Odor Fingerprint-Based Real-Time Temperature Monitoring for Asphalt Mixtures via E-Nose and Drift Compensation

Temperature control during hot-mix asphalt production and paving is critical to construction quality and emissions. However, conventional thermometry is susceptible to dust and vibration. This study proposes an indirect temperature monitoring system using electronic nose technology, exploiting the nonlinear correlation between asphalt VOC odor fingerprints and temperature. The system integrates a MOS sensor array with IoT feedback, with a proof-of-concept study involving three asphalt types at five temperature levels. Leave-One-Out Cross-Validation (LOOCV) was employed to mitigate overfitting. The hybrid modeling framework significantly outperformed the Multiple Linear Regression (MLR) baseline, achieving 88.9% three-class accuracy and a regression RMSE of ±6.2℃. Drift compensation improved accuracy by 16.4%. These results show the feasibility of multi-dimensional odor patterns for quantitative temperature prediction, offering a new paradigm for closed-loop, non-contact temperature control in smart, low-carbon pavement engineering.