Learned resonance analysis separates stiffness and damping feature patterns in coated microcantilevers
Source check: Source checked. Not independent scientific confirmation.
Finding: Random-forest analysis of a large measured silicon-microcantilever dataset linked resonance frequency chiefly to effective modulus, thickness and geometry, while quality-factor variation emphasized coating thickness and effective modulus followed by surface descriptors. These are model-derived associations within the sampled wafers. Coating, roughness and material descriptors were correlated, so feature importance does not isolate causal dissipation mechanisms.
Technology: Random forest feature importance; extra-trees regression
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- RF learned nonlinear associations between engineered geometry/surface descriptors and experimentally measured resonance endpoints, enabling the reported contrast in predictive feature patterns. Extra trees modeled the same measured response space.
- Test and result
- The study fabricated and measured 12,060 cantilevers across six wafers with uncoated silicon and four metal coatings. Endpoint-specific filtering retained 10,450–11,429 measurements. RF nonlinear feature importance distinguished features associated with fundamental resonance frequency from those associated with quality factor. Twelve regression families were compared using random within-design splits, with a 70:30 split used for later interpolation analyses.
- Limitation / still unknown
- Surface roughness was assigned per coating material and strongly correlated with surface category; several physical descriptors were nominal rather than device-specific measurements. Random splitting within sampled wafers tests interpolation, not unseen-material or independent-wafer generalization. Feature-importance rankings do not establish causal mechanisms, independently optimized devices or universal material laws.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://onlinelibrary.wiley.com/doi/10.1002/adem.71360
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Read indexed original publisher Methods on measured endpoints, engineered features, RF importance and model families, plus feature-pattern Results, interpolation evaluation and limitations. Exact publisher front DOI/title/date matched current publisher-deposited Crossref canonical metadata; current relationship fields and indexed publisher notice context checked. Supplement and complete publisher HTML were not read..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.