Khatti & Kontoni (16) and Jatoliya et al. (15) show that hybrid optimization (ABC, PSO) on standard ANNs achieves top performance. The question of when to prefer shallow vs. deep vs. hybrid-optimized architectures remains unresolved.
-
Single vs. ensemble digital twins: Bull et al. (6) argue for probabilistic ensembles of digital twins rather than single best-fit models. This challenges the prevailing industry practice of deterministic digital twins and raises questions about computational feasibility at scale.
-
Dimensional homogeneity in empirical scaling laws: Escala (10) provocatively argues that most allometric laws violate the similitude principle. This debate extends to geotechnical and hydraulic scaling (34) where model-to-prototype similarity often relies on incomplete dimensional analysis.
-
Data requirements for reliable ML in geotechnics: Stuyts (31) and Marin-Moreno (24) highlight that ML model reliability depends critically on training data coverage and soil type diversity. No consensus exists on minimum dataset sizes for different geotechnical prediction tasks.
GAPS¶
-
Idling and transient states under-studied: Schmidt et al. (28) note that meta-models for idling OWT have received almost no attention compared to normal operation, despite idling contributing significantly to lifetime fatigue.
-
Decommissioning risk quantification is nascent: Macias-Amador et al. (21) observe that simulation tools for OWF decommissioning are still being developed; process-risk integration with weather-window analysis is largely absent from the literature.
-
Cross-correlated random fields from sparse data: Guan & Wang (11) address a recognized gap -- most random field simulators require known correlation structures, yet real site investigations provide only sparse measurements. Data-driven generators remain rare.
-
Field validation of SHM under real environmental variability: Drangsfeldt (9) acknowledges that laboratory results degrade in uncontrolled environments. Large-scale field validation datasets for vibration-based SHM of wind turbine blades are still lacking.
-
Integration of geophysical and geotechnical data at centimetre scale: Marin-Moreno (24) identifies the need to align CPT and seismic data at high resolution; current practice averages over incompatible depth ranges.
METHODS¶
- Numerical: SPH (1), DEM (7), RANS CFD (25), nonlinear FEA/ABAQUS (20), FELA (30), COMSOL multiphysics (36)
- Surrogate/ML: Kriging (28, 29), ANN/ANN-PSO/ANFIS (15, 16, 30, 32, 38), DNN (24), autoencoders (3, 17), Extended Isolation Forest (13), YOLOv8n (39), BBO-MLP (38)
- Statistical/Probabilistic: MCMC (19), Monte Carlo (21), Bayesian regression (9), geostatistics/kriging interpolation (23, 27, 35), random field simulation (11)
- Experimental/Field: Geotechnical centrifuge (1, 7), in-situ vibration/accelerometer (2, 9, 13, 26), sonar (14), SPT campaigns (27), infrared thermography (36)
- Control: LPV H-infinity (37), impedance reshaping (40), gain scheduling (37)
- Dimensional analysis: Pi theorem (10), Shield's parameter (34)
BENCHMARKS¶
| Domain | Benchmark/Dataset | Papers Using It |
|---|---|---|
| Turbofan anomaly detection | NASA C-MAPSS FD001 | 3 |
| FOWT simulation | NREL 5MW / OC4 semisubmersible | 8, 19 |
| Spar FOWT | Zefyros 2.3MW (in-situ data) | 2 |
| Blade damage detection | DTU blade dataset | 39 |
| Scour prediction (tripod) | 99 experimental data points from literature | 15 |
| Pile bearing capacity | 194 literature data points | 16 |
| OWT lifetime fatigue | IEC 61400-3 standard load cases (~1M simulations) | 28, 29 |
| Geotechnical CPT-Vs correlation | 5284 public-domain CPT instances | 24 |
| Bridge scour monitoring | Colorado bridges F-05-R and P-01-G | 14 |
| Railway bridge scour | 30 piers, Tensor MS measurements | 26 |