Sensor Signal Decoupling Modeling Based on Hierarchical Cascade Residual Network

In-situ monitoring of equipment in harsh environments is crucial in various engineering fields. Traditional wired measurement methods are limited due to the highly enclosed and explosive nature of these environments, leading to the use of wireless passive sensors for monitoring. The coupling relationship between multi-parameter sensor signals makes data decoupling challenging. To improve the prediction accuracy of each variable's data, a temperature and humidity prediction method based on a hierarchical cascade residual neural network (HCRN) is proposed. This method addresses the introduction of prior information through hierarchical training and uses a residual structure in the humidity prediction model to enhance model performance. By collecting the resonant frequency under certain temperature and humidity conditions using a Complementary Split Ring Resonator (CSRR) sensor, this method models the data and verifies the model, achieving optimal multi-target prediction accuracy. Compared to extreme learning machines, support vector regression, and fully connected neural networks, this model performs the best.