Study on Rapid Detection of Original Distillate of Strong-Aroma Baijiu Acid and Ester Component Concentration by Machine Learning Combined with Fourier Transform Infrared

Accurate detection of the key acid and ester components in the original distillate of strong-aroma Baijiu is crucial for Baijiu quality control and provides a scientific basis for blending and storage processes, traditional detection methods primarily rely on sensory evaluation combined with gas chromatography analysis. These methods are time-consuming and subject to operator bias. To address this, we propose a novel hybrid framework integrating Fourier transform infrared (FTIR) spectroscopy and machine learning for rapid detection. FTIR spectral of original distillate of strong-aroma Baijiu samples from different distillation periods were collected and preprocessed using Savitzky-Golay (S-G) smoothing, first-order derivative, and standard normal variate (SNV) transformation. Feature selection was carried out through the least absolute shrinkage and selection operator (LASSO) and competitive adaptive reweighted sampling (CARS). Following this, predictive models were constructed using partial least squares regression (PLSR), long short-term memory (LSTM) networks, and least squares boosting (LSBoost). The parameters of these models were optimized by the Grey Wolf Optimization (GWO) algorithm. The results show that the LASSO-GWO-LSTM model achieved the best performance in predicting the concentrations of ethyl acetate and lactic acid. For the ethyl acetate model, the root mean square error of prediction (RMSEP), R square of prediction (RP2), and relative percentage difference‌ of prediction (RPDP) reached 0.0967, 0.9949, and 14.6591, respectively. For the lactic acid prediction model, the corresponding values were 0.0152, 0.9805, and 7.0378. These results far exceed those of conventional PLSR models, indicating that the combination of FTIR and machine learning (LASSO-GWO-LSTM) is a promising method for real-time, high-precision, non-destructive analysis of acid and ester concentrations in the original distillate of strong-aroma Baijiu.