An artificial neural network (ANN) based frequency controlled automated RF sensor system for the characterization of dispersive liquids is presented. The frequency of the designed structure is electronically controlled by varying the applied reverse-biased voltage to the complementary split ring resonator (CSRR) attached varactor diode. A hybrid modelling approach has mainly been devised to account for the effects of biasing circuitry and the parasitic elements. The designed RF sensor can provide a relatively higher tuning bandwidth (800 MHz) for applied DC biasing voltage of 0–25 V. The proposed sensor system employs an artificially intelligent feed-forward neural network architecture, which is trained by the Levenberg–Marquardt training algorithm for the complex permittivity estimation in the designated frequency band with reasonable accuracy. It employs the Bayesian Regularization for better input–output correlation and system performance. The ANN based characterization algorithm is integrated with a graphical user interface (GUI) and implemented using MATLAB.

提出了一种基于人工神经网络 (ANN) 的频率控制自动射频传感器系统，用于表征分散液体。设计结构的频率通过改变施加到互补裂环谐振器 (CSRR) 变容二极管的反向偏置电压来进行电子控制。主要设计了一种混合建模方法来考虑偏置电路和寄生元件的影响。设计的 RF 传感器可以为施加的 0-25 V 直流偏置电压提供相对较高的调谐带宽（800 MHz）。所提出的传感器系统采用人工智能前馈神经网络架构，该架构由 Levenberg-Marquardt 训练进行训练在指定频带内以合理的精度计算复介电常数的算法。它采用贝叶斯正则化以获得更好的输入-输出相关性和系统性能。基于 ANN 的表征算法与图形用户界面 (GUI) 集成并使用 MATLAB 实现。