A sensory-calibrated electronic nose method for odor intensity quantification and grade identification in...

Refrigerator odors, arising from spoiled food and material off-gassing, can impair perceived freshness and consumer experience. To address the practical limitations of existing assessment methods in objectivity, speed, and alignment with human perception, this study developed a sensory-calibrated electronic nose (E-nose) method for quantifying odor intensity and identifying odor grades in refrigerators. Human sensory evaluation was first conducted on 30 odor samples from 10 representative foods, establishing a perceptual "ruler" with odor intensity values from 1 to >20 (as dimensionless dilution factors) and three distinct odor grades (weak, moderate, strong). From an initial 14-sensors array, an optimized two-sensor combination (S3 and S8) was then identified using a principal component analysis-discrimination index (PCA-DI) algorithm. Based on this optimized array, a multivariate regression model predicted odor intensity accurately (adjusted R² = 0.8377, RMSE = 1.03), while a Boosted Tree classifier achieved 93.6% ± 0.8% training and 92.0% ± 2.3% validation accuracy for odor grade identification. These results demonstrate that the proposed sensory-calibrated E-nose approach enables rapid, objective, and human-perception-aligned assessment of complex refrigerator odors. This work provides a practical methodological framework for intelligent odor monitoring and management.