Scholarlink

<p>This study presents a novel analytical methodology for the rapid prediction and optimization of taste profiles in liquid-smoked beef. Apple wood-derived liquid smoke was applied to beef samples under 46 different processing conditions, designed via response surface methodology (RSM). Taste attributes were quantitatively analyzed using an electronic tongue. Principal component analysis and K-means clustering of the E-tongue data identified four distinct taste types. Key processing parameters, liquid smoke concentration, immersion time, and heating temperature, were found to significantly influence these taste attributes. Among tested machine learning models, the random forest algorithm demonstrated comparatively stable performance in predicting taste classifications based on process variables. Optimization via RSM defined the optimal parameters for replicating traditional smoked taste as curing with 30% liquid smoke at 5 °C for 0.5 h, followed by heating at 180 °C for 10 min. The integration of E-tongue analysis with multivariate statistics and machine learning establishes a robust, data-driven framework for the rapid design and quality control of food flavor.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rapid Prediction and Optimization of Smoked Beef Taste Using Electronic Tongue, Multivariate Analysis, and Machine Learning

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Xing Xin,
  • Qiju Chen,
  • Jie Zhou,
  • Zhenyu Wang,
  • Lei Wang,
  • Shuai Yu,
  • Xiaohui Cui

摘要

This study presents a novel analytical methodology for the rapid prediction and optimization of taste profiles in liquid-smoked beef. Apple wood-derived liquid smoke was applied to beef samples under 46 different processing conditions, designed via response surface methodology (RSM). Taste attributes were quantitatively analyzed using an electronic tongue. Principal component analysis and K-means clustering of the E-tongue data identified four distinct taste types. Key processing parameters, liquid smoke concentration, immersion time, and heating temperature, were found to significantly influence these taste attributes. Among tested machine learning models, the random forest algorithm demonstrated comparatively stable performance in predicting taste classifications based on process variables. Optimization via RSM defined the optimal parameters for replicating traditional smoked taste as curing with 30% liquid smoke at 5 °C for 0.5 h, followed by heating at 180 °C for 10 min. The integration of E-tongue analysis with multivariate statistics and machine learning establishes a robust, data-driven framework for the rapid design and quality control of food flavor.