遇见数据集

Open FEM Benchmark Database for CMUT Inverse Design (55,350 frequency–displacement samples, 369 design combinations)

收藏
Zenodo2026-07-10 更新2026-08-01 收录
官方服务:

资源简介:

# Open FEM Benchmark Database for CMUT Inverse Design **File:** `cmut_inverse_design_benchmark.csv` (55,350 rows) To the authors' knowledge, the first open FEM benchmark database forcapacitive micromachined ultrasonic transducer (CMUT) inverse design —the MEMS/ultrasonics counterpart of the photonic ceviche-challenges andinvrs-gym suites. ## Contents 369 unique thickness combinations of a unit-cell carbon-nanotube-implantedpiston-shaped CMUT, each mapped by COMSOL Multiphysics 6.2 parametric FEMsweeps to a 150-point frequency–displacement profile (2.3–12.15 MHz; peakdisplacements 0.013–0.084 µm). Columns: | Column | Meaning | Unit ||---|---|---|| `t_e` | top-electrode (Au) thickness | µm || `t_np` | nitride (Si3N4) passivation thickness | µm || `t_ox` | oxide (SiO2) thickness | µm || `t_w` | wall thickness | µm || `displacement_um` | membrane displacement amplitude | µm || `frequency_MHz` | excitation frequency | MHz | Fixed geometry: device length 63.5 µm, top/bottom electrode 33/35 µm,V = 5 V, p_max = 1 MPa (full geometry in the papers below). ## Recommended evaluation protocol Split **by design combination**, not by row: hold out ≥20% of the 369combinations entirely. Row-level splitting leaks within-combinationinterpolation and inflates R² (0.95 vs. 0.86 for the same surrogate — seethe PARL-ID paper, Sec. VI-A). ## Provenance and citation Generated for and first described in: > T. M. Mahim, M. M. Rahman, A.H.M.A. Rahim, "Hierarchical Inverse Design> Framework for Unit-Cell CMUTs With Attentive Gated Recurrent and Fully> Connected Dense Layers," IEEE Sensors Journal, 2025. Benchmark protocol, fabrication-corruption model and baseline results: > T. M. Mahim et al. "PARL-ID: A Fabrication-Aware> Physics-Informed Adjoint Framework With Reinforcement Learning in the Loop> for Inverse Design of CMUT and Photonic Sensors," (in preparation). Please cite both when using this dataset. ## License CC BY 4.0 (recommended — edit if you prefer another license before upload).

提供机构:
Zenodo
创建时间:
2026-07-10
二维码
社区交流群
二维码
科研交流群
商业服务