A Secondary Nondestructive Detection Method of Liquid Concentration for RFID Tag Array With Mutual Coupling

Radio Frequency Identification (RFID) technology emerges as a crucial tool for passive sensing due to the lightweight, taggable, and easily deployable attributes of tags. However, RFID multi-tag systems face significant accuracy degradation due to tag mutual coupling interference. To address this problem, we propose the RF-Copy method, a tagged coupling interference suppression approach. The method constructs RSSI and phase models with coupled terms, simplifies and estimates these parameters based on the models, and develops an interference suppression algorithm to remove coupling effects and extract ‘clean’ signal fingerprints. Experimental results demonstrate that the decoupled outputs (0.318 dB RSSI, 0.623 radians phase) closely approach the interference-free baseline values (0.392 dB, 0.649 radians). We apply the RF-Copy method to evaluate the tag mutual coupling suppression effectiveness in nondestructive white wine and wine concentration recognition. This process generates ‘clean’ RSSI and phase values, which serve as input features for training a Gated Recurrent Unit (GRU) neural network integrated with an attention mechanism. Experimental results demonstrate that our method achieves unopened-package detection accuracies of 99.0% (white wine) and 99.6% (wine). Compared to other methods, the RF-Copy-GRU not only improves sensing accuracy but also demonstrates robustness against interference.