Fully automated solid-phase extraction and edge-deployed spectroscopy-based machine learning platform for on-site detection of PFOS/PFOA in industrial wastewater.

Abstract

Rapid on-site screening of perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) in industrial wastewater is hindered by matrix interference, laborious pretreatment, and reliance on laboratory instrumentation. Here, we developed a portable device integrating automated solid-phase extraction, fluorescence detection using poly(3-alkylthiophenes) (P3ATs), and principal component analysis-support vector machine (PCA-SVM) classification within a 30-min workflow. Performance was evaluated using samples from three fluorochemical wastewater streams. Automated extraction achieved efficiencies above 70% and 70-fold enrichment while reducing interference from coexisting phenolic compounds. Fluorescence detection showed concentration-dependent responses to both analytes, with greater sensitivity toward PFOS and workflow detection limits of 48 ng/L for PFOS and 78 ng/L for PFOA. Complete emission spectra acquired by the portable spectrometer enabled spectral fingerprint analysis, and PCA-SVM classified PFOS concentration ranges with 98% accuracy within the evaluated dataset. Embedded software supported on-site operation and real-time reporting without external computing resources. These findings demonstrate the feasibility of integrating sample preparation, fluorescence spectroscopy, and machine learning for rapid preliminary PFOS and PFOA screening in industrial wastewater. The device could support sample prioritization for confirmatory liquid chromatography-mass spectrometry analysis, although broader matrix applicability and PFOA classification performance require further validation.

Abstract as supplied by the indexing source. © the authors / publisher.