AI-enabled self-powered humidity sensors based on PVA/CNTs/Mg(NO3)2 composites for smart applications

This research introduces the creation and thorough assessment a flexible self-powered humidity sensor based on a PVA/CNTs/Mg(NO₃)₂ composite, designed in a sandwich configuration with aluminum and copper adhesive tape electrodes, synthesized via the freeze-thaw technique to enhance mechanical stability and flexibility. The sensor leverages the hydrophilic and degradable properties of polyvinyl alcohol (PVA), the conductive pathways of carbon nanotubes (CNTs), and the high ionic conductivity of magnesium nitrate (Mg(NO₃)₂) to achieve efficient moisture adsorption and ion migration, generating stable voltage outputs with range of 11–97 %. Material characterization through XRD, SEM, FTIR, and water contact angle measurements confirms the composite structural integrity, uniform morphology, and controlled hydrophilicity, while electrochemical analysis reveals a peak voltage of 0.6 V at 97 % RH and response/recovery times of 12 and 16 s, respectively, with minimal hysteresis (3.5 % RH) and sustained stability over 500 min. The sensor demonstrates recyclability through water dissolution of the composite within 180 min, albeit with a 12.1 % performance reduction (from 0.58 V to 0.51 V), and versatility in applications such as touch sensing, non-contact detection up to 8 mm, and differentiation of bare versus gloved hand interactions. Integration of machine learning, with the Gated Recurrent Unit (GRU) model achieving an MSE of 0.0001 and R² of 0.9991, enhances humidity prediction accuracy by leveraging the sensor dynamic voltage responses. These findings establish the PVA/CNTs/Mg(NO₃) sensor as a robust, eco-friendly platform for wearable and environmental monitoring, with significant implications for sustainable sensing technologies and opportunities for further optimization in recycling efficiency and predictive modeling.