A predictive model for the evaluation of flavor attributes of raw and cooked beef based on sensor array analyses

Abstract

There are currently no standardized objective measures to evaluate beef flavor attributes, especially the comparison between raw beef and cooked beef. Beef flavor attribute is one of the most significant parameters for consumers. This study described a predictive model using a 12-ion-sensor array and sensory properties to evaluate beef flavor attributes based on potential. Then the number of sensors was reduced to six via variance of analysis, and these six sensors were reserved with the saturated calomel reference electrode to constitute a new sensor array. Sensitive flavors of each sensor were selected through multiple comparative analysis. Results showed that the accuracy rate of classifying five basic flavors (acidity, sweetness, bitterness, saltiness, freshness) using the new sensor array was 100%. The processing methods used were based on multivariate statistical methods done with the cluster analysis (CA). Results were compared to sensory evaluation using genetic algorithm (GA). From GA, the accuracy rates of raw and cooked beef were 85.0% and 90.0%, which was consistent with the sensory analysis results. Moreover, reducing the number of sensors could decrease the data dimensionality and detection time. Also raw beef instead of cooked beef could be used in flavor attributes evaluation. This model could become an important method for evaluating beef flavor attributes repeatedly and objectively.

Keywords: Cluster analysis; Cooked beef; Flavor attributes; Genetic algorithm; Predictive model; Row beef; Sensor array.