Batch 02 Agent 3: Literature Synthesis (positions 281-320)¶
Paper Extraction Table¶
| # | Author(s) | Year | Title | Core Finding | Method | Tags |
|---|---|---|---|---|---|---|
| 1 | Erinmwingbovo et al. | 2020 | Dynamic impedance spectroscopy of LiMn2O4 thin films | Two-stage intercalation process for Li-ion insertion identified via dynamic multi-frequency analysis | Dynamic impedance spectroscopy, DMFA | battery, thin-film, electrochemistry |
| 2 | Frederik et al. | 2020 | Periodic dynamic induction control of wind farms | Periodic sinusoidal variation of induction factor yields wind farm power gains; first wind-tunnel proof of concept | Simulation + scaled wind tunnel experiments | wind-farm-control, wake, DIC |
| 3 | Fu & Liu | 2020 | Fin configuration effects of dynamically installed anchors in clay | Short-wide-rectangular fins at rear of shaft achieve deeper embedment and higher holding capacity | Centrifuge modelling + theoretical approach | offshore-anchor, DIA, clay, centrifuge |
| 4 | Fukami et al. | 2020 | CNN-based hierarchical autoencoder for nonlinear mode decomposition of fluid fields | Hierarchical AE preserves contribution order of latent vectors; ordered autoencoder mode family handles turbulence | CNN autoencoder, hierarchical training | autoencoder, fluid-dynamics, ROM |
| 5 | Gaikwad et al. | 2020 | Mathematical modelling of material removal rate using Buckingham Pi theorem | Dimensionless groups derived via Pi theorem predict EDM material removal rate | Buckingham Pi dimensional analysis | dimensional-analysis, manufacturing |
| 6 | Gao et al. | 2020 | Deep learning replacing FDM for in-situ stress prediction | ES-Caps-FCN achieves MSE of 0.06% vs 0.62% for DNN; faster than conventional FDM | Capsule network + FCN (deep learning) | deep-learning, geotechnical, stress |
| 7 | Ghaemi & Zeraatgar | 2020 | Hull-propeller-engine interactions in regular waves | Conventional added-resistance method cannot capture dynamic system behaviour; governor/limiters affect fuel consumption | Mathematical modelling + experiment | ship-propulsion, wave-loading, dynamics |
| 8 | Gill et al. | 2020 | Offshore wind development effects on fish and fisheries | OWFs act as artificial reefs but also displace fishing; evidence gaps remain for population-level impacts | Review, stakeholder analysis | OWF, fisheries, environmental-impact |
| 9 | Hammond & Pokorny | 2020 | Gap size effects on natural regeneration in beech-spruce forest | Gap size is significant for beech regeneration; soil temperature/moisture control species-specific outcomes | Field monitoring, statistical analysis | forestry, ecology, soil-conditions |
| 10 | Wu (Haoyu) et al. | 2020 | Transient response of TLP-type FOWT under tendon failure | Tendon failure causes large transient surge/heave motions; remaining tendons may exceed design loads | Coupled aero-hydro-servo-elastic simulation | FOWT, TLP, tendon-failure, transient |
| 11 | Hart et al. | 2020 | Review of wind turbine main bearings | MB failure rates up to 30% over 20-year lifetime; MBs are under-studied relative to gearboxes | Comprehensive review | wind-turbine, bearing, reliability, review |
| 12 | Homaei | 2020 | Inelastic soil-foundation interface effects on seismic demand | Inelastic soil modelling reduces superstructure demand by >50% via shrinking-dominated rocking | Winkler model, SDOF superstructure, 20 ground motions | SSI, seismic, foundation, Winkler |
| 13 | Hou & Xia | 2020 | Review of vibration-based damage identification 2010-2019 | ML/AI techniques are emerging dominant tools; environmental variability remains a major challenge | Comprehensive review (modal, signal, FEM, ML, Bayesian) | SHM, damage-identification, review |
| 14 | Hsu et al. | 2020 | Wind turbine fault diagnosis via SPC and ML | Decision tree (92.7%) and random forest (92.0%) accurately predict turbine maintenance needs from 2.8M sensor records | SPC, DBSCAN, random forest, decision tree | wind-turbine, fault-diagnosis, ML |
| 15 | Hu & Chang | 2020 | Optimized Buckingham Pi for wind tunnel testing | Optimized dimensional analysis method improves physical quantity selection for wind tunnel similitude | Buckingham Pi theorem, CFD verification | dimensional-analysis, wind-tunnel |
| 16 | Huang et al. | 2020 | Quantum autoencoder for lossless compression | Lossless compression possible when max linearly independent input vectors <= latent space dimension | Quantum circuit, machine learning optimization | quantum-computing, autoencoder |
| 17 | Huang et al. | 2020 | Seismic resistance of utility tunnel in saline soil with new cementitious materials | New slag-gypsum-lime-magnite composite reduces tunnel displacement/acceleration under seismic loads | FEM + shaking table tests | seismic, foundation-reinforcement, FEM |
| 18 | Hutchison et al. | 2020 | EMF interactions between OWF cables and resource species | Knowledge gaps remain on population-level EMF impacts; no policies or regulations exist for EMFs | Review of lab and field studies | OWF, EMF, marine-ecology |
| 19 | Iwicki & Przewlocki | 2020 | 3D FEM of monopile and gravity foundations for OWT | Nonlinear static/dynamic FEM comparison of monopile vs gravity foundation performance under combined loading | 3D FEM (soil + tower + blades) | OWT-foundation, monopile, gravity, FEM |
| 20 | Kumar & Majid | 2020 | Renewable energy for sustainable development in India | India is a top renewable energy market; policy/investment barriers remain despite strong government push | Policy review | renewable-energy, India, policy |
| 21 | Jin et al. | 2020 | Edge-based strain smoothing particle FEM for large deformations | ES-PFEM handles large deformation geotechnical problems more stably than standard PFEM | Particle FEM with edge-based smoothing | numerical-methods, large-deformation, geotech |
| 22 | Jin & Yin | 2020 | Enhanced backtracking search for soil parameter identification | Improved BSA algorithm identifies soil parameters more efficiently than standard optimization | Backtracking search algorithm | soil-parameters, optimization |
| 23 | Johlas et al. | 2020 | LES of FOWT wakes for spar vs semi-submersible | Floating turbine wakes deflect upward vs fixed-turbine wakes due to mean platform tilt | LES with actuator line model (SOWFA + OpenFAST) | FOWT, wake, LES, CFD |
| 24 | Jonkman et al. | 2020 | Substructure flexibility in OpenFAST for FOWTs | Implemented member-level load capabilities and flexible substructure modelling in OpenFAST | OpenFAST framework extension | FOWT, simulation-tool, OpenFAST |
| 25 | Karimezadeh et al. | 2020 | Dynamic behaviour of unsaturated sand across strain ranges | Gmax increases beyond air-entry value; constrained modulus shows non-monotonic trend with suction | Modified cyclic simple shear + bender elements | unsaturated-soil, dynamic-properties, suction |
| 26 | Kerpen et al. | 2020 | Wave-induced microplastic distribution in surf zone | Wave action redistributes microplastic particles in the surf zone | Experimental (wave flume) | coastal, microplastic, environmental |
| 27 | Kim & Chung | 2020 | Multi-modal stacked denoising AE for missing healthcare data | Multi-modal stacked DAE achieves 92.2% accuracy at 25% missing data with fewer parameters than single-modal | Stacked denoising autoencoder | autoencoder, missing-data, healthcare |
| 28 | Koh et al. | 2020 | Cyclic response of suction caisson anchor in calcareous silt | Cyclic loading at varying mean loads (30-70%) tested via centrifuge; load inclination affects capacity | Beam centrifuge testing | suction-caisson, calcareous-silt, cyclic |
| 29 | Kubo | 2020 | Hybrid ML + GMPE predictor for ground-motion intensity | Hybrid ML-GMPE approach reduces underestimation of strong motions caused by data bias | Random forest + conventional GMPE | ground-motion, ML, seismology, hybrid |
| 30 | Kulczykowski | 2020 | Skirted foundation in sand under rapid uplift | Displacement rate significantly affects uplift resistance magnitude but not stress-displacement curve shape | 1g model tests | suction-caisson, uplift, sand, physical-model |
| 31 | Kulev et al. | 2020 | Non-resilient OWF behaviour from cyber-physical attacks | Cyber-physical attacks can cause cascade failures exceeding mechanical/electrical limits without timely recovery | Functional modelling, threat scenarios | OWF, cybersecurity, resilience |
| 32 | Kumar et al. | 2020 | Reliability of centrifuge modelling for liquefaction effects on shallow foundations | Nonuniformity in centrifuge models affects reliability of liquefaction-induced settlement predictions | Centrifuge modelling + reliability analysis | liquefaction, centrifuge, reliability |
| 33 | Lambinet & Sharif Khodaei | 2020 | Damage detection on composite patch repair under environmental effects | Environmental conditions (temperature, moisture) affect damage detection/localization accuracy | SHM, guided waves | SHM, composite, environmental-effects |
| 34 | Lemmer et al. | 2020 | Multibody modelling for concept-level FOWT design | Reduced-order multibody model completes 1-hour simulations in ~25 seconds; suitable for optimization | Newton-Euler multibody + frequency domain | FOWT, reduced-order-model, multibody |
| 35 | Li et al. | 2020 | Coupled dynamic response of floating multi-purpose platform | Rigid-body hypothesis confirmed feasible for support structure; structural vibration modes not excited by wave/wind | Aero-hydro-servo-elastic coupled analysis + FEM | multi-purpose-platform, FOWT, WEC |
| 36 | Liguori et al. | 2020 | Indoor environment time-series reconstruction via AE | Autoencoders outperform polynomial interpolation for gap-filling (RMSE: 0.42C, 1.30%, 78 ppm CO2) | Feed-forward/CNN/LSTM denoising autoencoders | autoencoder, building-data, missing-data |
| 37 | Liu et al. | 2020 | SMF decomposition for OWT time-frequency analysis | Single mode function decomposition overcomes HHT mode mixing for closely-spaced OWT frequencies | Signal decomposition, state-space model | SHM, OWT, time-frequency, signal-processing |
| 38 | Ly & Pham | 2020 | Soil shear strength prediction via SVM | SVM model predicts shear strength from direct shear test data, reducing lab testing cost | Support vector machine | soil-mechanics, ML, shear-strength |
| 39 | Marques et al. | 2020 | Constructive effect and SSI in tall buildings on sand | Ignoring incremental construction effects and SSI leads to designs violating stability code requirements | Multi-spring-mass SSI model | SSI, tall-building, shallow-foundation |
| 40 | Mayall | 2020 | Flume testing of OWT dynamics with scour/scour protection | (File empty - no content extracted) | Flume tank experiments | OWT, scour, physical-model |
1. CONSENSUS¶
Several convergent themes emerge across these 40 papers (all from 2020):
ML/AI as viable surrogates for physics-based methods. Multiple studies confirm that machine learning models can replace or augment conventional numerical methods: deep learning substituting finite difference for stress prediction (Gao), hybrid ML+GMPE outperforming either alone for ground motion (Kubo), SVM for soil shear strength (Ly), random forest/decision tree for wind turbine fault diagnosis (Hsu), and autoencoders for data reconstruction/compression (Fukami, Kim, Liguori, Huang-quantum). The consistent finding is that ML achieves comparable or superior accuracy at lower computational cost, though data bias and missing data remain limiting factors.
Autoencoders are a versatile tool across domains. From hierarchical mode decomposition of fluid fields (Fukami), to missing healthcare data imputation (Kim), indoor environment gap-filling (Liguori), and quantum information compression (Huang), autoencoders are broadly adopted. All studies confirm autoencoders outperform simpler baselines.
Offshore wind turbine foundations and structural integrity demand coupled analysis. Papers on FOWT dynamics (Haoyu-Wu, Lemmer, Li, Jonkman, Johlas), OWT foundations (Iwicki, Mayall), and bearing reliability (Hart) all agree that simplified decoupled models are insufficient. Coupled aero-hydro-servo-elastic analysis is now the standard expectation.
Soil-structure interaction materially affects structural response. Homaei, Marques, and Karimezadeh all demonstrate that ignoring nonlinear soil behaviour or SSI leads to erroneous (usually over-conservative or under-conservative) structural predictions.
Dimensional analysis (Buckingham Pi) retains utility in physical modelling. Both Hu and Gaikwad confirm that Pi theorem-based dimensional analysis remains essential for designing wind tunnel experiments and manufacturing process models, though quantity selection requires optimization.
2. DEBATES¶
Steady-state vs dynamic induction control for wind farms. Frederik et al. show that static induction control yields limited-to-no gains, while periodic dynamic induction control (DIC) produces measurable power increases. However, the increase in damage equivalent loads on the excited turbine, and whether DIC scales to large farms, remains contested.
Elastic vs inelastic soil modelling at foundation interfaces. Homaei shows >50% demand reduction with inelastic modelling, but conventional practice still uses elastic Winkler models. The debate centers on whether the computational complexity of nonlinear soil models is justified for routine design versus only for performance-based assessment.
ML-only vs hybrid approaches for geoscience prediction. Kubo explicitly demonstrates that pure ML predictors underestimate rare strong events due to data imbalance, advocating hybrid ML+physics models. This tension between data-driven and physics-informed approaches pervades the batch.
Population-level vs individual-level impacts of OWF on marine species. Gill and Hutchison both highlight that while individual EMF and habitat effects are documented, translating these to population-level fisheries impacts remains unresolved, creating regulatory uncertainty.
3. GAPS¶
- Floating turbine wake interactions in farm-scale arrays are studied only for isolated turbines (Johlas); multi-turbine floating farm wake modelling is absent.
- Main bearing failure root cause remains poorly understood despite 30% failure rates (Hart); condition monitoring techniques specific to main bearings are underdeveloped.
- Unsaturated soil dynamic properties beyond the residual suction zone are rarely measured (Karimezadeh); most dynamic soil databases cover only saturated conditions.
- Cybersecurity resilience of OWFs is identified as critical (Kulev) but no validated defence frameworks or standards exist.
- Centrifuge model nonuniformity affects reliability of liquefaction studies (Kumar) but no standardized correction protocols are proposed.
- Autoencoder mode ordering for turbulent flows at higher Reynolds numbers is untested (Fukami).
- Environmental compensation of SHM signals under varying temperature/moisture for composite repairs (Lambinet) lacks robust frameworks.
4. METHODS¶
Dominant experimental methods: - Centrifuge modelling: Fu (DIA in clay), Koh (suction caisson in calcareous silt), Kumar (liquefaction) - 1g physical model tests: Kulczykowski (skirted foundation uplift) - Wind tunnel experiments: Frederik (DIC) - Shaking table: Huang (utility tunnel seismic)
Dominant numerical methods: - FEM: Iwicki (OWT foundations), Huang (seismic), Li (multi-purpose platform) - Coupled aero-hydro-servo-elastic simulation: Haoyu-Wu (TLP FOWT), Jonkman (OpenFAST), Lemmer (multibody) - LES/CFD: Johlas (FOWT wakes) - Particle FEM: Jin (large deformation)
Dominant ML methods: - Autoencoders (CNN, stacked denoising, LSTM, hierarchical): Fukami, Kim, Liguori, Huang - Random forest / decision tree: Hsu, Kubo - SVM: Ly - Deep learning (capsule networks): Gao - Optimization algorithms: Jin-Yin (backtracking search)
Signal processing: Liu (single mode function decomposition replacing HHT)
5. BENCHMARKS¶
| Domain | Metric | Value | Source |
|---|---|---|---|
| In-situ stress prediction | MSE | 0.06% (ES-Caps-FCN) vs 0.62% (DNN) | Gao 2020 |
| Wind turbine fault diagnosis | Accuracy | 92.7% (decision tree), 92.0% (random forest) | Hsu 2020 |
| Missing healthcare data (25%) | Accuracy | 0.922 (multi-modal stacked DAE) | Kim 2020 |
| Indoor temperature reconstruction | RMSE | 0.42 deg-C | Liguori 2020 |
| Indoor humidity reconstruction | RMSE | 1.30% | Liguori 2020 |
| Indoor CO2 reconstruction | RMSE | 78.41 ppm | Liguori 2020 |
| Main bearing failure rate | 20-year rate | up to 30% | Hart 2020 |
| SSI demand reduction | Superstructure demand | >50% reduction with inelastic soil | Homaei 2020 |
| FOWT concept simulation speed | Wall-clock time | ~25 sec per 1-hour simulation | Lemmer 2020 |
| Suction caisson cyclic loading | Mean load tested | 30%, 50%, 70% of monotonic capacity | Koh 2020 |