中空圆柱形硅超材料传感器及其参数反演模型
By integrating deep learning with physical consistency constraints into a parameter inversion model, the problems of feature extraction mismatch and physical consistency in parameter inversion of hollow cylindrical QBIC sensors are solved, achieving efficient and accurate sensor parameter inversion and biomolecule detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA JILIANG UNIV
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning models suffer from problems such as feature extraction mismatch, lack of physical consistency, and poor generalization and robustness in parameter inversion of hollow cylindrical QBIC sensors. They are unable to capture both extremely narrow resonance peaks and global spectral features simultaneously, and the prediction results lack physical rationality.
A hollow cylindrical metamaterial sensor and its parameter inversion model were designed. By integrating deep learning and physical consistency constraints, ResNet50 and LSTM branches were used to extract features, and PINN physical head was combined for parameter inversion. This enabled the accurate capture of the resonance peaks and spectral temporal patterns unique to BIC. Constraints were imposed by a weighted function of loss function and physical consistency loss function.
It achieves efficient and accurate sensor parameter inversion, ensuring the consistency of prediction results in physical characteristics, improving the sensitivity and robustness of the sensor, and can accurately match the target spectral characteristics within a specific frequency band, making it suitable for highly sensitive biomolecule detection.
Smart Images
Figure CN122409565A_ABST