zNose™ technology for the classification of honey based on rapid aroma profiling

Rapid aroma profiling of food products is a first step towards at-line flavor quality control and off-flavor assessment. In this paper, the potential of the zNose™ was tested for the first time to address this application. Honey was chosen as the food product because of its characteristic aroma. Both a chromatogram and a spectral approach to the interpretation of the zNose™ signal were established. In the chromatogram approach, the signal was treated as a traditional chromatogram and relative peak areas were calculated and compared, while the whole aroma spectrum was considered in the spectral approach. Shifts in GC-column retention times initially led to misinterpretation of the results in the spectral approach. A data processing algorithm was, hence, developed to correct for these shifts. Data were analyzed with principal component analysis (PCA), and canonical discriminant analysis (CDA). With both relative peak areas and corrected spectra, the aroma of six different honey varieties and two types of sugar solutions were successfully discriminated. A classification model was built and validated externally, which resulted in a correct classification of 15 out of 16 honey aroma profiles (94%).