A nano-ZnS-decorated hierarchically porous carbon (ZSHPC) was mixed with MWCNTs to obtain ZSHPC/MWCNT nanocomposites. Then, ZSHPC/MWCNTs were used to modify a screen-printed electrode, and a portable electrochemical detection system combined with machine learning methods was used to investigate carbendazim (CBZ) residues in rice and tea. The electrochemical performance of the constructed electrode showed that the electrode had good electrocatalytic ability, large effective surface area, strong stability and anti-interference ability. Support Vector Machine (SVM), Least Square Support Vector Machine (LS-SVM) and Back Propagation-Artificial Neural Network (BP-ANN) were used to establish the prediction model for CBZ residues in rice and tea, and the traditional linear regression was developed. The investigated results showed that the LS-SVM model had the best prediction performance and the lowest prediction error compared with the traditional linear regression, BP-ANN and SVM models. The R², RMSE, and MAE for the training set samples were 0.9969, 0.3605 and 0.2968, respectively. The R², RMSE, MAE and RPD for the prediction set samples were 0.9924, 0.6190, 0.5360 and 10.3097, respectively. The average recovery range of CBZ in tea and rice was 98.77–109.32% and that of RSD was 0.47–2.58%, indicating that the rapid analysis of CBZ pesticide residues in agricultural products based on a portable electrochemical detection system combined with machine learning was feasible.

The surface-enhanced Raman spectroscopy (SERS) technique is being increasingly used for the detection of pesticide residues in agricultural products. However, there are large amounts of fluorescence-producing substances in agricultural products, which seriously affect the Raman signal of the analyte. In this paper, the QuEChERS method was used to remove interfering fluorescent substances in the analyte, and the purification effects of different doses of nano bamboo charcoal (NBC) and Fe₃O₄ magnetic nanoparticle (Fe₃O₄ MNP) adsorbents were studied. Meanwhile, the Raman spectral acquisition conditions (AuNPs, test solution, and NaCl) were optimized based on the orthogonal test method. The results showed that 300 µL AuNPs, 40 µL test solution, and 100 µL 1.5% NaCl gave the best SERS response effect. 12.5 mg NBC combined with 10 mg Fe₃O₄ MNPs could effectively remove the interfering substances from citrus. The Raman spectra of chlorpyrifos molecules were theoretically modeled using density-functional theory (DFT). By comparing the DFT results with the actual tests, five feature peaks, at 338, 522, 558, 672, and 1600 cm⁻¹, were obtained for the detection of chlorpyrifos pesticide residues in citrus. Based on the Raman feature peak intensity at 672 cm⁻¹, the concentration of chlorpyrifos in citrus showed a good linear relationship (R² = 0.9979) in the concentration range of 3–20 mg kg⁻¹. The recovery rate was 92.12% to 98.38%, and the relative standard deviation (RSD) was 1.77% to 5.29%. The lowest detection concentration was about 3 mg kg⁻¹, and the detection time of a single sample could be completed within 15 min. This study showed that the combination of SERS and QuEChERS preprocessing methods could achieve rapid detection of chlorpyrifos pesticide residues in citrus.

In this study, ultra-thin metal–organic framework nanosheets (MOFNs) were used for preparing an enhanced substrate to rapidly detect imidacloprid residues in citrus based on surface-enhanced Raman scattering (SERS) technology. The ultrasonic stripping method was used to prepare Co-MOFNs. Subsequently, a CoAg-MOFN-enhanced substrate was prepared by loading AgNPs onto the Co-MOFNs to enhance the SERS signal. The effects of the concentration of AgNPs, the ratio of Co-MOFNs to AgNPs, and the mixing time between the as-prepared enhanced base and imidacloprid standard solution on the SERS peak intensity were investigated, which showed that when the concentrated volume of AgNPs was 20 times and the mixed volume ratio of Co-MOFNs and AgNPs was 1 : 1, the prepared CoAg-MOFNs exhibited a good SERS enhanced effect. Notably, SERS response was the best when the CoAg-MOFN nanocomposite was mixed with an imidacloprid pesticide for 1 min. Utilizing density functional theory (DFT), four Raman characteristic peaks, specifically at 764, 784, 942 and 1035 cm⁻¹, were identified for the qualitative and quantitative assessment of imidacloprid pesticide residues in citrus fruits. A quantitative analysis curve representing imidacloprid residues was developed, correlating the intensity at 1035 cm⁻¹ over a range of 0.5–50 mg kg⁻¹ (y = 2.7964x + 26.211, R² = 0.9923), which demonstrated a strong linear relationship. The limit of detection (LOD) was 0.5 mg kg⁻¹ and the limit of quantitation (LOQ) was 0.5 mg kg⁻¹, which could meet the maximum residue limit (MRL) of 1 mg kg⁻¹ stipulated by the Codex Alimentarius Commission (CAC) and European Union (EU). The recovery rates were 93.02–101.07%, and the RSD was 2.57–7.71%, indicating that the prepared CoAg-MOFN nanocomposite combined with the SERS method is suitable for the analysis of imidacloprid pesticide residues in citrus.