RESOLUTION OF HIGHLY OVERLAPPING DIFFERENTIAL-PULSE ANODIC-STRIPPING VOLTAMMETRIC SIGNALS USING MULTICOMPONENT ANALYSIS AND NEURAL NETWORKS

This paper reports and discusses the results obtained by using multico mponent analysis methods based on multiple linear regression and neura l network procedures to resolve highly overlapping signals obtained by differential pulse anodic stripping voltammetry by using a static dro p electrode. The former procedures were applied to the well-known chem ical model composed of Pb(II), TI(I), In(III) and Cd(II) in binary, te rnary and quaternary mixtures. Different network architectures are inv estigated using the back propagation algorithm. Versatile software for data processing was developed. The proposed methodology was used to d etermine these four metals in tap water.