MODELOS FUZZY E NEURO-FUZZY PARA PREDIÇÃO DE VARIÁVEIS FISIOLÓGICAS EM OVINOS COM BASE EM DADOS MICROMETEOROLÓGICOS

The use of computational models and thermography enables automated systems to monitor thermal comfort in livestock. This study developed fuzzy and neuro-fuzzy models to predict physiological variables in sheep based on microclimatic variables. Microclimatic data were obtained using a black globe thermometer, while physiological data included rectal temperature and ocular, flank, and rib surface temperatures. A total of 396 data points were collected from 18 sheep during two daily periods. The models were created in MATLAB, using air temperature and humidity as inputs and physiological data as outputs. Performance evaluation considered statistics such as mean absolute error (MAE), mean absolute percentage error (MAPE), mean squared error (MSE), root mean squared error (RMSE), and Nash-Sutcliffe efficiency index (NSE). The neuro-fuzzy model outperformed the fuzzy model, although both delivered satisfactory results. These models eliminate invasive data collection, save time and human resources, and enable automation with greater precision in thermal stress analysis and animal welfare assessment. The approach proved efficient in estimating physiological variables, contributing to advancements in monitoring thermal comfort in sheep.