Scientific ML for aerodynamics: neural operators and neural fields for CFD surrogate modeling at Cenaero
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
This axis would measure publication volume and productivity over time — paper counts and output trajectory. The underlying bibliometric data is not yet available in this snapshot; only two reviewed arXiv preprints anchor this profile, so no output score can be computed.
Impact—
This axis would capture scholarly influence — citations, citation velocity, and uptake of methods by other groups. Citation data is not available in this snapshot, and with two 2026 preprints the work is too recent for impact to be assessed.
Focus0/100
The focus axis registers 0/100, while the raw field distribution splits evenly — 0.5 to scientific machine learning for aerodynamics (CFD surrogate modeling) and 0.5 to scientific machine learning / turbomachinery aerodynamics. Since these are two closely adjacent labels drawn from only two reviewed papers, the zero score should be read as a small-sample artifact rather than genuinely scattered interests; topically the work is tightly concentrated on neural surrogates for aerodynamics.
Collaboration41/100
The collaboration axis scores 41/100, with 7 distinct coauthors identified in the snapshot. Both catalog papers are first-authored (non-corresponding) and share a Cenaero-centered team; Tariq Benamara, Caroline Sainvitu, Joachim Dominique, Théo Flament, and Franck Mastrippolito recur in the co-author network. The moderate score reflects a compact, stable collaboration circle around a small two-paper sample rather than a broad international network.
Activity—
This axis would measure yearly publication cadence and recent research activity. The snapshot lacks time-series publication data beyond the two 2026 preprints, so activity level and its trend cannot be scored here.
Reproducibility55/100
The reproducibility axis scores 55/100, a mean computed over 2 assessed papers, both rated medium readiness. Evolve's verdicts note concrete limitations: the P-FNM study evaluates a single geometry with no external baselines, and the ONERA CRM winning entry used stepwise ablations but made leaderboard-driven selection choices. The medium ratings leave room for stronger artifacts such as released code, multi-geometry validation, and external comparisons.
Research summary
Lionel Salesses is a researcher at Cenaero (Belgium) working at the interface of scientific machine learning and computational aerodynamics. His reviewed output to date consists of two first-authored arXiv preprints from 2026, both cataloged and reviewed by Evolve. The first introduces periodic neural mapping (P-FNM), an extension of the Fourier Neural Operator that injects periodic time encoding to predict unsteady rotor-blade pressure and aeroelastic loads in turbomachinery, cutting pressure and generalized-aerodynamic-force errors against an internal VAE-RNN baseline. The second documents the winning solution to the ONERA CRM Wall Distribution 2025 Challenge: a conditional neural-field ensemble with Fourier feature encoding that scored 8.81 against an 8.64 baseline on the CRM-WBPN configuration. Together the two papers mirror his field distribution, which splits evenly (0.5/0.5) between scientific machine learning for CFD surrogate modeling and scientific machine learning for turbomachinery aerodynamics. He publishes within a compact team of 7 distinct coauthors — including Tariq Benamara, Caroline Sainvitu, Joachim Dominique, Théo Flament, and Franck Mastrippolito — consistent with a Cenaero-centered research group. Evolve rates both papers medium on reproducibility readiness, with a mean score of 55/100 across the pair. Citation, publication-volume, and yearly-activity metrics are not yet available in this snapshot, so impact and productivity remain to be established as the profile matures.
Collaboration pattern
The snapshot shows a compact, Cenaero-centered collaboration pattern: 7 distinct coauthors across the record, with Tariq Benamara (listed twice in the raw frequent-coauthor data), Caroline Sainvitu, Joachim Dominique, Théo Flament, and Franck Mastrippolito recurring. Both reviewed papers are first-authored but not corresponding, implying correspondence is typically handled by a senior colleague or group lead. The radar collaboration score is 41/100, consistent with a small stable team rather than an extensive network.
Research keywords
Highlights
Periodic neural mapping (P-FNM) for unsteady rotor-blade loads
Winning solution to the ONERA CRM Wall Distribution 2025 Challenge
Neural fields and implicit representations as aerodynamic surrogates
Timeline
- 2026First-authored arXiv preprint introduces P-FNM, adding periodic time encoding to the Fourier Neural Operator for unsteady rotor-blade pressure and aeroelastic load prediction.
- 2026First-authored arXiv report documents the winning solution (8.81 vs 8.64 baseline) to the ONERA CRM Wall Distribution 2025 Challenge using a conditional neural-field ensemble.
- 2026Both first-authored papers are reviewed into the Evolve catalog, each rated medium reproducibility readiness with a mean score of 55/100 across the pair.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric scores and the co-author count are derived from OpenAlex-based data; citation, h-index, and publication-volume metrics are absent from this snapshot.
- The profile rests on two 2026 arXiv preprints reviewed by Evolve; no journal or conference publication records are included, so career-stage inference is limited.
- Impact, output, and activity radar axes are missing because the underlying citation and yearly-productivity data were insufficient at capture time.
- Identity is name-resolved from public bibliographic data linked to ORCID 0000-0003-4061-9716; the raw frequent-coauthor list contained duplicate entries (including the author's own name), which were removed.