Portable beef-freshness detection platform based on colorimetric sensor array technology and bionic algorithms for total volatile basic nitrogen (TVB-N) determination

Colorimetric sensor array (CSA) and bionic algorithms were integrated to form a facile platform for total volatile basic nitrogen (TVB-N) determination. First, a CSA containing twelve color-sensitive materials was prepared to obtain scent information of beef and generate scent fingerprints for visualization. Second, four bionic optimi-zation algorithms, ant colony optimization (ACO), particle swarm optimization (PSO), simulated annealing (SA), and whale optimization algorithm (WOA), were used to extract the characteristic fingerprint variables from the CSA. Finally, the back-propagation neural network (BPNN) model combined with characteristic color compo-nents was constructed to determine the TVB-N during beef storage, with improved precision, robustness, and generalization performance. The results demonstrated that WOA had the best optimization performance, fol-lowed by PSO, ACO, and SA. The WOA-BPNN optimized only two materials to detect TVB-N during beef storage. The BPNN constructed by three variables from the two selected materials had the best determination results, with the RMSEC, Rc, RMSEP, Rp, and RPD were 2.502 +/- 0.083 mg/100 g, 0.966 +/- 0.002, 2.903 +/- 0.143 mg/100 g, 0.952 +/- 0.006, and 3.430 +/- 0.185, respectively. Therefore, the WOA-BPNN model could realize high-precision quantitative determination of TVB-N during beef storage and save resources for CSA preparation. The combi-nation of CSA and the excellent bionic algorithm is expected to become a facile on-site sensing platform for food freshness monitoring.

Automated summary

In this study, a colorimetric sensor array (CSA) and bionic algorithms were combined to create a simple platform for determining total volatile basic nitrogen (TVB-N). The CSA collected scent information of beef and generated scent fingerprints for visualization. Bionic optimization algorithms were used to extract characteristic fingerprint variables from the CSA. A back-propagation neural network (BPNN) model combined with characteristic color components was constructed for TVB-N determination, showing improved precision and robustness. The WOA-BPNN model demonstrated high-precision quantitative determination of TVB-N during beef storage and potentially served as an efficient on-site sensing platform for food freshness monitoring.