Machine learning optimized design of THz piezoelectric perovskite-based biosensor for the detection of formalin in aqueous environments

This investigation presents the development and characterization of an advanced piezoelectric perovskite-based biosensing platform optimized for formalin detection in aqueous media through the implementation of Locally Weighted Linear Regression (LWLR) machine learning algorithms. The sensor architecture operates within the terahertz spectral region and incorporates an advanced nanomaterial composite system comprising black phosphorus, gold nanostructures, graphene, and barium titanate to maximize detection sensitivity and operational performance metrics. The engineered platform integrates a circular graphene metasurfaces configuration with a gold-based H-resonator assembly and concentrically arranged circular ring resonators. Computational simulations demonstrate vigorous sensing capabilities across three discrete frequency bands, achieving remarkable sensitivity parameters of 444 GHzRIU⁻¹, accompanied by a quality factor of 5.970 and detection accuracy of 7.576. The integration of LWLR-based optimization protocols substantially enhances prediction accuracy while reducing computational time by ≥ 85% as well as cutting down the required resources. The proposed sensor architecture presents significant potential for environmental monitoring and clinical applications, offering a highly sensitive and efficient methodology for quantitative formalin detection in aqueous environments.