Wide-range quantitative detection of Dehydroacetic acid via terahertz metamaterial sensor integrated with support vector regression

Excessive intake of the food additive dehydroacetic acid (DHA) poses potential risks to human health, making rapid and precise detection of its residues essential for ensuring food safety. Addressing the limitations of complex sample pretreatment and restricted sensitivity in traditional detection methods, this paper proposes a terahertz metamaterial sensor based on an asymmetric single split-ring resonator (SRR) structure. Simulation and experimental results demonstrate a distinct resonance feature near 0.411 THz, with a theoretical refractive index sensitivity of up to 105 GHz/RIU. By integrating Terahertz Time-Domain Spectroscopy (THz-TDS), trace detection of DHA was achieved over a wide concentration range of 0.02 mg/L to 1000 mg/L. To address the non-linear characteristics of the resonance frequency shift with respect to concentration, a quantitative prediction model was established using the Support Vector Regression (SVR) algorithm. The results indicate high prediction accuracy, with a coefficient of determination (R 2 ) of 0.9602 and the Normalized Root Mean Square Error (NRMSE) of 5.1% for the testing set. The experimental Limit of Detection (LOD) is as low as 0.02 mg/L, representing a sensitivity enhancement of approximately 5 × 10 5 times compared to direct THz-TDS detection. This study validates the potential of terahertz metamaterial sensing technology combined with SVR algorithms for the rapid, low-cost, and high-precision quantitative detection of food additives.