Improving the accuracy of stripping voltammetry detection of Cd2+ and Pb2+ in the presence of Cu2+ and Zn2+ by machine learning: Understanding and inhibiting the interactive interference among multiple heavy metals

The application of square-wave anodic stripping voltammetry (SWASV) for the accurate detection of Cd2+ and Pb2+ in soils presents tremendous challenges because of the poor anti-interference on the electrochemical interaction among multiple heavy metal ions (HMIs). To improve the SWASV detection accuracy, it is necessary to deeply understand the interactive interference among multiple ions and then propose an efficient method to inhibit the above interference. In this study, the two-dimensional correlation spectroscopy (2D-COS) method was employed to gain insight into the change degree and sequence in peak currents of various HMIs when subjected to the interference of Cu2+, Zn2+, Pb2+, and Cd2+. The 2D-COS results highlighted the severity and complexity of the interactive interferences that could not be comprehensively reflected by the limited information in stripping peak currents of HMIs. Therefore, the feature currents were mined, which contained abundant information about stripping voltammetry. Then, combining the feature currents with machine learning models, the study built the Feature-RF and Feature-SVR models that significantly improved the detection accuracies of Cd2+ and Pb2+ in the presence of Cu2+ and Zn2+. Finally, the proposed method was used to detect Cd2+ and Pb2+ concentrations in real soil extracts, yielding results close to those of the inductively coupled plasma mass spectrometry (ICP-MS) and recoveries close to 100%, validating its practicability. This study provides new insight into interactive interferences among multiple heavy metal ions in SWASV signals and a new method to improve the SWASV detection accuracy of HMIs in complex matrices.