pith. machine review for the scientific record. sign in

arxiv: 2604.09112 · v1 · submitted 2026-04-10 · 💻 cs.IR

Recognition: unknown

Hybrid Cold-Start Recommender System for Closure Model Selection in Multiphase Flow Simulations

S. H\"ansch , A. Sajdokov\'a , A. R\k{e}bowski , F. Mi\v{s}ka\v{r}\'ik , K. Ramakrishna , F. Schlegel , V. Ryb\'a\v{r} , R. Alves
show 1 more author
P. Kord\'ik
Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:49 UTC · model grok-4.3

classification 💻 cs.IR
keywords cold-start recommender systemsclosure model selectionmultiphase CFDmatrix completionhybrid recommendationmodel selectionscientific decision supportCFD simulations
0
0 comments X

The pith

A hybrid recommender using case metadata and matrix completion selects closure models for new multiphase CFD simulations with lower regret than popularity or expert baselines.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper treats closure model selection in multiphase flow simulations as a cold-start recommender problem where each new CFD case needs a model combination chosen from many options. It builds a hybrid system that first measures similarity between cases using their descriptive metadata features and then applies matrix completion to infer performance from past simulations of similar cases. This produces recommendations for entirely unseen flow scenarios without requiring prior runs on those exact cases. Evaluation across 136 cases and 100 model combinations shows the method improves ranking quality and cuts a domain-specific regret measure relative to both popularity-based and expert-designed baselines. A reader would care because poor model choices in CFD lead to inaccurate results or wasted computation on expensive simulations.

Core claim

The central claim is that the hybrid recommender framework, which integrates metadata-driven case similarity with collaborative inference via matrix completion, enables effective case-specific recommendations for closure model combinations on entirely new CFD cases and reduces regret across varying data sparsities, outperforming both popularity-based and expert-designed reference models.

What carries the argument

The hybrid recommendation framework that combines metadata-driven case similarity and collaborative inference via matrix completion.

If this is right

  • Case-specific model recommendations become feasible for new CFD cases using only their descriptive features and historical data from similar cases.
  • Regret relative to the per-case optimum decreases across different levels of data sparsity in the simulation history.
  • Recommendation quality improves on ranking-based metrics compared with simpler reference strategies.
  • The approach supports complex scientific decision tasks that involve expensive evaluations and structured metadata with limited prior observations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same hybrid structure could transfer to model selection problems in other high-cost simulation domains that have descriptive case metadata.
  • Enriching the metadata features with additional flow diagnostics might further tighten the link between case similarity and model performance.
  • The regret metric could be expanded to weight computational cost or prediction uncertainty alongside accuracy loss.

Load-bearing premise

Descriptive metadata features of CFD cases sufficiently capture the similarities that determine which closure model combinations will perform well, enabling reliable generalization to new flow scenarios.

What would settle it

A new collection of flow scenarios where the hybrid recommender's top-ranked model combinations produce no lower average regret than the popularity-based or expert-designed baselines would falsify the central claim.

Figures

Figures reproduced from arXiv: 2604.09112 by A. R\k{e}bowski, A. Sajdokov\'a, F. Mi\v{s}ka\v{r}\'ik, F. Schlegel, K. Ramakrishna, P. Kord\'ik, R. Alves, S. H\"ansch, V. Ryb\'a\v{r}.

Figure 1
Figure 1. Figure 1: The ground-truth performance matrix constructed from 100 items evaluated for 136 CFD cases, with the performance values indicated by colour coding. Case-item interactions that lead to numerical instabilities present themselves in dark blue (performance value of zero). Features are mostly distinct for one CFD experiment forming experiment-specific clusters. In particular, categorical features are identical … view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of feature-based and performance-based representations of the 136 CFD cases, visualized in two dimensions using multidimensional scaling (MDS). Colors indicate the 17 distinct CFD experiments. 𝑛) represents item 𝑖, column 𝑗 (out of 𝑚) represents case 𝑗 and the entry (𝑖, 𝑗) of 𝑅 the performance of item 𝑖 in case 𝑗 that evaluates the predictive quality of a particular closure model for a given cas… view at source ↗
Figure 3
Figure 3. Figure 3: Schematic of the recommender system pipeline for CFD model selection for new CFD cases. Formally, for a query case 𝑞 with feature vector 𝑥𝑞 , we identify the set of 𝑘 nearest neighbours (𝑘-NN) 𝑘 (𝑞) = arg min ⊂ ||=𝑘 ∑ 𝑖∈ 𝑑(𝑥𝑞 , 𝑥𝑖 ), (4) where  denotes the set of available (historical) CFD cases and 𝑑(⋅, ⋅) is the chosen distance metric. We explore different values of 𝑘 and distance metrics in cross-… view at source ↗
Figure 4
Figure 4. Figure 4: Relevant items achieving a performance within a 0.05 threshold of the per-case top item performance, with their ranking indicated by colour coding. 4.2. Evaluation metrics Based on the relevant items, we make use of the ranking metric reciprocal rank (RR) at the top k items, which is generally computed as: 𝑅𝑅@𝑘 = { 1 𝑟𝑎𝑛𝑘𝑖 if relevant item 𝑖 is among the top k 0 if no relevant item is in the top k. (6) Thi… view at source ↗
Figure 5
Figure 5. Figure 5: Flow diagram of the nested cross-validation procedure. The inner cross-validation loop (right) is used for 𝑘-NN hyperparameter selection and feeds back into the main test loop after selecting the optimal configuration. 5. Results 5.1. Optimised 𝑘-NN configurations Distinct optimised 𝑘-NN configurations are obtained for each sparsity level and for each individual CFD test experiment [PITH_FULL_IMAGE:figure… view at source ↗
Figure 6
Figure 6. Figure 6: Reciprocal rank (RR) per test experiment (averaged over sub-cases), evaluated on a 75% sparse performance matrix with 95% confidence intervals. Panel (a) compares RR@1 across reference model, popular-item, and recommender model, while panel (b) reports the corresponding RR@3 results for the popular-item and the recommender model. As expected, the popularity baseline degrades monotonically with increasing s… view at source ↗
Figure 7
Figure 7. Figure 7: MRR@3 for the popularity baseline from observed entries and after matrix completion compared to the recommender system predictions, plotted over the different sparsity levels explored. At the highest sparsity level (𝑠 = 0.90), the margin between RS and MC narrows, which might reflect the intrinsic difficulty of reliable ranking under extremely limited observations. Nevertheless, even in this regime, the hy… view at source ↗
Figure 8
Figure 8. Figure 8: Regret per experiment (consisting of a certain number of CFD sub-cases) for the reference model, the recommender-predicted and popular best item derived from a 75% sparse matrix with 95% confidence intervals. 5.2.2. Performance regret While the MRR@k metrics quantify the ranking quality, they do not capture the magnitude of performance loss when a sub-optimal model is selected. To address this we additiona… view at source ↗
Figure 9
Figure 9. Figure 9: Average Regret for the popularity baseline from observed entries and after matrix completion compared to the recommender system predictions, plotted over the different sparsity levels. for robust model selection: while matrix completion improves global performance estimates, it lacks the case-specific discrimination enabled by feature-based 𝑘-NN neighbourhood selection. The ablation therefore provides clea… view at source ↗
Figure 10
Figure 10. Figure 10: Validation plots for a case example with high Regret: Experiment ID 11/Case ID 97 with 𝑅𝑒𝑔𝑟𝑒𝑡RS 𝑐 = 0.15 (𝑝 RS 𝑐 = 0.65, 𝑝 ⋆ 𝑐 = 0.80). The recommendation via hybrid recommender system (RS, green dashed line) and the true best item (gray solid line) compared to the experimental validation data (red points). recommended by the recommender model across 100 independent scenarios at a sparsity level of 𝑠 = 0.… view at source ↗
Figure 11
Figure 11. Figure 11: Validation plots for a case example with low Regret: Experiment ID 8/Case ID 49 with 𝑅𝑒𝑔𝑟𝑒𝑡RS 𝑐 = 0.01 (𝑝 RS 𝑐 = 0.81 , 𝑝 ⋆ 𝑐 = 0.82). The recommendation via hybrid recommender system (RS, green dashed line) and the true best item (gray solid line) compared to the experimental validation data (red points). an information-filtering problem, CFD cases are interpreted as users and closure model combinations … view at source ↗
Figure 12
Figure 12. Figure 12: MRR@1 results across different sparsity levels plotted for each individual CFD experiment. Comparison of the popularity baseline, the recommender prediction and the reference item. S. Hänsch et al.: Preprint submitted to Elsevier Page 28 of 27 [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: MRR@3 results across different sparsity levels plotted for each individual CFD experiment. Comparison of the popularity baseline, the recommender prediction and the reference item. S. Hänsch et al.: Preprint submitted to Elsevier Page 29 of 27 [PITH_FULL_IMAGE:figures/full_fig_p029_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Regret results across different sparsity levels plotted for each individual CFD experiment. Comparison of the popularity baseline, the recommender prediction and the reference item. S. Hänsch et al.: Preprint submitted to Elsevier Page 30 of 27 [PITH_FULL_IMAGE:figures/full_fig_p030_14.png] view at source ↗
read the original abstract

Selecting appropriate physical models is a critical yet difficult step in many areas of computational science and engineering. In multiphase Computational Fluid Dynamics (CFD), practitioners must choose among numerous closure model combinations whose performance varies strongly across flow conditions. Sub-optimal choices can lead to inaccurate predictions, simulation failures, and wasted computational resources, making model selection a prime candidate for data-driven decision support. This work formulates closure model selection as a cold-start recommender system problem in a high-cost scientific domain. We propose a hybrid recommendation framework that combines (i) metadata-driven case similarity and (ii) collaborative inference via matrix completion. The approach enables case-specific model recommendations for entirely new CFD cases using their descriptive features, while leveraging historical simulation results from similar cases. The methodology is evaluated on 13,600 simulations across 136 validation cases and 100 model combinations. A nested cross-validation protocol with experiment-level holdout is employed to rigorously assess generalisation to unseen flow scenarios under varying levels of data sparsity. Recommendation quality is measured using ranking-based metrics and a domain-specific regret measure capturing performance loss relative to the per-case optimum. Results show that the proposed hybrid recommender consistently outperforms popularity-based and expert-designed reference models and reduces regret across the investigated sparsities. These findings demonstrate that recommender system methodology can effectively support complex scientific decision-making tasks characterised by expensive evaluations, structured metadata, and limited prior observations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper formulates closure model selection for multiphase CFD simulations as a cold-start recommender problem. It proposes a hybrid approach combining metadata-driven case similarity with matrix completion for collaborative inference, enabling recommendations for new cases. Evaluation uses 13,600 simulations over 136 cases and 100 model combinations, with nested cross-validation employing experiment-level holdout, ranking metrics, and a domain-specific regret measure. Results claim consistent outperformance over popularity-based and expert-designed baselines across sparsity levels.

Significance. If the central claims hold, the work provides a concrete demonstration of recommender-system techniques applied to expensive scientific model selection under data scarcity, with strengths in the experiment-level holdout protocol and regret metric that directly ties recommendations to simulation performance loss. This could inform similar decision-support tasks in computational engineering. However, significance is limited by the unverified assumption that chosen metadata features align with the physics governing model performance, which is load-bearing for generalization claims.

major comments (2)
  1. [Abstract and evaluation protocol] Abstract and evaluation protocol: The claim of reliable generalization to entirely new flow scenarios via case similarity is load-bearing but rests on the untested assumption that descriptive metadata features (e.g., flow conditions) sufficiently encode the physics determining which of the 100 closure combinations perform well. In multiphase CFD, optimal models depend on fine-grained aspects such as bubble size distributions, interfacial forces, and turbulence modulation that may not be captured by high-level metadata. The nested CV with experiment-level holdout tests performance within the observed metadata manifold but provides no direct evidence (e.g., via feature ablation or physics-based similarity checks) that metadata similarity predicts performance similarity outside it. With only 136 cases total, this risks overestimating robustness.
  2. [Methods and results sections] Methods and results sections: The hybrid framework's outperformance over baselines and regret reduction is reported across sparsities, but without explicit details on how metadata features are constructed, normalized, or selected (and whether they were tuned post-hoc), it is difficult to assess if the gains are driven by genuine similarity capture or by the matrix completion component alone. A concrete test—such as reporting correlation between metadata distance and optimal-model overlap—would strengthen the central claim.
minor comments (2)
  1. [Methods] Clarify the exact construction of the 13,600 simulations (e.g., how many per case, how sparsity is induced in the nested CV) and provide a table summarizing the metadata feature set and its dimensionality.
  2. [Results] The abstract states 'reduces regret across the investigated sparsities'—add quantitative regret values and confidence intervals in the main results table for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and detailed review. The comments highlight important aspects of our evaluation protocol and methodological transparency. We address each major comment below and describe the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract and evaluation protocol] Abstract and evaluation protocol: The claim of reliable generalization to entirely new flow scenarios via case similarity is load-bearing but rests on the untested assumption that descriptive metadata features (e.g., flow conditions) sufficiently encode the physics determining which of the 100 closure combinations perform well. In multiphase CFD, optimal models depend on fine-grained aspects such as bubble size distributions, interfacial forces, and turbulence modulation that may not be captured by high-level metadata. The nested CV with experiment-level holdout tests performance within the observed metadata manifold but provides no direct evidence (e.g., via feature ablation or physics-based similarity checks) that metadata similarity predicts performance similarity outside it. With only 136 cases total, this risks overestimating robustness.

    Authors: We agree that the metadata features are high-level descriptors of flow conditions and do not explicitly encode all fine-grained physics such as bubble size distributions or detailed turbulence modulation. The experiment-level holdout in the nested CV protocol evaluates generalization to unseen cases that lie within the sampled metadata manifold, and the domain-specific regret metric directly links recommendations to simulation performance loss. We acknowledge that this does not constitute direct evidence for extrapolation outside the observed manifold. In the revised manuscript we will add an explicit discussion of these assumptions and limitations, including the rationale for the chosen metadata based on domain knowledge of multiphase flow, and we will outline directions for physics-informed extensions. We will also report the precise list of metadata features and their construction. revision: partial

  2. Referee: [Methods and results sections] Methods and results sections: The hybrid framework's outperformance over baselines and regret reduction is reported across sparsities, but without explicit details on how metadata features are constructed, normalized, or selected (and whether they were tuned post-hoc), it is difficult to assess if the gains are driven by genuine similarity capture or by the matrix completion component alone. A concrete test—such as reporting correlation between metadata distance and optimal-model overlap—would strengthen the central claim.

    Authors: We will expand the Methods section to provide full details on metadata feature construction, normalization, and selection criteria. These features were predefined from standard flow-condition descriptors used in multiphase CFD literature and were not tuned post-hoc on the validation results. To directly address the suggested test, we will add an analysis in the Results section reporting the correlation between metadata-based distances and the overlap of optimal model combinations across cases. This will help quantify the extent to which metadata similarity aligns with performance similarity. revision: yes

Circularity Check

0 steps flagged

No circularity: standard recommender application with independent CV evaluation

full rationale

The paper applies established matrix completion and metadata similarity techniques to a new domain (CFD closure selection) and evaluates via nested experiment-level holdout on held-out cases. No derivation reduces a claimed prediction to a fitted parameter or self-citation by construction; the regret and ranking metrics are computed directly against per-case optima on unseen data. The methodology is self-contained against external benchmarks and does not invoke load-bearing self-citations or ansatzes that loop back to the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based on the abstract alone, no explicit free parameters, axioms, or invented entities are described; the hybrid framework presumably employs standard similarity metrics and low-rank assumptions from matrix completion, but these are not itemized.

pith-pipeline@v0.9.0 · 5608 in / 1317 out tokens · 57450 ms · 2026-05-10T16:49:21.549938+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

67 extracted references · 33 canonical work pages

  1. [1]

    Investigation of drag models for simulation of gas-liquid two- phase flow systems

    Adzaklo, S.Y., Sarkodie, K., Agyei, E., Yamoah, S., Asiedu, N.Y., 2025. Investigation of drag models for simulation of gas-liquid two- phase flow systems. Scientific Reports 15, 39440. URL:https://www.nature.com/articles/s41598-025-23038-3, doi:10.1038/ s41598-025-23038-3

    2025

  2. [2]

    Regionalization-basedcollaborativefiltering:Harnessinggeographicalinformationinrecommenders

    Alves,R.,2024. Regionalization-basedcollaborativefiltering:Harnessinggeographicalinformationinrecommenders. ACMTransactionson Spatial Algorithms and Systems 10, 1–23

    2024

  3. [3]

    Analysis of phase distribution in fully developed laminar bubbly two-phase flow

    Antal, S.P., Lahey Jr, R., Flaherty, J., 1991. Analysis of phase distribution in fully developed laminar bubbly two-phase flow. International Journal of Multiphase Flow 17, 635–652

    1991

  4. [4]

    Turbulent bubbly two-phase flow in a triangular duct

    Lopez de Bertodano, M., 1992. Turbulent bubbly two-phase flow in a triangular duct. Ph.D. thesis. Rensselaer Polytechnic Institution

    1992

  5. [5]

    Computational fluid dynamics modelling of two-phase bubble columns: A comprehensive review

    Besagni, G., Varallo, N., Mereu, R., 2023. Computational fluid dynamics modelling of two-phase bubble columns: A comprehensive review. Fluids 8, 91. URL:https://www.mdpi.com/2311-5521/8/3/91, doi:10.3390/fluids8030091. open Access under CC BY 4.0

    work page doi:10.3390/fluids8030091 2023

  6. [6]

    The Favre averaged drag model for turbulent dispersion in Eulerian multi-phase flows, in: 5th International Conference on Multiphase Flow

    Burns, A., Frank, T., Hamill, I., Shi, J.M., 2004. The Favre averaged drag model for turbulent dispersion in Eulerian multi-phase flows, in: 5th International Conference on Multiphase Flow

    2004

  7. [7]

    Benchmarkingofcomputationalfluiddynamicmodelsforbubblyflows

    Colombo,M.,Rzehak,R.,Fairweather,M.,Liao,Y.,Lucas,D.,2021. Benchmarkingofcomputationalfluiddynamicmodelsforbubblyflows. Nuclear Engineering and Design 375, 111075. doi:https://doi.org/10.1016/j.nucengdes.2021.111075

    work page doi:10.1016/j.nucengdes.2021.111075 2021

  8. [8]

    Sommerfeld, M., Tsuji, Y., 2011

    Crowe, C., Schwarzkopf, J., M. Sommerfeld, M., Tsuji, Y., 2011. Multiphase Flows with Droplets and Particles

    2011

  9. [9]

    Large eddy simulation of the gas–liquid flow in a square cross-sectioned bubble column

    Deen, N.G., Solberg, T., Hjertager, B.H., 2001. Large eddy simulation of the gas–liquid flow in a square cross-sectioned bubble column. Chemical Engineering Science 56, 6341–6349. doi:10.1016/S0009-2509(01)00255-4

    work page doi:10.1016/s0009-2509(01)00255-4 2001

  10. [10]

    Fine-tuning a large language model for automating computational fluid dynamics simulations

    Dong, Z., Lu, Z., Yang, Y., 2025. Fine-tuning a large language model for automating computational fluid dynamics simulations. Theoretical and Applied Mechanics Letters 15

    2025

  11. [11]

    Practical Recommender Systems

    Falk, K., 2019. Practical Recommender Systems. Manning Publications / Simon & Schuster, New York, NY

    2019

  12. [12]

    Advanced Intelligent Discovery , year =

    Fan, E., Hu, K., Wu, Z., Ge, J., Miao, J., Zhang, Y., Sun, H., Wang, W., Zhang, T., 2026. Chatcfd: A large language model-driven agent for end-to-end computational fluid dynamics automation with structured knowledge and reasoning. Advanced Intelligent Discovery URL: https://doi.org/10.1002/aidi.202500174, doi:10.1002/aidi.202500174

    work page doi:10.1002/aidi.202500174 2026

  13. [13]

    Advances in computational fluid dynamics (CFD) of 3-dimensional gas-liquid multiphase flows, in: NAFEMS Seminar: Simulation of Complex Flows (CFD)-Applications and Trends, pp

    Frank, T., 2005. Advances in computational fluid dynamics (CFD) of 3-dimensional gas-liquid multiphase flows, in: NAFEMS Seminar: Simulation of Complex Flows (CFD)-Applications and Trends, pp. 1–18

    2005

  14. [14]

    Numerical methods for multiphase flows

    Garcia-Villalba, M., Colonius, T., Desjardins, O., Lucas, D., Mani, A., Marchisio, D., Matar, O.K., Picano, F., Zaleski, S., 2025. Numerical methods for multiphase flows. International Journal of Multiphase Flow 191, 105285. URL:https://www.sciencedirect.com/ science/article/pii/S0301932225001636, doi:10.1016/j.ijmultiphaseflow.2025.105285. S. Hänsch et a...

    work page doi:10.1016/j.ijmultiphaseflow.2025.105285 2025

  15. [15]

    Deep Learning

    Goodfellow, I., Bengio, Y., Courville, A., 2016. Deep Learning. MIT Press

    2016

  16. [16]

    Multidimensional modeling of turbulent two-phase flows in stirred vessels

    Gosman, A.D., Lekakou, C., Politis, S., Issa, R.I., Looney, M.K., 1992. Multidimensional modeling of turbulent two-phase flows in stirred vessels. AIChE Journal 38, 1946–1956

    1992

  17. [17]

    AcomparativestudyofclosurerelationsforCFDmodellingofbubblyflowinaverticalpipe

    Gray,G.S.,Ormiston,S.J.,2021. AcomparativestudyofclosurerelationsforCFDmodellingofbubblyflowinaverticalpipe. OpenJ.Fluid Dyn. 11, 98–134

    2021

  18. [18]

    Lyu, H., Meller, R., Schlegel, F., Tekavčič, M., 2025a

    Hänsch, S., Draw, M., Evdokimov, I., Khan, H., Krull, B., Lehnigk, R., Liao, Y., H. Lyu, H., Meller, R., Schlegel, F., Tekavčič, M., 2025a. Multiphase Cases Repository by HZDR for OpenFOAM Foundation Software.https://rodare.hzdr.de/record/3862. doi:10.14278/rodare.3862

    work page doi:10.14278/rodare.3862

  19. [19]

    A workflow for the sustainable development of closure models for bubbly flows

    Hänsch, S., Evdokimov, I., Schlegel, F., Lucas, D., 2021. A workflow for the sustainable development of closure models for bubbly flows. Chemical Engineering Science 244, 116807

    2021

  20. [20]

    Data-driven closure model selection for multiphase cfd via matrix completion

    Hänsch, S., Sajdoková, A., Rabau, A., Rybář, V., Alves, R., Kordík, P., 2025b. Data-driven closure model selection for multiphase cfd via matrix completion. AI Thermal Fluids 4, 100019. doi:https://doi.org/10.1016/j.aitf.2025.100019

    work page doi:10.1016/j.aitf.2025.100019 2025

  21. [21]

    Lift force coefficient of ellipsoidal single bubbles in water

    Hessenkemper, H., Ziegenhein, T., Rzehak, R., Lucas, D., Tomiyama, A., 2021. Lift force coefficient of ellipsoidal single bubbles in water. International Journal of Multiphase Flow 138, 103587

    2021

  22. [22]

    Axial interfacial area transport of vertical bubbly flows

    Hibiki, T., Ishii, M., Xiao, Z., 2001. Axial interfacial area transport of vertical bubbly flows. International Journal of Heat and Mass Transfer 44, 1869–1888. doi:10.1016/S0017-9310(00)00217-5

    work page doi:10.1016/s0017-9310(00)00217-5 2001

  23. [23]

    Multi-fluid simulation of turbulent bubbly pipe flows

    Hosokawa, S., Tomiyama, A., 2009. Multi-fluid simulation of turbulent bubbly pipe flows. Chemical Engineering Science 64, 5308–5318. doi:10.1016/j.ces.2009.08.018

    work page doi:10.1016/j.ces.2009.08.018 2009

  24. [24]

    Bubble-inducedpseudoturbulenceinlaminarpipeflows

    Hosokawa,S.,Tomiyama,A.,2013. Bubble-inducedpseudoturbulenceinlaminarpipeflows. InternationalJournalofHeatandFluidFlow40, 97–105. URL:http://www.sciencedirect.com/science/article/pii/S0142727X13000088,doi:https://doi.org/10.1016/ j.ijheatfluidflow.2013.01.004

    2013

  25. [25]

    Lateralmigrationofsinglebubblesduetothepresenceofwall.,in:ASME2002 Joint US-European Fluids Engineering Division Conference

    Hosokawa,S.,Tomiyama,A.,Misaki,S.,Hamada,T.,2002. Lateralmigrationofsinglebubblesduetothepresenceofwall.,in:ASME2002 Joint US-European Fluids Engineering Division Conference

    2002

  26. [26]

    Distilling experience into a physically interpretable recommender system for computational model selection

    Huang, X., Chyczewski, T., Xia, Z., Kunz, R., Yang, X., 2023. Distilling experience into a physically interpretable recommender system for computational model selection. Scientific Reports 13, 2225. doi:https://doi.org/10.1038/s41598-023-27426-5

    work page doi:10.1038/s41598-023-27426-5 2023

  27. [27]

    Thermo-Fluid Dynamics of Two-Phase Flow

    Ishii, M., Hibiki, T., 2011. Thermo-Fluid Dynamics of Two-Phase Flow. 2 ed., Springer, New York. doi:10.1007/978-1-4419-7985-8

    work page doi:10.1007/978-1-4419-7985-8 2011

  28. [28]

    Drag coefficient and relative velocity in bubbly, droplet or particulate flows

    Ishii, M., Zuber, N., 1979. Drag coefficient and relative velocity in bubbly, droplet or particulate flows. AIChE Journal 25, 843–855. URL: https://doi.org/10.1002/aic.690250513, doi:10.1002/aic.690250513

    work page doi:10.1002/aic.690250513 1979

  29. [29]

    Logistic matrix factorization for implicit feedback data

    Johnson, C.C., et al., 2014. Logistic matrix factorization for implicit feedback data. Advances in Neural Information Processing Systems 27, 1–9

    2014

  30. [30]

    Study of bubble-induced turbulence in upward laminar bubbly pipe flows measured with a two-phase particle image velocimetry

    Kim, M., Lee, J.H., Park, H., 2016. Study of bubble-induced turbulence in upward laminar bubbly pipe flows measured with a two-phase particle image velocimetry. Experiments in Fluids 57, 186. doi:10.1007/s00348-016-2272-3

    work page doi:10.1007/s00348-016-2272-3 2016

  31. [31]

    Orthogonalinductivematrixcompletion

    Ledent,A.,Alves,R.,Kloft,M.,2021. Orthogonalinductivematrixcompletion. IEEETransactionsonNeuralNetworksandLearningSystems 34, 2259–2270

    2021

  32. [32]

    The lift force on a spherical bubble in a viscous linear shear flow

    Legendre, D., Magnaudet, J., 1998. The lift force on a spherical bubble in a viscous linear shear flow. J. Fluid Mech. 368, 81 – 126

    1998

  33. [33]

    Sustainabledevelopmentofsimulationsetupsandaddons for openfoam for nuclear reactor safety research

    Lehnigk,R.,Bruschewski,M.,Huste,T.,Lucas,D.,,Rehm,M.,Schlegel,F.,2023. Sustainabledevelopmentofsimulationsetupsandaddons for openfoam for nuclear reactor safety research. Kerntechnik 88, pp. 131–140. doi:10.1515/kern-2022-0107

    work page doi:10.1515/kern-2022-0107 2023

  34. [34]

    Baselineclosuremodelfordispersedbubblyflow:Bubblecoalescenceandbreakup

    Liao,Y.,Rzehak,R.,Lucas,D.,Krepper,E.,2015. Baselineclosuremodelfordispersedbubblyflow:Bubblecoalescenceandbreakup. Chem- ical Engineering Science 122, 336 – 349. URL:http://www.sciencedirect.com/science/article/pii/S0009250914005466, doi:https://doi.org/10.1016/j.ces.2014.09.042

    work page doi:10.1016/j.ces.2014.09.042 2015

  35. [35]

    Eulerian-Eulerian two-fluid model for laminar bubbly pipe flows: Validation of the baseline model

    Liao, Y., Upadhyay, K., Schlegel, F., 2020. Eulerian-Eulerian two-fluid model for laminar bubbly pipe flows: Validation of the baseline model. Computers & Fluids 202, 104496. URL:http://www.sciencedirect.com/science/article/pii/S0045793020300694, doi:https://doi.org/10.1016/j.compfluid.2020.104496

    work page doi:10.1016/j.compfluid.2020.104496 2020

  36. [36]

    The role of bubble size on liquid phase turbulent structure in two-phase bubbly flow, in: Proceedings of the 3rd International Conference on Multiphase Flow (ICMF’98), Lyon, France

    Liu, T.J., 1998. The role of bubble size on liquid phase turbulent structure in two-phase bubbly flow, in: Proceedings of the 3rd International Conference on Multiphase Flow (ICMF’98), Lyon, France

    1998

  37. [37]

    Structure of air–water bubbly flow in a vertical pipe

    Liu, T.J., Bankoff, S.G., 1993. Structure of air–water bubbly flow in a vertical pipe. ii: Void fraction, bubble velocity and bubble size distribution. International Journal of Heat and Mass Transfer 36, 1061–1072. doi:10.1016/0017-9310(93)80029-8

    work page doi:10.1016/0017-9310(93)80029-8 1993

  38. [38]

    doi:10.1016/j.ijthermalsci.2009.11.006

    Lucas,D.,Beyer,M.,Szalinski,L.,Schütz,P.,2010.Anewdatabaseontheevolutionofair–waterflowsalongalargeverticalpipe.International Journal of Thermal Sciences 49, 664–674. doi:10.1016/j.ijthermalsci.2009.11.006

    work page doi:10.1016/j.ijthermalsci.2009.11.006 2010

  39. [39]

    Developmentofco-currentair–waterflowinaverticalpipe

    Lucas,D.,Krepper,E.,Prasser,H.M.,2005. Developmentofco-currentair–waterflowinaverticalpipe. InternationalJournalofMultiphase Flow 31, 1304–1328. doi:10.1016/j.ijmultiphaseflow.2005.07.002

    work page doi:10.1016/j.ijmultiphaseflow.2005.07.002 2005

  40. [40]

    A strategy for the qualification of multi-fluid approaches for nuclear reactor safety

    Lucas, D., Rzehak, R., Krepper, E., Ziegenhein, T., Liao, Y., Kriebitzsch, S., Apanasevich, P., 2016. A strategy for the qualification of multi-fluid approaches for nuclear reactor safety. Nuclear Engineering and Design 299, 2–11. doi:10.1016/j.nucengdes.2015.07.007

    work page doi:10.1016/j.nucengdes.2015.07.007 2016

  41. [41]

    Directnumericalsimulation-basedReynolds-averagedclosureforbubble- induced turbulence

    Ma,T.,Santarelli,C.,Ziegenhein,T.,Lucas,D.,Fröhlich,J.,2017. Directnumericalsimulation-basedReynolds-averagedclosureforbubble- induced turbulence. Physical Review Fluids 2

    2017

  42. [42]

    Spectral regularization algorithms for learning large incomplete matrices

    Mazumder, R., Hastie, T., Tibshirani, R., 2010. Spectral regularization algorithms for learning large incomplete matrices. The Journal of Machine Learning Research 11, 2287–2322

    2010

  43. [43]

    Ten years of industrial experience with the SST turbulence model

    Menter, F., Kuntz, M., Langtry, R., 2003. Ten years of industrial experience with the SST turbulence model. Turbulence, Heat and Mass Transfer 4, 625 – 632

    2003

  44. [44]

    Lateral forces on spheres in turbulent uniform shear flow

    Moraga, F.J., Bonetto, F.J., Lahey, R.T., 1999. Lateral forces on spheres in turbulent uniform shear flow. International Journal of Multiphase Flow 25, 1321 – 1372

    1999

  45. [45]

    Uniformflowinbubblecolumns

    Mudde,R.F.,Harteveld,W.K.,vandenAkker,H.E.A.,2009. Uniformflowinbubblecolumns. Industrial&EngineeringChemistryResearch 48, 148–158. doi:10.1021/ie800630v. S. Hänsch et al.:Preprint submitted to ElsevierPage 26 of 27 Recommender for Model Selection

    work page doi:10.1021/ie800630v 2009

  46. [46]

    Evaluating XAI: A Comparison of Rule-Based and Example-Based Ex- planations

    Mısır, M., Sebag, M., 2017. Alors: An algorithm recommender system. Artificial Intelligence 244, 291–314. doi:10.1016/j.artint. 2016.12.001

    work page doi:10.1016/j.artint 2017

  47. [47]

    Investigations on bubbly two-phase flow in a constricted vertical pipe

    Neumann-Kipping, M., Bieberle, A., Hampel, U., 2020. Investigations on bubbly two-phase flow in a constricted vertical pipe. International Journal of Multiphase Flow 130, 103340. doi:10.1016/j.ijmultiphaseflow.2020.103340

    work page doi:10.1016/j.ijmultiphaseflow.2020.103340 2020

  48. [48]

    Openfoamgpt: a rag-augmented llm agent for openfoam-based computational fluid dynamics

    Pandey, S., Xu, R., Wang, W., Chu, X., 2025. Openfoamgpt: a rag-augmented llm agent for openfoam-based computational fluid dynamics. Physics of Fluids 37

    2025

  49. [49]

    Scikit-learn: Machine learning in python

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., Duchesnay, É., 2011. Scikit-learn: Machine learning in python. Journal of Machine Learning Research 12, 2825–2830

    2011

  50. [50]

    Hydrodynamicsimulationsoflaboratoryscalebubblecolumns:Fundamentalstudies of the eulerian–eulerian modelling approach

    Pfleger,D.,Gomes,S.,Gilbert,N.,Wagner,H.G.,1999. Hydrodynamicsimulationsoflaboratoryscalebubblecolumns:Fundamentalstudies of the eulerian–eulerian modelling approach. Chemical Engineering Science 54, 5091–5099. doi:10.1016/S0009-2509(99)00279-2

    work page doi:10.1016/s0009-2509(99)00279-2 1999

  51. [51]

    Recommender systems: Techniques, applications, and challenges, in: Recommender Systems Handbook

    Ricci, F., Rokach, L., Shapira, B., 2022. Recommender systems: Techniques, applications, and challenges, in: Recommender Systems Handbook. 3rd ed.. Springer, New York, NY. doi:10.1007/978-1-0716-2197-4_1

    work page doi:10.1007/978-1-0716-2197-4_1 2022

  52. [52]

    Unified modeling of bubbly flows in pipes, bubble columns, and airlift columns

    Rzehak, R., Ziegenhein, T., Kriebitzsch, S., Krepper, E., Lucas, D., 2017. Unified modeling of bubbly flows in pipes, bubble columns, and airlift columns. Chemical Engineering Science 157, 147–158. URL:https://www.sciencedirect.com/science/article/pii/ S0009250916302226, doi:10.1016/j.ces.2016.04.056

    work page doi:10.1016/j.ces.2016.04.056 2017

  53. [53]

    Deep learning in music recommendation systems

    Schedl, M., 2019. Deep learning in music recommendation systems. Frontiers in Applied Mathematics and Statistics 5, 44

    2019

  54. [54]

    Multiphase Code Repository by HZDR for OpenFOAM Foundation Software.https: //rodare.hzdr.de/record/3284

    Schlegel, F., Bilde, K., Draw, M., Evdokimov, I., Hänsch, S., Kamble, V., Khan, H., Krull, B., Lehnigk, R., Li, J., Lyu., H., Meller, R., Petelin, G., Kota, S., Tekavčič, M., 2024. Multiphase Code Repository by HZDR for OpenFOAM Foundation Software.https: //rodare.hzdr.de/record/3284. doi:10.14278/rodare.3284

    work page doi:10.14278/rodare.3284 2024

  55. [55]

    Multiphase Python Repository by HZDR.https: //rodare.hzdr.de/record/3569

    Schlegel, F., Fombonne, C., Hänsch, S., Krull, B., Lehnigk, R., Meller, R., 2025. Multiphase Python Repository by HZDR.https: //rodare.hzdr.de/record/3569. doi:10.14278/rodare.3569

    work page doi:10.14278/rodare.3569 2025

  56. [56]

    Bubble and liquid turbulence characteristics of bubbly flow in a large diameter vertical pipe

    Shawkat, M.E., Ching, C.Y., Shoukri, M., 2008. Bubble and liquid turbulence characteristics of bubbly flow in a large diameter vertical pipe. International Journal of Multiphase Flow 34, 767–785. doi:10.1016/j.ijmultiphaseflow.2008.02.002

    work page doi:10.1016/j.ijmultiphaseflow.2008.02.002 2008

  57. [57]

    SnakemakeDevelopers,2025.Snakemake:Aframeworkforreproducibledataanalysis.URL:https://snakemake.github.io/.accessed: July 9, 2025

    2025

  58. [58]

    Hydrodynamics in a bubble column – part 1: Two-phase flow

    Sommer, A.E., Draw, M., Wang, L., Schmidtpeter, J., Hessenkemper, H., Gatter, J., Nam, H., Eckert, K., Rzehak, R., 2023. Hydrodynamics in a bubble column – part 1: Two-phase flow. Chemical Engineering & Technology 46, 1763–1772. doi:10.1002/ceat.202300130

    work page doi:10.1002/ceat.202300130 2023

  59. [59]

    Spišák, M., Alves, R., Kelleher, T., Sheppard, J., Fiedler, O., Kosovrasti, E., Vančura, V., Kasalick`y, P., Kordík, P., 2025. Segment-aware analyticsforreal-timeeditorialsupportinmediagroups,in:INRA2025:13thInternationalWorkshoponNewsRecommendationandAnalytics co-located with the 19th ACM Conference on Recommender Systems, Prague, Czech Republic, Septemb...

    2025

  60. [60]

    Structure of turbulent bubbly jets—ii

    Sun, T.Y., Faeth, G.M., 1986. Structure of turbulent bubbly jets—ii. International Journal of Multiphase Flow 12, 115–126. doi:10.1016/ 0301-9322(86)90010-4

    1986

  61. [61]

    Struggle with computational bubble dynamics

    Tomiyama, A., 1998. Struggle with computational bubble dynamics. Multiphase Science and Technology 10, 369–405

    1998

  62. [62]

    Transverse migration of single bubbles in simple shear flows

    Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002. Transverse migration of single bubbles in simple shear flows. Chemical Engineering Science 57, 1849 – 1858

    2002

  63. [63]

    3-dturbulencestructureandphasedistributionmeasurementsinbubblytwo-phase flows

    Wang,S.K.,Lee,S.J.,Jones,O.C.,J.,Lahey,R.T.,J.,1987. 3-dturbulencestructureandphasedistributionmeasurementsinbubblytwo-phase flows. International Journal of Multiphase Flow 13, 327–343. doi:10.1016/0301-9322(87)90024-0

    work page doi:10.1016/0301-9322(87)90024-0 1987

  64. [64]

    Computational Techniques for Multiphase Flows

    Yeoh, G.H., Tu, J., 2019. Computational Techniques for Multiphase Flows. 2 ed., Butterworth-Heinemann, Oxford

    2019

  65. [65]

    Deep learning based recommender system: A survey and new perspectives

    Zhang, S., et al., 2019. Deep learning based recommender system: A survey and new perspectives. ACM Computing Surveys

    2019

  66. [66]

    gcimpute:Apackageformissingdataimputation

    Zhao,Y.,Udell,M.,2024. gcimpute:Apackageformissingdataimputation. JournalofStatisticalSoftware108. doi:10.18637/jss.v108. i04

    work page doi:10.18637/jss.v108 2024

  67. [67]

    The critical bubble diameter of the lift force in technical and environmental, buoyancy-driven bubbly flows

    Ziegenhein, T., Lucas, D., 2019. The critical bubble diameter of the lift force in technical and environmental, buoyancy-driven bubbly flows. International Journal of Multiphase Flow 116, 26–38. doi:10.1016/j.ijmultiphaseflow.2019.03.002. S. Hänsch et al.:Preprint submitted to ElsevierPage 27 of 27 Recommender for Model Selection Figure 12:MRR@1results ac...

    work page doi:10.1016/j.ijmultiphaseflow.2019.03.002 2019