Utilização de redes neurais artificiais para determinação e avaliação de parâmetros que influenciam na absorção de água em carcaças de frangos

Abstract

With the increase in chicken meat production and consumption, consumers have become more demanding, with the quality of the meat becoming a fundamental factor. Among the concerns, the quality of the carcass and the excess water absorbed during the slaughter and cooling processes stand out. In this context, this study sought to measure the impacts of known parameters that can interfere with water absorption in the post-slaughter process of a slaughterhouse through data analysis using Python. The main objective was to develop a predictive model using artificial neural networks for water absorption in chicken carcasses, considering variables: scalding tank temperature 1 and 2; retention time in scalding tanks 1 and 2; temperature of the 1st, 2nd and 3rd plucking machine; initial carcass weight after bleeding; initial and final pre-chiller temperature; initial and final chiller temperature; pre-chiller retention time; total cooling time; chiller retention time; final weight of the carcass after the chiller. The methodology involved data collection, standardization, model training using neural networks, optimization, performance evaluation and analysis of the importance of variables (temperature of the 1st and 2nd scalding tanks; initial temperature of the pre-chiller; final temperature of the pre-chiller -chiller; plucking temperature 1, 2 and 3). The model achieved an impact of 0.004868 for the temperature of the 1st scalding tank and 0.014184 for the 2nd scalding tank, 0.007995 for the initial pre-chiller temperature, 0.004819 for the final pre-chiller temperature, 0.003780 for the plucking temperature 1, 0.003487 plucking temperature 2 and 0.004342 plucking temperature 3, highlighting the effectiveness in predicting absorption. It showed a significant correlation of 0.8913 for water absorption in chicken carcasses. Therefore, parameters related to the cooling stages demonstrated an impact on absorption. Although promising, the model requires continuous improvements due to the complexity of the poultry environment.