Digital transformation of Jiangxiangxing Baijiu production: integrating flavor compound analysis, machine learning recognition, and genetic algorithm blending of multi-rounds

Abstract

The blending of Jiangxiangxing Baijiu is achieved through the specific proportional blending of base liquors from distinct distillation rounds, characterized by significant flavor variations attributed to differing volatile and non-volatile flavor compounds. To explore digital-assisted blending methods, sensory profiles of each round were first established to identify characteristic descriptors. Flavoromics analysis was then systematically applied to determine the types, concentrations, and variation patterns of flavor compounds. An 8-sensor electronic nose was built for multi-round Jiangxiangxing Baijiu. Based on flavoromics data, two intelligent models were innovatively developed: a machine learning-based recognition model utilizing electronic nose and chromatographic data for accurate round identification, and a genetic algorithm-based optimization model incorporating a multidimensional flavor compound dataset for intelligent blending. The optimized genetic algorithm model, validated by sensory evaluation and chromatographic analysis, was selected as the superior model for Jiangxiangxing Baijiu blending, providing a scientific basis for the digital transformation of traditional blending techniques.

Keywords: Baijiu blending; Electronic nose; Genetic algorithm; Jiangxiangxing baijiu; Machine learning; Non-volatile compounds; Volatile compounds.