Application of Weibull analysis and artificial neural networks to predict the useful life of the vacuum packed soft cheese

The objective of this work was to evaluate the capability of artificial neural networks (ANN) to predict shelf life and the acidity on vacuum packed fresh cheese. First, cheese samples, of 200 g per unit, were prepared; then these samples were stored for 2 to 4 days at temperatures of 4, 10 and 16 degrees C and relative humidity of 67.5%. Throughout the storage, the acidity (AC) and sensorial acceptability were determined; this acceptability was used to determine the Shelf life time (SLT) by modified Weibull sensory risk method. A set of artificial neural networks (ANN) was created and trained; temperatures (T), maturation time (M) and failure possibility (F(x)) were used as inputs and SLT and AC as outputs. From this set, the networks with the lowest mean squared error (MSE) and best fit (R-2) were selected. These networks showed correlation coefficients (R2) of 0.9996 and 0.6897 for SLT and AC respectively, and good accuracy compared with regression models. It is shown that the ANN can be used to adequately model the SLT and, to a lesser degree, the AC of vacuum-packed fresh cheeses.