Modeling of a short-path distillation process to remove persistent organic pollutants in fish oil based on process parameters and quantitative structure properties relationships

A factorial experimental design based on temperature, feed rate and addition of "working fluid" (WF; fatty acid ethyl esters) was used to model a short-path distillation process applied for removal of persistent organic pollutants (POPs) in fish oil. Decontamination effect was assessed by measuring the level of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs), dioxin-like polychlorinated biphenyls (DL-PCBs) and polybrominated diphenylethers (PBDEs). The average reduction in chemical concentration of the individual PCDD, PCDF, DL-PCB and PBDE congeners was linearly dependent on the number of chlorine or bromine substitutions within each homologue group. DL-PCB and PBDE congeners could also be separated based on ortho-substitution. The quantitative structure properties relationships (QSPR) were combined with process parameters to establish response surface models for each homologue congener group based on partial least squares regression (PLSR). Cross validated predictive ability of the models was in the 4-9% range. Generally high temperature, low feed rate and WF addition improved the decontamination efficiency. The WHO-PCDD/F-PCB-TEQ level could be reduced by up to 98% based on the best experimental settings with residual concentration considerably below present maximum permitted levels in European food and feed legislations.