High-precision quantitative analysis of 3-nitro-1,2,4-triazol-5-one (NTO) concentration based on ATR-FTIR spectroscopy and machine learning

3-Nitro-1,2,4-triazol-5-one (NTO) is a typical high-energy, low-sensitivity explosive, and accurate concentration monitoring is critical for crystallization process control. In this study, a high-precision quantitative analytical model for NTO concentration in ethanol solutions was developed by integrating real-time ATR-FTIR spectroscopy with chemometric and machine learning techniques. Dynamic spectral data were obtained by designing multi-concentration gradient heating-cooling cycle experiments, abnormal samples were eliminated using the isolation forest algorithm, and the effects of various preprocessing methods on model performance were systematically evaluated. The results show that partial least squares regression (PLSR) exhibits superior generalization ability compared to other models. Vibrational bands corresponding to C=O and –NO2 were identified as key predictors for concentration estimation. This work provides an efficient and reliable solution for real-time concentration monitoring during NTO crystallization and holds significant potential for process analytical applications in energetic material manufacturing.