An electronic nose combined with qualitative-quantitative two-stage hybrid modeling for microbial quantitative prediction in automotive air conditioners

Cabin air pollution caused by microorganisms in automobile air conditioning systems can pose a significant health risk to drivers and passengers. Therefore, it is crucial to assess microbial contamination in vehicle air conditioning systems. This study introduces a novel approach utilizing a qualitative-quantitative two-stage hybrid algorithmic model for rapid detection of microbial populations in vehicle air-conditioning systems using an electronic nose device. The methodology involves qualitative analysis of air conditioner filter samples with varying levels of microbial contamination, followed by quantitative microbial predictions based on the qualitative findings. The qualitative analysis model compared four classifiers (SVM, RF, RBF, and BPNN), with BPNN achieving the highest recognition accuracy of 99.35 % +/- 1.36 %. For the quantitative prediction models, ANN, RF, ELM, and SVM were evaluated for regression analysis, with SVM identified as the optimal regression model, achieving R2 values of 0.97 +/- 0.03 and 0.93 +/- 0.07 for the two sample types. The BPNN-SVM hybrid model proposed in this study has an R2 of 0.94 in the validation set and can accurately predict samples as low as 40 CFU/cm2. These findings demonstrate that the electronic nose method presented in this research can effectively and efficiently determine microbial contamination in vehicle air conditioning systems.