Non-destructive determination of beef freshness based on colorimetric sensor array and multivariate analysis

Abstract

Rapid and reliable detection of beef freshness is essential for diet safety and resource-saving. In this study, the colorimetric sensor array (CSA) integrated with whale optimization algorithm (WOA) and back-propagation neural network (BPNN) had been innovatively developed for the quantitative determination of beef freshness, regarding the total volatile basic nitrogen (TVB-N) and total viable counts (TVC) contents. Firstly, the CSA comprising twelve color-sensitive dyes was designed to acquire the scent fingerprints (RGB triplets) during beef storage. Secondly, WOA-BPNN was used to optimize the color components combination from the preprocessed CSA to acquire the dominant color components. Finally, the BPNN models were constructed based on the optimized color components, with the BPNN topology of 8–12–1 for TVB-N prediction and 6–6–1 for TVC prediction. The results revealed that the BPNN model combined with optimized color components could be utilized to quantitatively determine beef freshness. The overall results demonstrated that the CSA integrated with appropriate multivariable analysis methods could realize rapid and reliable quantitative determination of beef freshness. Furthermore, the WOA-BPNN could effectively extract the dominant color components, which was beneficial for improving detection accuracy and robustness of the BPNN, as well as time- and cost-saving.